Discussion on the possibility of multi-layer intelligent technologies to achieve the best recover of musculoskeletal injuries: Smart materials, variable structures, and intelligent therapeutic planning

Abstract

Although intelligent technologies has facilitated the development of precise orthopaedic, simple internal fixation, ligament reconstruction or arthroplasty can only relieve pain of patients in short-term. To achieve the best recover of musculoskeletal injuries, three bottlenecks must be broken through, which includes scientific path planning, bioactive implants and personalized surgical channels building. As scientific surgical path can be planned and built by through AI technology, 4D printing technology can make more bioactive implants be manufactured, and variable structures can establish personalized channels precisely, it is possible to achieve satisfied and effective musculoskeletal injury recovery with the progress of multi-layer intelligent technologies (MLIT).

1 Introduction

FIGURE 1

As shown in Figure 2, similarly to the construction process of a building, MLIT for musculoskeletal injuries consists of three stages: scheme design, scheme implementation, and enhancements (Table 1).

FIGURE 2

2 Status of multi-layer intelligent technologies

2.1 Scientific surgical path can be planned and built by through AI technology

Surgical placement and appropriate implant fixation had equal importance as the inherent implant characteristics in maintaining long-term implant stability (Li et al., 2019). Biomechanical surgical path must be planned on basis of the pathogenesis of musculoskeletal diseases, to reach the satisfied musculoskeletal injury recovery. This requires physical parameters to be predicted precisely, to establish build kinesiology models of musculoskeletal system. In the past decade, AI technologies has made dramatic advances on Orthopaedic in the following four aspects (Zhou et al., 2020a; Lopez et al., 2021a) (Figure 3): automatic reconstruction, physical parameters predicting, preoperative planning and intraoperative images registration.

FIGURE 3

2.1.1 Automatic reconstruction algorithm

In the early study, threshold segmentation or seed growth combined with manual repair are often used to segment medical images like CT, MRI etc. This semi-automatic method is very inefficient. AI technology, especially the U-net (Li et al., 2017; Falk et al., 2019), has promoted the rapid development of medical image segmentation. In general, variants of the U-net can be divided into four categories (Figure 4): ① Design encoder or decode structures. Cascade decoder conducted more effective decoding of hierarchically encoded features (Liang et al., 2019). The encoder of TernausNet removed the fully connected layers and replace them with a single convolutional layer of 512 channels that serves as a bottleneck central part of the network, and the decoder of TernausNet transposed convolutions layers that doubles thesize of a feature map (Iglovikov and Shvets, 2018). TernausNet was helpful to prevent over-fitting. ②Optimize connection modes of encoding module and decoding module. MNet proposed a skip connection method to balance the spatial representation inter axes via learning (Dong et al., 2022); UNet++ redesigned skip connections to exploit multiscale features in image segmentation (Zhou et al., 2020b).③ Set new loss function. Loss function is a method to measure the quality of model prediction, which is of course important to an AI model. Dice loss is a common loss function for medical image segmentation. However, it may cause oscillation during training when the prediction is close to ground truth (Chen et al., 2018b). Cos-Dice loss function was used in W-net to make the network more stable (Chen et al., 2018b). The adjustable penalty weights of the misclassified voxels were used in dice coefficient to adapt to unbalanced class frequency (Huang et al., 2018). Loss functions based on the Tversky index (Salehi et al., 2017; Abraham and Khan, 2018; Huang et al., 2018; Das and Zhang, 2020)were used to address the issue of data imbalance. ④Import attention mechanism. Swin Unet was a semantic segmentation of brain tumors in MRI Images using a swin transformer encoder which can extract features at five different resolutions by utilizing shifted windows for computing self-attention (Hatamizadeh et al., 2021). RA-UNet proposed a 3D hybrid residual attention-aware segmentation method to precisely extract the liver volume of interests (VOI) and segment tumors from the liver VOI (Jin et al., 2018). TransUNet was a variant of U-net using a transformer encodes which tokenized image patches from a convolution neural network (CNN) feature map as the input sequence for extracting global contexts (Chen et al., 2021). UNEt TRansformers (UNETR) utilized a transformer as the encoder to learn sequence representations of the input volume and effectively capture the global multi-scale information (Hatamizadeh et al., 2022). Particularly in 2021, the nn-Unet proposed by Isensee F. et al. Isensee et al. (2021) at Heidelberg University in Germany surpasses most existing approaches on 23 public datasets used in international biomedical segmentation competitions. Variants of the U-net or other deep learning algorithms are widely used in segmentations for pelvic, spine, femur, knee arthroscopy and other orthopedics fields (Zeng et al., 2017; Balagopal et al., 2018; Kolařík et al., 2019; Isensee et al., 2021).

FIGURE 4

2.1.2 Physical parameters predicting method

Extraordinarily important is that AI technology perform well in predicting physical parameters of organizations. In 2019, a review by Rogers M.A. Rogers and Aikawa (2019) of Harvard University in the United States showed that AI technology can analyze the cardiovascular calcification on quantitatively large data sets. In 2020, Black K.M. et al. Black et al. (2020) of the University of Michigan automatically detected kidney stones composition from digital photographs of stones by ResNet-10, and precisions for each stone type were above 75%. In 2021, the study of Lopez F. et al. Lopez et al. (2021b) showed that the DCCN approach is the best method with a precision of 98% for four kidney stones classification with in-vivo endoscopic images. In 2021, Molnar D. Molnar et al. (2021) et al. of Gothenburg in Sweden proposed a Crop-Net for fat prediction, with a precision of 99.4%.

In the field of orthopedics, AI technology can predict bone density accurately (Figure 5). Hsieh C.I. from Chang Gung Hospital in Taiwan (Hsieh et al., 2021) proposed the Dual-energy X-ray A bsorptiometry (DXA)to predict bone mineral density and fracture risk, and the accuracy of hip osteoporosis prediction reached 95%. In 2020, the BMDCNN is applied to predict the bone mineral density (BMD) of the lumbar spine by Yasaka K. Yasaka et al. (2020) from the University of Tokyo, and the internal and external AUC of osteoporosis AUC 0.965 and 0.970.

FIGURE 5

2.1.3 Preoperative planning

Preoperative planning is an essential part of Clinical Decision Support System(CDSS). Screw placement position (Cai et al., 2019), implants size (Dong et al., 2021; Polce et al., 2021), osteotomy morphology are three common preoperative planning needs in Orthopaedic surgery (Figure 6). Target reconstructions, landmark/anatomical components recognization (Cai et al., 2019; Siemionow et al., 2021), physical parameters (bone mineral density (Caprara et al., 2021), morphological parameters (Soodmand et al., 2019)), finite element analysis (Zheng et al., 2018a; Caprara, 2021), personal information and other risk constraints may be possible inputs for preoperative planning. AI technologies can play a role in the processing of planning inputs. AI applications in bone reconstructions and physical parameters predictions have been discussed in Section 2.1.1 and 2.1.2. It is well-attended to predict how bone adapts to different loads in surgical planning (Jordi et al., 2022). FE is commonly performed in biomechanics analysis. It is necessary to build statistical shape models (SSM) based on a set of landmark points (Zheng and Yu, 2017) for FE analysis. An SSM built by the training dataset is a set of annotated images (Zheng and Yu, 2017). AI technologies make the landmarks/anatomical components recognization automatically (Siemionow et al., 2021). Nathan proposed a deep learning method based on the PointNet++ architecture for biomechanical modeling of facial tissue deformation in orthognathic surgical planning (Lampen et al., 2022). Xiao Xiao et al. (2021) proposed an estimating Reference Bony Shape Models for Orthognathic surgical planning using 3D point-cloud deep learning. Due to the incompleteness of information and uncertainty of healing prediction, preoperative planning dose not have a rapid development like reconstruction technologies.

FIGURE 6

2.1.4 Intraoperative registration

The accuracy of robotic assisted surgery mainly depends on the space registration. The 2D-3D registration is commonly used in spinal surgery or pelvic surgery and the 3D-3D registration is commonly used in TKA (total knee arthroplasty) surgery. There are always tow workflows of 2D-3D registration (Figure 7): 1) Projection based 2D-3D registration. Digitally reconstructed radiographs (DRR) are generated from CT images, and then mutual information, normalized cross correlation, sum of square differences or other similarity measurements are used to calculate the relationship between X‐ray images and DRRs (Guo et al., 2019). 2) Reconstruction based 2D-3D registration. 3D point clouds are reconstructed from 2D x-ray or ultrasonic images. Transformation matrix are calculated through tow 3D point clouds (Wang et al., 2018). The intraoperative point clouds of knee are commonly collected by probes with markers in TKA surgery. The preoperative and intraoperative point clouds are matched through ICP or CPD algorithms.

FIGURE 7

Due to the sparsity inconsistency of point clouds between preoperative images and intraoperative images, deformation differences caused by different parameters of different imaging devices, the accuracy of surgical registration methods has been paid a great attention, but not yet been well solved. Registration methods based on deep-learning (Lu et al., 2021) have shown be capable of addressing the limitations of conventional registration methods (Figure 7), as they have a better performance in predicting deformations, learning 3D point clouds features, generating DRRs and optimizing registration strategy. Unsupervised learning (Chen et al., 2022) estimate voxel-to-voxel deformable transformation by maximizing image similarity (Fan et al., 2019). Weakly-supervised regularization loss is able to map thecomplex appearance to a common space (Blendowski et al., 2021). Self-supervised anatomical embedding (SAM) (Yan et al., 2022) is capable of computing dense anatomical/semantic correspondences between two images at the pixel level. Feature learning methods use the deep neural network to learn a robust feature correspondence search (Huang et al., 2021a), which includes a local PPF feature using the distribution of neighbour points (Deng et al., 2018), a rotation-invariant hand-craft feature (Gojcic et al., 2019), a global features (Qi et al., 2017) etc. There are three registration strategies using AI (Huang et al., 2021a): ICP-based variations, graph-based, GMM-based and semi-definite registration methods. Deep Closest Point(DCP) uses deep features to estimate correspondences to avoid spurious local optima of most ICP algorithms (Wang and Solomon, 2019). The surface registration are solved effectively by transforming the registration problem into a graph matching problem (Le-Huu and Paragios, 2017). For Gaussian mixture models (GMM), DeepGMR (Yuan et al., 2020) uses a neural network to learn pose-invariant point-to-distribution parameter correspondences. DRRs are simulated by algorithm which may differ from the real X-ray images. To improve the quality of DRRs, GAN-based training, material decomposition, and incremental learning are proposed. A GAN-based disentanglement learning framework (Han et al., 2021) can transfer the rib structural priors from DRRs. DeepDRR is a framework for DRRs embedding material decomposition and scatter estimation in 3D and 2D, combined with analytic forward projection and noise injection (Unberath et al., 2018). Incremental learning can improve continuously acquire new knowledge by continuously acquiring new knowledge (Jiang et al., 2021).

2.2 Responsive bioactive implants made through 3D printing technology can provide signals for endogenous repair

Autograft, allografts and metal-based material scaffolds are commonly used in the clinical treatment of bone defects. However, chronic inflammation, immune rejection or stress shield and inflammation have hindered their clinical application. New bioactive responsive implants made through 3D printing technology and adaptively expanded, which can provide signals for endogenous repair, is highly necessary to promote bone regeneration (Figure 8). Bioactive materials, especially bioactive polymers high osteoinduction, excellent angiogenesis, biocompatibility and unlimited size, can interact with proteins, cells or tissues in vivo and cause biological reactions.

FIGURE 8

2.2.1 Responsive bioactive materials can promote the reconstruction of organizations

Bioactive materials can guide bone regeneration (Mistry et al., 2015) (Figure 9). Dhivya S.et al. Dhivya et al. (2015) explored a nanoparticulate mineralized collagen glycosaminoglycan scaffold that induces healing of critical-sized rabbit cranial defects. These inorganic nano materials delivery of bioactive agents. However, their widespread employment may be reduced due to the possible toxicity and the lack of biodegradability (Makvandi et al., 2020).

FIGURE 9

Polymer materials consisting of inorganic nanomaterials and organic polymer materials, with high ductility, biocompatibility and biodegradability, are widely used in bone repair scaffolds (Figure 11). Their specific structure and surface properties can be specifically recognized and interact with target biomolecules. Makvandi P.et al. Makvandi et al. (2020) developed a graphene oxide (GO)–Chitosan (CS)–Hyaluronic acid (HA) based bioactive composite scaffold containing an osteogenesis-inducing drug simvastatin (SV). The elongated morphology of the cells after 48 h incubation has demonstrated the osteoblast migration using the SV loaded GO–CS–HA scaffolds. A collagen–hydroxyapatite(HA) scaffold (Unnithan et al., 2017) with a degree of interconnectivity of 99%, can not only deliver biological factors, promote stem cell differentiation and ossification (Thitiset et al., 2013; Villa et al., 2015), but also theoretically be applied in a load bearing application when combined with mechanical fixation (Zhang et al., 2018). Mouse BMSCs isolated from the femur and tibia of CD1 wild type animals, were seeded dropwise on to the top of either the Col–HA scaffold. Cells well were attached on the scaffolds after 12 h of seeding. In the study of in vivo bone formation in a mouse calvaria defect, scaffolds were implanted. After 3 weeks of implantation, calvaria containing critical size defects filled with BMSCs combined with either the Col–HA scaffold in X-ray images (Unnithan et al., 2017). The research of Ren X. et al. Ren et al. (2016) demonstrated that a nanoparticulate mineralized collagen glycosaminoglycan scaffold showed more efficient mineralization of MC-GAG scaffolds than non-mineralized Col-GAG scaffolds in either the histologic analyses or mCT scanning images. Wu H. et al. Wu et al. (2017) has fabricated a chitosan-based microsphere delivery system to controlled release of alendronate (AL), which can release AL for up to 30 days. CM-ALs (10%) scaffolds showed better performance in large-sized bone defects repairs than CM-ALs (0%). The new bone increasing ratio (NBIR) of HLA/HA-βTCP samples was 1.78 times higher than the blank group at week 2 (Chang et al., 2016). The study of Roohaniesfahani S.I. et al. Roohani-Esfahani et al. (2012) showed that, BCP/PCL scaffolds with silk layer is more favorable than BCP/PCL scaffolds with collagen layer in mechanical properties and biological properties.

More importantly, responsive bioactive materials, especially shape memory polymers (SMP) (Figure 10), can remember the temporary shape and return to the original shape under the condition of the external stimuli, like heat, pH, electricity, magnetic field, etc (Zhang et al., 2012; Rezwan et al., 2006; Zheng et al., 2018b). This specific shape memory could simplify complex transplant procedures, with excellent chemical stability, biocompatibility and biodegradability, which can stimulate specific cellular responses. Deng Z.et al. Deng et al. (2016) designed and synthesized a series of shape memory copolymers with electroactivity, super stretchability and tunable recovery temperature based on poly(e-caprolactone) (PCL) with different molecular weight and conductive amino capped aniline trimer. They proved that they can enhance myogenic differentiation from C2C12 myoblast cells. Xie R.et al. Xie et al. (2017) prepared a novel polyurethane or hydroxyapatite based SMP porous foam for the treatment of load-bearing bone defects by gas foaming. The foam can match the trabecular bone, possess the feasibility of minimally invasive delivery. And it can also overcome the disadvantages of traditional polymer foams in terms of insufficient mechanical properties, inadequate pore structures, low biocompatibility and inconvenience in operation. The rabbit femoral defect model (Xie et al., 2018) demonstrated that a SMP foam bone scaffold could play an important role in promoting neovascularization and bone remodeling. A shape memory PCL porous scaffold was made from photocrosslinking (ε-caprolactone) (PCL) polydiacrylate through SCPL method (Zhang et al., 2014), and it evidently promoted the adhesion and proliferation of osteoblasts. SMP scaffolds show strong self adaptability, as evidenced by the following facts. One of these is that smaller SMP scaffolds than bone defect size can match bone defect boundary after shape expansion (Zhang et al., 2012; Xie et al., 2017), the other is that larger SMP scaffolds can promote the inward growth of bone, due to the binding force with bone tissue (Zhang et al., 2014; Bao et al., 2016).

FIGURE 10

2.2.2 3D bioprinting technology can personalize implants

Three-dimensional printing technologies that can fabricate the microstructure of materials precisely, have shown distinct advantages to personalize implants in bone tissue engineering (Luo et al., 2015). In clinical medicine, 3D printing technologies can be divided into two stages (Figure 11): one is conventional 3D printing objects without cellular information exchange, like surgical guiding paltes (Butscher et al., 2013), prosthesis (Xu et al., 2013; Turnbull et al., 2018); the other is bioactive 3D printing objects can provide signals for endogenous repair, which particularly is of great significance to bone regeneration and vascular reconstruction, such as degradable scaffolds (Xu et al., 2019a), living cell printing (Kolesky et al., 2014), and stimuli-responsive printing materials with programmable behavior (Gladman et al., 2016) (Qi et al., 2014).

FIGURE 11

Degradable 3D printing scaffolds can gradually release the occupied space with the recovery of the damaged part (Figure 7.C) (Li et al., 2020). Kawai T. et al. Kawai et al. (2018) of Stanford University designed and 3D printed a functionally graded scaffold (FGS) made of polycaprolactone (PCL) and b-tricalcium phosphate (b-TCP) to treat Osteonecrosis of the femoral head. The in vivo degradation rate and the bone ingrowth ratio of the scaffold is significantly higher than the empty-tunnel group. Li et al. Li et al. (2019) at University of Sydney developed a 3D-printed Sr-HT-Gahnite scaffolds implanted into critical-sized segmental defects in sheep tibia. Compared with bone autografts, the scaffolds possessing both osteoconductive and osteoinductive properties, can induce substantial bone formation and defect bridging. Spiral fractures were observed in the study. The fractures may have a negative effect on the recovery. Thus, implants should be fixed at appropriate position.

2.2.3 Bio implants through bio-inks can migrate in planned direction

Bioinks is a formulation of cells suitable for processing by an automated biofabrication technology that may also contain biologically active components and biomaterials (Groll et al., 2018). Biomaterials may consist of cells, collagen, and other bioactive ingredients; Auxiliary biomaterials, such as gelatin, alginate hydrogel, carbomer glue, etc, can be used to improve mechanical strength of implants, maintain the shape of the printed object, and ensure the adhesion and survival rate of cells. In 2015, Kang H.W. et al. Kang et al. (2016) of the Wake Forest Institute for Regenerative Medicine developed a system that deposits cell-laden hydrogels together with synthetic biodegradable polymers that impart mechanical strength. The printed mandible, skull, cartilage and skeletal muscle can incorporate multiple cell types at precise locations to recapitulate native structure and function. In 2016, Zhang Y. S. Zhang et al. (2016) of Harvard Medical School proposed a novel hybrid strategy based on 3D bioprinting to fabricate endothelializedmyocardium. Endothelial cells controlled anisotropy, can gradually migrate towards the peripheries of the microfibers.

With the continuous upgrading of material technology, biological 3D printing strategy is constantly updated. In 2019, Van Belleghem S.et al. Van Belleghem et al. (2020), from the University of Maryland College Park, proposed a 3D printing strategy for dual bioinks. The graft consisting of both degradable and nondegradable parts, providing long term mechanical integrity and shape retention. A pre-programmed responsive bioactive materials can be used in various application where the human intervention is not possible. This technology may lead to satisfied repair. The materials which responds to external stimuli is called 4D printing (Tibbits, 2014). A porous PLA/HAP scaffolds can stand up to three compression-heating-compression cycles without delamination. The scaffolds can narrow the cracks during heating, which may resulted in ‘self-healing’. The significant changes of post implantation have higher requirements to ensure the stability during deformation process of the implants, and that the deformation not cause additional damage to the surrounding tissues.

2.3 Personalized surgical channels can be established through intelligent robotics precisely

In clinical, there are linear channels established, due to the limitation of the structure of instruments (Nicolas et al., 2016; Zhang et al., 2019). Fracture or scoliosis treatments are great different between children and the elderly, because children have better self-healing in growth. Linear channels may be not meet the biomechanics needs, especially for the elderly. As smart materials and variable structures developed, such as memory alloys, concentric tube structures, etc., curved-trajectory, with less trauma and better therapeutic effect, attracts the researchers’ great attention (Figure 12). Moreover, intelligent sensing technology and intelligent operation technology also make the surgery more efficiency.

FIGURE 12

2.3.1 Personalized surgical channels can be established through variable structures

Personalized surgical channels are linear channels, curved- channels or linear- curved mixed channels based on the status of patients. In 2011, Watanabe H. Watanabe et al. (2011) et al. developed a steerable drill for ACL reconstruction to construct an arbitrary trajectory of a bone tunnel. Gilbert H. Gilbert et al. (2014) adopted a needle-sized tentacle-like robot that require access through constrained paths in transnasal skull base surgery. In 2012, Gosline, A. H. et al. Gosline et al. (2012) manufactured a steerable curved concentric tube robot that can enter the heart through the vasculature, using a unique metal MEMS process.

Since 2014, Alambeigi F. of Johns Hopkins University et al. Alambeigi et al. (2014), Alambeigi et al. (2016a), Alambeigi et al. (2016b); Wilkening et al., 2017; Bakhtiarinejad et al., 2020) has researched the novel steerable drill using a continuum dexterous manipulator (CDM), and carried out experiments on both simulated and human cadaveric bones. Bakhtiarinejad M. Bakhtiarinejad et al. (2020) studied the use of curved drilling technique for treatment of osteonecrosis of femoral head. The biomechanical study demonstrated that a novel robot-assisted curved core decompression (CCD) technique is introduced to provide surgeons with direct access to the lesions causing minimal damage to the healthy bone. The progress of bistable structure, shape memory polymer and intelligent variable structure promoted the adaptability of surgical robot, which enable the biomechanical channels for orthopaedics.

2.3.2 Intelligent sensing technology can lead to precise trajectory in complex environment

Unlike industrial scene, the intraoperative environments are always complex. The robot cannot perceive, explain and understand the surrounding environment as well as surgeons is one of the key factors that the robotics ca not perform surgery automatically. Deformation predicting and tracking, and haptic feedback are tow common intelligent sensing technologies in surgical robotics (Figure 13). Shademan Shademan et al. (2016) demonstrated in vivo supervised autonomous soft tissue surgery which can suture a wound automatically. Deep reinforcement learning policies were used in Pattern Cutting task to induce tension in the material as cutting proceeds (Thananjeyan et al., 2017). Force sensors were developed to estimate interaction forces in robotic surgery (Kesner and Howe, 2011; Marban et al., 2018). And a Semi-Supervised Deep Neural Network Model was applied to understand the operation pattern (Marban et al., 2018).

FIGURE 13

2.3.3 Robotic skill learning can improve the automation level of surgical robotics

The surgical robotics in clinical are commonly semi-automatic. They can perform intraoperative actions based on the double closed-loop control of force position mixing in puncture (Wells et al., 2016), cochlear implant (Wang et al., 2020), fracture reduction (Lei et al., 2019) or other surgeries. However, it is impossible to dynamically adjust the task strategy according to the actual situation (Graham et al., 2008; Zhang, 2018). Skill learning (Silver et al., 2016; Mahler et al., 2019) make robots execute task which difficult by traditional control methods (Levine et al., 2016; Sun et al., 2017; Mandlekar et al., 2018), like pouring water and screwing screws, automatically.

3 Future research directions for satisfied repair of musculoskeletal injuries

The multi-layer intelligent technologies (MLIT) offer possibilities to achieve satisfied repair of musculoskeletal injuries. However, there are still no wide range of clinical applications, because of the following reasons: 1) the pathogenesis like the femoral head necrosis or the spinal tumor, the healing process like the ligament or the bone repair, and the biomechanical or kinematics models are not clarified in detail. 2) the real environment is hard to simulate. As there are differences between the simulation environment and the real environment, further verification of some research results should be implemented. 3) some theoretical designed surgical plan cannot be realized as expected. Further research will be needed for satisfied repair of musculoskeletal injuries:

3.1 The research on the pathogenesis or the healing process can be more effective through new AI technologies

3.1.1 Biomechanical models based on multi-physical parameter prediction will be built

Current biomechanical models are always based on 3D reconstructed models. However, morphology and parameters of the ligament, meniscus (Antico et al., 2020) and other tissue are always artificially devised (Viris et al., 2016), and the same density of different bone tissues like cortical bone, cancellous bone and trabecular bone structures are commonly set in biomechanical analysis. It is said that kinematics simulations are greatly different from real scene (Bell et al., 1996). To build real biomechanical models, physical parameter such as the bone density, ligament elastic modulus, must be predicted precisely.

Although some physical parameters predicting methods are proposed, which are discussed in Section2.1.2, lots of parameters still cannot be predicted for the lack of data with high-quality labels. Standard labeling procedure, and cross check of data labels are necessary.

3.1.2 Causal relationship between symptoms and physiological factors will be discovered by causal discovery

R Ganz (Siebenrock et al., 2010) proposed a surgical plan which can retain the vascular supply of the femoral head, to guarantee the oxygen and nutritional supply to the cells (Kramer et al., 2009). However, excessive vascular supply may be not conducive to recovery. Further research on the pathogenesis or the healing process for femoral head necrosis, especially the causality, but not only the correlation between the symptoms and physiological factors should be deserved. Now causal discovery, which can promote the research of disease mechanism, has been applied in hippocampal function analysis (Mannoor et al., 2013; Sanchez-Romero et al., 2018). Causal discovery can be a useful theory for the research on the pathogenesis or the healing process.

3.2 Intelligent technologies will make add the functionality of implants and make the implants manufacture process more efficiently

3.2.1 Implant with intelligent sense will provide real recovery status

Implants bioprinting combined with intelligent sense is also a research hotspot (Figure 14). In 2013, Mannoor M.S.et al. Mannoor et al. (2013) generated a bionic ear via additive manufacturing of biological cells with structural and nanoparticle derived electronic elements. In 2019, Jodat, Y.A. et al. Jodat et al. (2020) designed an odor-perceptive nose-like hybrid, with 3D cartilage-like tissue constructs. It is composed of a mechanically robust cartilage-like construct and a biocompatible biosensing platform. Patients can not recover only by surgery without rehabilitation training as the human body systems has changed after operations. Doctors can obtain the patient’s status through regular follow-up. These data were often temporary and fragmented, and could not truly reflect the patient’s postoperative status. The smart prosthesis persona IQ of Zimmer Biomet has passed FDA certification (Zimmerbiomet, 2021). Persona IQ can not only implant in patients like the traditional prosthesis, but also record the range, speed and other index of gait, for assessing postoperative recovery progress. Implant with intelligent sense will provide real recovery status for changing rehabilitation plans.

FIGURE 14

3.2.2 Intelligent models 3D/4D printing process will optimize the manufacturing process

The performance of 3d/4d printed matter is not only related to its own physical, mechanical and biological properties, but also affected by the process parameters of the printing process (Naghieh et al., 2020). Study of Zheng W. et al. Zheng et al. (2021) showed the following facts: ① Higher pressures caused instability of extruded biomaterial, and subsequently, poor printability; ②Fell under or above the range of nozzle speeds caused poor strand printability; ③ The more viscous the biomaterial, the more appropriate printability could be achieved. Due to the complex conditions, it is hard to build an accurate mathematical printing model. Data driven AI technology (Shi et al., 2019; Yu and Jiang, 2020; Ruberu et al., 2021) has obvious advantages in the multi-objective optimization of biological printing process (Figure 14).

3.3 Advanced materials and variable structures will improve the treatment performance of robotics

As the physiological state of patients are different and changing (Ionov, 2018), such as temperature, pH (Ca2+ concentration), to achieve satisfied repair in long-term therapy, personalized surgical channels should be conducted (Xu et al., 2019b; Zappetti et al., 2020; Wang et al., 2022) or the endogenous repairs should be guided by the micro nano robots (Aghakhani et al., 2020; Lee et al., 2020). Combining the improvement of intraoperative operation skills and the providing signals for postoperative repair, the treatment effect can be improved through robots.

4 Conclusion

Satisfied recovery, a key goal of precise orthopaedics, has became possible, as AI technology in orthopaedics has progressed in the following apsects: physical parameters can be predicted precisely through AI technology, bioactive implants made through 3D printing technology can provide signals for endogenous repair, and personalized surgical channels can be established through intelligent robotic technology. However, the multi-layer intelligent technologies (MLIT) has not been used widely. With further study on biomechanical models for multi-physical parameter prediction, stimulus response mechanism of bioactive implants, smart implants, intelligent modeling of 3D/4D printing process and variable structure in long-term therapy (Senatov et al., 2016; Lin et al., 2019; Su et al., 2019), satisfied repair may be achieved by the multi-layer intelligent technologies (MLIT).

References

• 1

  AbrahamN.KhanN. M.A Novel Focal Tversky loss function with improved Attention U-Net for lesion segmentation. Preprint, 2018.

  • Google Scholar

• 2

  AghakhaniA.YasaO.WredeP.SittiM. (2020). Acoustically powered surface-slipping mobile microrobots. Proc. Natl. Acad. Sci. U. S. A.117 (7), 3469–3477. 10.1073/pnas.1920099117

  • CrossRef
  • Google Scholar

• 3

  AhrendM. D.TeunisT.NoserH.SchmidutzF.RichardsG.GueorguievB.et al (2021). 3D computational anatomy of the scaphoid and its waist for use in fracture treatment. J. Orthop. Surg. Res.16 (1), 216. 10.1186/s13018-021-02330-8

  • CrossRef
  • Google Scholar

• 4

  AlambeigiF.MurphyR. J.BasafaE.TaylorR. H.ArmandM. (2014). Control of the coupled motion of a 6 DoF robotic arm and a continuum manipulator for the treatment of pelvis osteolysis. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc.36, 6521–6525. 10.1109/EMBC.2014.6945122

  • CrossRef
  • Google Scholar

• 5

  AlambeigiF.WangY.MurphyR. J.IordachitaI.ArmandM. (2016). Toward robot-assisted hard osteolytic lesion treatment using a continuum manipulator. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc.2016, 5103–5106. 10.1109/EMBC.2016.7591875

  • CrossRef
  • Google Scholar

• 6

  AlambeigiF.SefatiS.MurphyR. J.IordachitaI.ArmandM.Design and characterization of a debriding tool in robot-assisted treatment of osteolysis. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), Stockholm, Sweden, 2016, pp. 5664–5669. 10.1109/ICRA.2016.7487787

  • CrossRef
  • Google Scholar

• 7

  AnticoM.SasazawaF.DunnhoferM.CampsS.JaiprakashA.PandeyA.et al (2020). Deep learning-based femoral cartilage automatic segmentation in ultrasound imaging for guidance in robotic knee arthroscopy. Ultrasound Med. Biol.46 (2), 422–435. 10.1016/j.ultrasmedbio.2019.10.015

  • CrossRef
  • Google Scholar

• 8

  BakhtiarinejadM.AlambeigiF.ChamaniA.UnberathM.KhanujaH.ArmandM. “A biomechanical study on the use of curved drilling technique for treatment of osteonecrosis of femoral head,” in Proceedings of the Medical Image Computing and Computer-Assisted Intervention, 2020, Cham (Springer).

  • Google Scholar

• 9

  BalagopalA.KazemifarS.DanN.LinM. H.HannanR.OwrangiA.et al (2018). Fully automated organ segmentation in male pelvic CT images.Phys. Med. Biol.63 (24), 245015. 10.1088/1361-6560/aaf11c

  • CrossRef
  • Google Scholar

• 10

  BaoM.WangX.YuanH.LouX.ZhaoQ.ZhangY. (2016). HAp incorporated ultrafine polymeric fibers with shape memory effect for potential use in bone screw hole healing. J. Mat. Chem. B4 (31), 5308–5320. 10.1039/C6TB01305h

  • CrossRef
  • Google Scholar

• 11

  BellK. L.GarrahanN.KneisselM.LoveridgeN.GrauE.StantonM.et al (1996). Cortical and cancellous bone in the human femoral neck: Evaluation of an interactive image analysis system. Bone19 (5), 541–548. 10.1016/S8756-3282(96)00245-1

  • CrossRef
  • Google Scholar

• 12

  BernardoI.EdoardoB. (2021). Robotics in orthopaedic surgery: Why, what and how?Archives Orthop. Trauma Surg.141 (12), 2035–2042.

  • Google Scholar

• 13

  BhandarkarS.DhatrakP. (2022). Optimization of a knee implant with different biomaterials using finite element analysis. Mater. Today Proc.59 (1), 459–467. 10.1016/j.matpr.2021.11.466

  • CrossRef
  • Google Scholar

• 14

  BlackK. M.LawH.AldoukhiA.DengJ.GhaniK. R. (2020). Deep learning computer vision algorithm for detecting kidney stone composition. BJU Int.125 (6), 920–924. 10.1111/bju.15035

  • CrossRef
  • Google Scholar

• 15

  BlendowskiM.HansenL.HeinrichM. P . (2021). Weakly-supervised learning of multi-modal features for regularised iterative descent in 3D image registration. Med. Image Anal.67, 101822. 10.1016/j.media.2020.101822

  • CrossRef
  • Google Scholar

• 16

  ButscherA.BohnerM.DoebelinN.HofmannS.MüllerR. (2013). New depowdering-friendly designs for three-dimensional printing of calcium phosphate bone substitutes. Acta biomater.9 (11), 9149–9158. 10.1016/j.actbio.2013.07.019

  • CrossRef
  • Google Scholar

• 17

  CaiD.WangZ.LiuY., Automatic path planning for navigated pedicle screw surgery based on deep neural network, Proceedings of the 2019 WRC Symposium on Advanced Robotics and Automation (WRC SARA). 2019.

  • Google Scholar

• 18

  CapraraS.FasserM. R.SpirigJ. M., (2021). Bone density optimized pedicle screw instrumentation improves screw pull-out force in lumbar vertebrae. Comput. Methods Biomechanics Biomed. Eng.203, 1–11. 10.1080/10255842.2021.1959558

  • CrossRef
  • Google Scholar

• 19

  CapraraS. (2021). Towards the integration of computational methods in spinal surgical planning: A combination of deep learning, statistical, and finite element methods. Switzerland: ETH Zurich Research Collection. 10.3929/ethz-b-000475269

  • CrossRef
  • Google Scholar

• 20

  ChangY. L.LoY. J.FengS. W.HuangY. C.TsaiH. Y.LinC. T.et al (2016). Bone healing improvements using hyaluronic acid and hydroxyapatite/beta-tricalcium phosphate in combination: An animal study. Biomed Res. Int.2016, 1–8. 10.1155/2016/8301624

  • CrossRef
  • Google Scholar

• 21

  ChenF.ZhaoZ.GaoC.LiuJ.SuX.ZhaoJ.et al (2018). Clustering of morphological features for identifying femur cavity subtypes with difficulties of intramedullary nail implantation. IEEE J. Biomed. Health Inf.22 (4), 1209–1217. 10.1109/JBHI.2017.2761980

  • CrossRef
  • Google Scholar

• 22

  ChenW.ZhangY.HeJ.QiaoY.ChenY.ShiH.et alW-net: Bridged U-net for 2D medical image segmentation. 2018, preprint.

  • Google Scholar

• 23

  ChenJ.LuY.YuQ., (2021). TransUNet: Transformers make strong encoders for medical image segmentation. Comput. Vis. Pattern Recognit.10.48550/arXiv.2102.04306

  • CrossRef
  • Google Scholar

• 24

  ChenZ.WeiJ.LiR., (2022). “Unsupervised multi-modal medical image registration via discriminator-free image-to-image translation,” in Proceedings of the International Joint Conference on Artificial Intelligence, Vienna, Austria.

  • Google Scholar

• 25

  DasK.ZhangQ. . Convolutional recurrent residual U-net embedded with attention mechanism and focal Tversky loss function for cancerous nuclei detection. Preprint, 2020.

  • Google Scholar

• 26

  DengZ.GuoY.ZhaoX.LiL.DongR.GuoB.et al (2016). Stretchable degradable and electroactive shape memory copolymers with tunable recovery temperature enhance myogenic differentiation. Acta Biomater.46, 234–244. 10.1016/j.actbio.2016.09.019

  • CrossRef
  • Google Scholar

• 27

  DengH.BirdalT.IlicS. (2018). “Ppfnet: Global context aware local features for robust 3d point matching,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 18-23 June 2018, Salt Lake City, UT, USA, 195–205.

  • Google Scholar

• 28

  DePhillipoN. N.MoatsheG.BradyA.ChahlaJ.AmanZ. S.DornanG. J.et al (2018). Effect of meniscocapsular and meniscotibial lesions in ACL-deficient and ACL-reconstructed knees: A biomechanical study. Am. J. Sports Med.46 (10), 2422–2431. 10.1177/0363546518774315

  • CrossRef
  • Google Scholar

• 29

  DhivyaS.AjitaJ.SelvamuruganN. (2015). Metallic nanomaterials for bone tissue engineering. J. Biomed. Nanotechnol.11 (10), 1675–1700. 10.1166/jbn.2015.2115

  • CrossRef
  • Google Scholar

• 30

  DingH.MiaoQ.LiuH.WangH.TuQ.ChenJ.et al (2017). Preoperative design and clinical application of TKA assisted by CAD combined with 3D printing navigation templates. Chin. Orthop. J. Clin. Basic Res.9 (5), 261–268. 10.3969/j.issn.1674-666X.2017.05.001

  • CrossRef
  • Google Scholar

• 31

  DongQ.HuangZ.SunY., Prediction of accuracy and screw size by pedicle anatomic parameters and screws in idiopathic scoliosis with freehand screw placement based on machine learning. Preprint, 2021.

  • Google Scholar

• 32

  DongZ.HeY.QiX., “MNet: Rethinking 2D/3D networks for anisotropic medical image segmentation,” in Proceedings of the International Joint Conference on Artificial Intelligence, Vienna, Austria, 10 May 2022.

  • Google Scholar

• 33

  FalkT.MaiD.BenschR.ÇiçekÖ.AbdulkadirA.MarrakchiY. (2019). U-net: Deep learning for cell counting, detection, and morphometry. Nat. Methods16 (1), 67–70. 10.1038/s41592-018-0261-2

  • CrossRef
  • Google Scholar

• 34

  FanJ.CaoX.WangQ.YapP.ShenD. (2019). Adversarial learning for mono- or multi-modal registration. Med. image Anal.58, 101545. 10.1016/j.media.2019.101545

  • CrossRef
  • Google Scholar

• 35

  GhezelbashF.SchmidtH.Shirazi-AdlA.El-RichM. (2020). Internal load-sharing in the human passive lumbar spine: Review of in vitro and finite element model studies. J. Biomech.102, 109441. 10.1016/j.jbiomech.2019.109441

  • CrossRef
  • Google Scholar

• 36

  GilbertH.HendrickR.RemirezA.WebsterR.III (2014). A robot for transnasal surgery featuring needle-sized tentacle-like arms. Expert Rev. Med. devices11 (1), 5–7. 10.1586/17434440.2013.854702

  • CrossRef
  • Google Scholar

• 37

  GladmanA. S.MatsumotoE. A.NuzzoR. G.MahadevanL.LewisJ. A. (2016). Biomimetic 4D printing.Nat. Mat.15 (4), 413–418. 10.1038/nmat4544

  • CrossRef
  • Google Scholar

• 38

  GojcicZ.ZhouC.WegnerJ. D.WieserA. (2019). “The perfect match: 3d point cloud matching with smoothed densities,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Long Beach, CA, June 15–20, 2019, 5545–5554.

  • Google Scholar

• 39

  GoslineA. H.VasilyevN. V.VeeramaniA.WuM.SchmitzG.ChenR.et al (2012). Metal MEMS tools for beating-heart tissue removal. IEEE Int. Conf. Robot. Autom, Saint Paul, MN, May 14–18, 2012, 1921–1926. 10.1109/ICRA.2012.6225210

  • CrossRef
  • Google Scholar

• 40

  GrahamA. E.XieS. Q.AwK. C.MukherjeeA. S.XuW. L. (2008). Bone–muscle interaction of the fractured femur. J. Orthop. Res.26 (8), 1159–1165. 10.1002/jor.20611

  • CrossRef
  • Google Scholar

• 41

  GrollJ.BurdickJ. A.ChoD. W.DerbyB.GelinskyM.HeilshornS. C.et al (2018). A definition of bioinks and their distinction from biomaterial inks. Biofabrication11 (1), 013001. 10.1088/1758-5090/aaec52

  • CrossRef
  • Google Scholar

• 42

  GuoN.YangB.JiX.WangY.HuL.WangT. (2019). Intensity-based 2D-3D registration for an ACL reconstruction navigation system. Int. J. Med. Robot.15, e2008. 10.1002/rcs.2008

  • CrossRef
  • Google Scholar

• 43

  HanL.LyuY.PengC., GAN-based disentanglement learning for chest X-ray rib suppression. Preprint, 2021.

  • Google Scholar

• 44

  HatamizadehA.NathV.TangY.YangD.RothH. R.XuD. (2021). Swin UNETR: Swin transformers for semantic segmentation of brain tumors in MRI images. Brainlesion Glioma, Multiple Scler. Stroke Trauma. Brain Inj. BrainLes2021, 272–284. 10.1007/978-3-031-08999-2_22

  • CrossRef
  • Google Scholar

• 45

  HatamizadehA.YangD.RothH., (2022). “Unetr: Transformers for 3D medical image segmentation,” in 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Waikoloa, HI, 15 February 2022. 10.1109/WACV51458.2022.00181

  • CrossRef
  • Google Scholar

• 46

  HsiehC. I.ZhengK.LinC.MeiL.LuL.LiW.et al (2021). Automated bone mineral density prediction and fracture risk assessment using plain radiographs via deep learning. Nat. Commun.12, 5472. 10.1038/s41467-021-25779-x

  • CrossRef
  • Google Scholar

• 47

  HuangQ.SunJ.HuiD.WangX.WangG. (2018). Robust liver vessel extraction using 3D U-Net with variant dice loss function. Comput. Biol. Med.101, 153–162. 10.1016/j.compbiomed.2018.08.018

  • CrossRef
  • Google Scholar

• 48

  HuangX.MeiG.ZhangJ., A comprehensive survey on point cloud registration. Preprint, 2021.

  • Google Scholar

• 49

  HuangB.CaoB.ZhangJ.FengY.WangL.ChenX.et al (2021). Induction of ferroptosis in human nasopharyngeal cancer cells by cucurbitacin B: Molecular mechanism and therapeutic potential.Cell Death Dis.1, 237–267. 10.1038/s41419-021-03516-y

  • CrossRef
  • Google Scholar

• 50

  HughesA. J.DeBuitleirC.SodenP.O’DonnchadhaB.TanseyA.AbdulkarimA.et al (2017). 3D printing aids acetabular reconstruction in complex revision hip arthroplasty. Adv. Orthop.2017, 1–7. 10.1155/2017/8925050

  • CrossRef
  • Google Scholar

• 51

  IglovikovV.ShvetsA.TernausNet: U-net with VGG11 encoder pre-trained on ImageNet for image segmentation, Preprint, 2018.

  • Google Scholar

• 52

  IonovL. (2018). 4D biofabrication: Materials, methods, and applications. Adv. Healthc. Mat.7 (17), e1800412. 10.1002/adhm.201800412

  • CrossRef
  • Google Scholar

• 53

  IsenseeF.JaegerP. F.KohlS. A. A.PetersenJ.Maier-HeinK. (2021). nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat. Methods18 (2), 203–211. 10.1038/s41592-020-01008-z

  • CrossRef
  • Google Scholar

• 54

  JamwalP. K.HussainS.GhayeshM. H. (2021). Intrinsically compliant parallel robot for fractured femur reduction: Mechanism optimization and control. Robotics Aut. Syst.141, 103787. 10.1016/j.robot.2021.103787

  • CrossRef
  • Google Scholar

• 55

  JiangJ.CetinE.CeliktutanO. (2021). “IB-DRR: Incremental learning with information-back discrete representation replay,” in Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Nashville, TN, USA, 19-25 June 2021.

  • Google Scholar

• 56

  JinQ.MengZ.SunC.LeyiW. (2018). Ran SRA-UNet: A hybrid deep attention-aware network to extract liver and tumor in CT scans. J. latex Cl. files octber. 10.3389/fbioe.2020.605132

  • CrossRef
  • Google Scholar

• 57

  JodatY. A.KiaeeK.JarquinD. V.HernándezR.WangT.JoshiS.et al (2020). A 3D-printed hybrid nasal cartilage with functional electronic olfaction. Adv. Sci. (Weinh).7 (5), 1901878. 10.1002/advs.201901878

  • CrossRef
  • Google Scholar

• 58

  JordiM.AnneE.Maureenv. E.RubenP.TymourF.KeesJ. B.et al (2022). A review on the application of deep learning for CT reconstruction, bone segmentation and surgical planning in oral and maxillofacial surgery. Dentomaxillofacial Radiol.51, 20210437. 10.1259/dmfr.20210437

  • CrossRef
  • Google Scholar

• 59

  KamJ.GanC.DimouS.AwadM.KavarB.NairG.et al (2019). Learning curve for robot-assisted percutaneous pedicle screw placement in thoracolumbar surgery. Asian Spine J.13 (6), 920–927. 10.31616/asj.2019.0033

  • CrossRef
  • Google Scholar

• 60

  KangH. W.LeeS. J.KoI. K.KenglaC.YooJ. J.AtalaA. (2016). A 3D bioprinting system to produce human-scale tissue constructs with structural integrity. Nat. Biotechnol.34 (3), 312–319. 10.1038/nbt.3413

  • CrossRef
  • Google Scholar

• 61

  KawaiT.ShanjaniY.FazeliS.BehnA. W.OkuzuY.GoodmanS. B.et al (2018). Customized, degradable, functionally graded scaffold for potential treatment of early stage osteonecrosis of the femoral head. J. Orthop. Res.36 (3), 1002–1011. 10.1002/jor.23673

  • CrossRef
  • Google Scholar

• 62

  KesnerS. B.HoweR. D. (2011). “Force control of flexible catheter robots for beating heart surgery,” IEEE Int Conf Robot Autom., 2011, 1589–1594. 10.1109/ICRA.2011.5979690

  • CrossRef
  • Google Scholar

• 63

  KolaříkM.BurgetR.UherV.ŘíhaK.DuttaM. K. (2019). Optimized high resolution 3D dense-U-net network for brain and spine segmentation. Appl. Sci. (Basel).9, 404. 10.3390/app9030404

  • CrossRef
  • Google Scholar

• 64

  KoleskyD. B.TrubyR. L.GladmanA. S.BusbeeT. A.HomanK. A.LewisJ. A. (2014). 3D bioprinting of vascularized, heterogeneous cell-laden tissue constructs. Adv. Mat.26 (19), 3124–3130. 10.1002/adma.201305506

  • CrossRef
  • Google Scholar

• 65

  KramerJ.ScheureckerG.ScheureckerA.et al (2009). Femoral head necrosis. Dordrecht, Netherlands: Springer Netherlands.

  • Google Scholar

• 66

  LampenN.KimD.FangX.XuX.KuangT.DengH. H.et al (2022). Deep learning for biomechanical modeling of facial tissue deformation in orthognathic surgical planning. Int. J. Comput. Assist. Radiol. Surg.17, 945–952. 10.1007/s11548-022-02596-1

  • CrossRef
  • Google Scholar

• 67

  LeeH. S.GoG.ChoiE.KangB.ParkJ. O.KimC. S. (2020). Medical microrobot - wireless manipulation of a drug delivery carrier through an external ultrasonic actuation: Preliminary results. Int. J. Control Autom. Syst.18 (1), 175–185. 10.1007/s12555-019-0239-6

  • CrossRef
  • Google Scholar

• 68

  Le-HuuD. K.ParagiosN. (2017). “Alternating direction graph matching,” in Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Hawaii, United States, July 2017, (IEEE), 4914–4922. 10.1109/CVPR.2017.522

  • CrossRef
  • Google Scholar

• 69

  LeiJ.ZhengG.HuL., (2019). “Force/position control simulation of robot-assisted fracture reduction,” in Proceedings of the 2019 WRC Symposium on Advanced Robotics and Automation (WRC SARA) (IEEE), 21-22 August 2019, Beiging, China.

  • Google Scholar

• 70

  LevineS.FinnC.DarrellT.AbbeelP. (2016). End-to-End training of deep visuomotor policies. J. Mach. Learn. Res.17 (1), 1334–1373. 10.1016/j.dam.2015.09.011

  • CrossRef
  • Google Scholar

• 71

  LiX.HaoC.QiX.QiD.FuC. W.HengH-DenseUNetP. A. (2017). H-DenseUNet: Hybrid densely connected UNet for liver and tumor segmentation from CT volumes. IEEE Trans. Med. Imaging37 (12), 2663–2674. 10.1109/TMI.2018.2845918

  • CrossRef
  • Google Scholar

• 72

  LiJ.DunstanC. R.EntezariA.LiQ.SteckR.SaifzadehS.et al (2019). A novel bone substitute with high bioactivity, strength, and porosity for repairing large and load-bearing bone defects. Adv. Healthc. Mat.8 (8), 1801298. 10.1002/adhm.201801298

  • CrossRef
  • Google Scholar

• 73

  LiY.PavanramP.ZhouJ.LietaertK.TaheriP.LiW.et al (2020). Additively manufactured biodegradable porous zinc. Acta Biomater.101, 609–623. 10.1016/j.actbio.2019.10.034

  • CrossRef
  • Google Scholar

• 74

  LiangP.ChenJ.ZhengH., (2019). “Cascade decoder: A universal decoding method for biomedical image segmentation,” in Proceedings of the 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019), Venice, Italy, April 8–11, 2019 (IEEE), pp. 339–342. 10.1109/ISBI.2019.8759430

  • CrossRef
  • Google Scholar

• 75

  LinC.LvJ.LiY.ZhangF.LiuY., (2019). 4D-Printed biodegradable and remotely controllable shape memory occlusion devices. Adv. Funct. Mat.29 (51), 1906569. 10.1002/adfm.201906569

  • CrossRef
  • Google Scholar

• 76

  LiuY.WangR.ChenS.XuZ.WangQ.YuanP.et al (2020). Heparan sulfate loaded polycaprolactone-hydroxyapatite scaffolds with 3D printing for bone defect repair. Int. J. Biol. Macromol.148, 153–162. 10.1016/j.ijbiomac.2020.01.109

  • CrossRef
  • Google Scholar

• 77

  LiuB.SunC.HouG.YangZ.LiuZ.TianG.et al (2021). Clinical application of 3D printing technology: Bone defect reconstruction of limbs based on ‘precision orthopaedics’. Chin. J. Minim. Invasive Surg.21 (4), 303–308. 10.3969/j.issn.1009-6604.2021.04.004

  • CrossRef
  • Google Scholar

• 78

  LoebelC.BurdickJ. A. (2018). Engineering stem and stromal cell therapies for musculoskeletal tissue repair. Cell Stem Cell22 (3), 325–339. 10.1016/j.stem.2018.01.014

  • CrossRef
  • Google Scholar

• 79

  LopezC.GazgalisA.BoddapatiV.ShahR. P.CooperH. J.GellerJ. A. (2021). Artificial learning and machine learning decision guidance applications in total hip and knee arthroplasty: A systematic review. Arthroplast. Today11, 103–112. 10.1016/j.artd.2021.07.012

  • CrossRef
  • Google Scholar

• 80

  LopezF.VarelaA.HinojosaO.MendezM.TrinhD. H.ElBezeY.et alAssessing deep learning methods for the identification of kidney stones in endoscopic images. Proceedings of the 2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), Mexico, 2021: 2778–2781. 10.1109/EMBC46164.2021.9630211

  • CrossRef
  • Google Scholar

• 81

  LuJ.OfverstedtJ.LindbladJ.SladojeN.Is image-to-image translation the panacea for multimodal image registration? A comparative study. 2021, ArXiv, 2021 abs/2103.16262.

  • Google Scholar

• 82

  LuoY.ZhaiD.HuanZ.ZhuH.XiaL.ChangJ.et al (2015). Three-dimensional printing of hollow-struts-packed bioceramic scaffolds for bone regeneration. ACS Appl. Mat. Interfaces7 (43), 24377–24383. 10.1021/acsami.5b08911

  • CrossRef
  • Google Scholar

• 83

  MahlerJ.MatlM.SatishV.DanielczukM.DeRoseB.McKinleyS.et al (2019). Learning ambidextrous robot grasping policies. Sci. Robot.4 (26), eaau4984. 10.1126/scirobotics.aau4984

  • CrossRef
  • Google Scholar

• 84

  MakvandiP.WangC.ZareE. N.BorzacchielloA.NiuL.TayF. R. (2020). Metal-based nanomaterials in biomedical applications: Antimicrobial activity and cytotoxicity aspects. Adv. Funct. Mat.30 (22), 1910021. 10.1002/adfm.201910021

  • CrossRef
  • Google Scholar

• 85

  MandlekarA.ZhuY.GargA., Robo Turk: A crowdsourcing platform for robotic skill learning through imitation. Proceedings of the International Conference on Robot Learning, Zürich, Switzerland, 2018:1–15.

  • Google Scholar

• 86

  MannoorM. S.JiangZ.JamesT.KongY.MalatestaK. A.SoboyejoW. O.et al (2013). 3D printed bionic ears. Nano Lett.13 (6), 2634–2639. 10.1021/nl4007744

  • CrossRef
  • Google Scholar

• 87

  MarbanA.SrinivasanV.SamekW., (2018). “Estimation of interaction forces in robotic surgery using a semi-supervised deep neural network model,” in Proceedings of the 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (IEEE), 01-05 Oct. 2018, Madrid, Spain.

  • Google Scholar

• 88

  MistryS.KunduD.DattaS.BasuD. (2015). Comparison of bioactive glass coated and hydroxyapatite coated titanium dental implants in the human jaw bone. Aust. Dent. J.56 (1), 68–75. 10.1111/j.1834-7819.2010.01305.x

  • CrossRef
  • Google Scholar

• 89

  MolnarD.EnqvistO.UlénJ.LarssonM.BrandbergJ.JohnssonA. A.et al (2021). Artificial intelligence based automatic quantification of epicardial adipose tissue suitable for large scale population studies. Sci. Rep.11 (1), 23905. 10.1038/s41598-021-03150-w

  • CrossRef
  • Google Scholar

• 90

  NaghiehS.SarkerM. D.SharmaN. K.BarhoumiZ.ChenX. (2020). Printability of 3D printed hydrogel scaffolds: Influence of hydrogel composition and printing parameters. Appl. Sci. (Basel).10 (1), 292. 10.3390/app10010292

  • CrossRef
  • Google Scholar

• 91

  NeumannD. A. (2010). Kinesiology of the musculoskeletal system. Maryland Heights, Missouri, United States: Mosby, Elsevier.

  • Google Scholar

• 92

  NicolasN.EmilieC. S.VincentC.BenoitB.MatthieuV.JulienB. (2016). Robot-assisted spine surgery: Feasibility study through a prospective case-matched analysis. Eur. Spine J.25 (3), 947–955. 10.1007/s00586-015-3758-8

  • CrossRef
  • Google Scholar

• 93

  PeiG. X.YanY. B. (2014). Current status and progress of digital orthopaedics in China. J. Orthop. Transl.2 (3), 107–117. 10.1016/j.jot.2014.05.001

  • CrossRef
  • Google Scholar

• 94

  PolceE. M.KunzeK. N.PaulK. M. (2021). Machine learning predicts femoral and tibial implant size mismatch for total knee arthroplasty. Arthroplasty Today8 (11), 268–277. 10.1016/j.artd.2021.01.006

  • CrossRef
  • Google Scholar

• 95

  QiG.DunnC. K.QiH. J.DunnM. L. (2014). Active origami by 4D printing. Smart Mater. Struct.23 (9), 094007. 10.1088/0964-1726/23/9/094007

  • CrossRef
  • Google Scholar

• 96

  QiC. R.SuH.MoK.GuibasL. J. (2017). “Pointnet: Deep learning on point sets for 3d classification and segmentation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Hawaii, United States, July 2017, pp. 77–85. 10.1109/CVPR.2017.16

  • CrossRef
  • Google Scholar

• 97

  RavivD.ZhaoW.McknellyC.PapadopoulouA.KadambiA.ShiB.et al (2014). Active printed materials for complex self-evolving DeformationsActive printed materials for complex self-evolving deformations. Sci. Rep.4, 7422. 10.1038/srep07422

  • CrossRef
  • Google Scholar

• 98

  RenX.TuV.BischoffD.WeisgerberD. W.LewisM. S.YamaguchiD. T.et al (2016). Nanoparticulate mineralized collagen scaffolds induce in vivo bone regeneration independent of progenitor cell loading or exogenous growth factor stimulation. Biomaterials89, 67–78. 10.1016/j.biomaterials.2016.02.020

  • CrossRef
  • Google Scholar

• 99

  RezwanK.ChenQ. Z.BlakerJ. J.BoccacciniA. R. (2006). Biodegradable and bioactive porous polymer/inorganic composite scaffolds for bone tissue engineering. Biomaterials27 (18), 3413–3431. 10.1016/j.biomaterials.2006.01.039

  • CrossRef
  • Google Scholar

• 100

  RogersM. A.AikawaE. (2019). Cardiovascular calcification: Artificial intelligence and big data accelerate mechanistic discovery. Nat. Rev. Cardiol.16 (5), 261–274. 10.1038/s41569-018-0123-8

  • CrossRef
  • Google Scholar

• 101

  Roohani-EsfahaniS. I.LuZ. F.LiJ. J.Ellis-BehnkeR.KaplanD. L.ZreiqatH. (2012). Effect of self-assembled nanofibrous silk/polycaprolactone layer on the osteoconductivity and mechanical properties of biphasic calcium phosphate scaffolds. Acta biomater.8 (1), 302–312. 10.1016/j.actbio.2011.10.009

  • CrossRef
  • Google Scholar

• 102

  RuberuK.SenadeeraM.RanaS.GuptaS.ChungJ.YueZ.et al (2021). Coupling machine learning with 3D bioprinting to fast track optimisation of extrusion printing. Appl. Mater. Today22, 100914. 10.1016/j.apmt.2020.100914

  • CrossRef
  • Google Scholar

• 103

  SalehiS. S. M.ErdogmusD.GholipourA. “Tversky loss function for image segmentation using 3D fully convolutional deep networks,” in Proceedings of the Machine Learning in Medical Imaging, Quebec City, QC, September 2017 (Springer). 10.1007/978-3-319-67389-9_44

  • CrossRef
  • Google Scholar

• 104

  Sanchez-RomeroR.RamseyJ. D.ZhangK., Causal discovery of feedback networks with functional magnetic resonance imaging, Preprint, 2018.

  • Google Scholar

• 105

  SenatovF. S.NiazaK. V.ZadorozhnyyM. Y.MaksimkinA. V.KaloshkinS. D.EstrinY. Z., (2016). Mechanical properties and shape memory effect of 3D-printed PLA-based porous scaffolds. J. Mech. Behav. Biomed. Mater.57, 139–148. 10.1016/j.jmbbm.2015.11.036

  • CrossRef
  • Google Scholar

• 106

  ShademanA.DeckerR. S.OpfermannJ. D.LeonardS.KriegerA.KimP. C. W. (2016). Supervised autonomous robotic soft tissue surgery. Sci. Transl. Med.8 (337), 337ra64. 337ra64-337ra64. 10.1126/scitranslmed.aad9398

  • CrossRef
  • Google Scholar

• 107

  ShiJ.SongJ.SongB.LuW. (2019). Multi-objective optimization design through machine learning for drop-on-demand bioprinting. Engineering5 (3), 586–593. 10.1016/j.eng.2018.12.009

  • CrossRef
  • Google Scholar

• 108

  SiebenrockK. A.PowellJ. N.GanzR. (2010). Osteochondritis dissecans of the femoral head. HIP Int.20 (4), 489–496. 10.1177/112070001002000412

  • CrossRef
  • Google Scholar

• 109

  SiemionowK. B.ForsthoefelC. W.FoyM. P.GawelD.LucianoC. J. (2021). Autonomous lumbar spine pedicle screw planning using machine learning: A validation study. J. Craniovertebral Junction Spine12 (3), 223–227.

  • Google Scholar

• 110

  SilverD.HuangA.MaddisonC. J.GuezA.SifreL.van den DriesscheG.et al (2016). Mastering the game of Go with deep neural networks and tree search. Nature529 (7587), 484–489. 10.1038/nature16961

  • CrossRef
  • Google Scholar

• 111

  SoodmandE.ZhengG.SteensW.BaderR.NolteL.KluessD. (2019). Surgically relevant morphological parameters of proximal human femur: A statistical analysis based on 3D reconstruction of CT data. Orthop. Surg.11, 135–142. 10.1111/os.12416

  • CrossRef
  • Google Scholar

• 112

  SuJ. W.GaoW.TrinhK.KendereS. M.PulatsuE. T.ZhangC.et al (2019). 4D printing of polyurethane paint-based composites. Int. J. Smart Nano Mater. (29), 1–12. 10.1080/19475411.2019.1618409

  • CrossRef
  • Google Scholar

• 113

  SunF.LiuC.HuangW., (2017). Object classification and grasp planning using visual and tactile sensing. IEEE Trans. Syst. Man Cybern. Syst.46 (7), 969–979. 10.1109/TSMC.2016.2524059

  • CrossRef
  • Google Scholar

• 114

  ThananjeyanB.GargA.KrishnanS., (2017). “Multilateral surgical pattern cutting in 2D orthotropic gauze with deep reinforcement learning policies for tensioning,” in Proceedings of the 2017 IEEE International Conference on Robotics and Automation (ICRA) (IEEE). 29 May 2017-03 June 2017, Singapore.

  • Google Scholar

• 115

  ThitisetT.DamrongsakkulS.BunaprasertT.LeeanansaksiriW.HonsawekS. (2013). Development of collagen/demineralized bone powder scaffolds and periosteum-derived cells for bone tissue engineering application. Int. J. Mol. Sci.14 (1), 2056–2071. 10.3390/ijms14012056

  • CrossRef
  • Google Scholar

• 116

  TianW.LiuY.LiuB.HeD.WuJ.HanX.et al (2019). Guideline for thoracolumbar pedicle screw placement assisted by orthopaedic surgical robot. Orthop. Surg.11 (2), 153–159. 10.1111/os.12453

  • CrossRef
  • Google Scholar

• 117

  TibbitsS. (2014). 4D printing: Multi-material shape change. Archit. Des.84 (1), 116–121. 10.1002/ad.1710

  • CrossRef
  • Google Scholar

• 118

  TurnbullG.ClarkeJ.PicardF.RichesP.JiaL.HanF.et al (2018). 3D bioactive composite scaffolds for bone tissue engineering. Bioact. Mater.3 (3), 278–314. 10.1016/j.bioactmat.2017.10.001

  • CrossRef
  • Google Scholar

• 119

  UnberathM.ZaechA.-N.LeeS. C., (2018). “DeepDRR – a catalyst for machine learning in fluoroscopy-guided procedures,” in Medical image computing and computer assisted intervention – MICCAI 2018. MICCAI 2018. Lecture notes in computer science. Editors FrangiA.SchnabelJ.DavatzikosC.Alberola-LópezC.FichtingerG. (Cham: Springer), 11073. 10.1007/978-3-030-00937-3_12

  • CrossRef
  • Google Scholar

• 120

  UnnithanA. R.SasikalaA. R. K.ParkC. H.KimC. S. (2017). A unique scaffold for bone tissue engineering: An osteogenic combination of graphene oxide–hyaluronic acid–chitosan with simvastatin. J. Industrial Eng. Chem.46, 182–191. 10.1016/j.jiec.2016.10.029

  • CrossRef
  • Google Scholar

• 121

  Van BelleghemS.TorresL.SantoroM.MahadikB.WolfandA.KofinasP.et al (2020). Hybrid 3D printing of synthetic and cell‐laden bioinks for shape retaining soft tissue grafts. Adv. Funct. Mat.30 (3), 1907145. 10.1002/adfm.201907145

  • CrossRef
  • Google Scholar

• 122

  VijayanR.SilvaT. D.HanR.ZhangX.UneriA.DoerrS.et al (2019). Automatic pedicle screw planning using atlas-based registration of anatomy and reference trajectories.Phys. Med. Biol.64 (16), 165020. 10.1088/1361-6560/ab2d66

  • CrossRef
  • Google Scholar

• 123

  VillaM. M.WangL.HuangJ.RoweD. W.WeiM. (2015). Bone tissue engineering with a collagen-hydroxyapatite scaffold and culture expanded bone marrow stromal cells. J. Biomed. Mat. Res.103 (2), 243–253. 10.1002/jbm.b.33225

  • CrossRef
  • Google Scholar

• 124

  VirisA.StefanoudakisG.PetousisM.VidakisN.TsainisA. M.KandylaB. (2016). Evaluation of an intact, an ACL-deficient, and a reconstructed human knee joint finite element model.Comput. Methods Biomech. Biomed. Engin.19 (3), 263–270. 10.1080/10255842.2015.1015526

  • CrossRef
  • Google Scholar

• 125

  WangY.SolomonJ. M. . Deep closest point: Learning representations for point cloud registration. Preprint, 2019.

  • Google Scholar

• 126

  WangT.WangL.TangP., (2018). Automatic bone segmentation and ultrasound—CT registration for robotic assisted femoral shaft fracture reduction. J. Med. Imaging Health Inf.8 (1), 151–156(6). 10.1166/jmihi.2018.2249

  • CrossRef
  • Google Scholar

• 127

  WangJ.LiuH.KeJ.HuL.ZhangS.YangB.et al (2020). Image-guided cochlear access by non-invasive registration: A cadaveric feasibility study. Sci. Rep.10, 18318. 10.1038/s41598-020-75530-7

  • CrossRef
  • Google Scholar

• 128

  WangS.XieZ.YuanF.LiL.LiuY.WangT.et al (2022). Bio-inspired physical intelligence for soft robotics. Chin. Sci. Bull.67 (10), 959–975. 10.1360/TB-2021-1217

  • CrossRef
  • Google Scholar

• 129

  WatanabeH.KanouK.KobayashiY.FujieM. G. (2011). “Development of a “steerable drill” for ACL reconstruction to create the arbitrary trajectory of a bone tunnel[C],” in IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2011, San Francisco, CA, September 25–30, 2011 (IEEE).

  • Google Scholar

• 130

  WellsT. S.YangS.MaclachlanR. A.LobesL. A.MartelJ. N.RiviereC. N. (2016). Hybrid position/force control of an active handheld micromanipulator for membrane peeling. Int. J. Med. Robot. Comput. Assist. Surg.12 (1), 85–95. 10.1002/rcs.1659

  • CrossRef
  • Google Scholar

• 131

  WilkeningP.AlambeigiF.MurphyR. J.TaylorR. H.ArmandM. (2017). Development and experimental evaluation of concurrent control of a robotic arm and continuum manipulator for osteolytic lesion treatment. IEEE Robot. Autom. Lett.2 (3), 1625–1631. 10.1109/LRA.2017.2678543

  • CrossRef
  • Google Scholar

• 132

  WuH.LeiP.LiuG.ZhangY. S.YangJ.ZhangL.et al (2017). Reconstruction of large-scale defects with a novel hybrid scaffold made from poly(L-lactic acid)/Nanohydroxyapatite/Alendronate-loaded chitosan microsphere: In vitro and in vivo studies. Sci. Rep.7 (1), 359. 10.1038/s41598-017-00506-z

  • CrossRef
  • Google Scholar

• 133

  WubnehA.TsekouraE. K.AyranciC.UludağH. (2018). Current state of fabrication technologies and materials for bone tissue engineering. Acta Biomater.80, 1–30. 10.1016/j.actbio.2018.09.031

  • CrossRef
  • Google Scholar

• 134

  XiaoD.LianC.DengH.KuangT.LiuQ.MaL.et al (2021). Estimating reference Bony shape models for orthognathic surgical planning using 3D point-cloud deep learning. IEEE J. Biomed. Health Inf.25 (8), 2958–2966. Epub 2021 Aug 5. PMID: 33497345; PMCID: PMC8310896. 10.1109/JBHI.2021.3054494

  • CrossRef
  • Google Scholar

• 135

  XieR.HuJ.NgF.TanL.QinT.ZhangM.et al (2017). High performance shape memory foams with isocyanate-modified hydroxyapatite nanoparticles for minimally invasive bone regeneration. Ceram. Int.43 (6), 4794–4802. 10.1016/j.ceramint.2016.11.216

  • CrossRef
  • Google Scholar

• 136

  XieR.HuJ.HoffmannO.ZhangY.NgF.QinT.et al (2018). Self-fitting shape memory polymer foam inducing bone regeneration: A rabbit femoral defect study. Biochimica Biophysica Acta - General Subj.1862 (4), 936–945. 10.1016/j.bbagen.2018.01.013

  • CrossRef
  • Google Scholar

• 137

  XuT.BinderK. W.AlbannaM. Z.DiceD.ZhaoW.YooJ. J.et al (2013). Hybrid printing of mechanically and biologically improved constructs for cartilage tissue engineering applications. Biofabrication5 (1), 015001. 10.1088/1758-5082/5/1/015001

  • CrossRef
  • Google Scholar

• 138

  XuX.TaoJ.WangS.YangL.ZhangJ.ZhangJ.et al (2019). 3D printing of nerve conduits with nanoparticle-encapsulated RGFP966. Appl. Mater. Today16, 247–256. 10.1016/j.apmt.2019.05.014

  • CrossRef
  • Google Scholar

• 139

  XuB.HanX.HuY.LouY.ChenC.ChenZ.et al (2019). A remotely controlled transformable soft robot based on engineered cardiac tissue construct. Small15 (18), e1900006. 10.1002/smll.201900006

  • CrossRef
  • Google Scholar

• 140

  YanK.CaiJ.JinD.MiaoS.GuoD.HarrisonA. P.et al (2022). Sam: Self-Supervised learning of pixel-wise anatomical embeddings in radiological images. IEEE Trans. Med. Imaging, 1. 10.1109/TMI.2022.3169003

  • CrossRef
  • Google Scholar

• 141

  YasakaK.AkaiH.KunimatsuA.KiryuS.AbeO. (2020). Prediction of bone mineral density from computed tomography: Application of deep learning with a convolutional neural network. Eur. Radiol.30 (6), 3549–3557. PMID: 32060712. 10.1007/s00330-020-06677-0

  • CrossRef
  • Google Scholar

• 142

  YeR.ChenY. (2009). “Path planning for robot assisted femur shaft fracture reduction: A preliminary investigation,” in Proceedings of the 2009 IEEE International Conference on Virtual Environments, Human-Computer Interfaces and Measurements Systems (IEEE), 11-13 May 2009, Hong Kong, China, 113–117. 10.1109/VECIMS.2009.5068876

  • CrossRef
  • Google Scholar

• 143

  YuC.JiangJ. (2020). A perspective on using machine learning in 3D bioprinting. Int. J. Bioprint.6 (1), 253. 10.18063/ijb.v6i1.253

  • CrossRef
  • Google Scholar

• 144

  YuanW.EckartB.KimK.JampaniV.FoxD.KautzJ. (2020). “DeepGMR: Learning Latent Gaussian Mixture Models for Registration,” in Computer Vision – ECCV 2020. ECCV 2020. Lecture Notes in Computer Science, October 29, 2020. Editors VedaldiA.BischofH.BroxT.FrahmJ. M. (Springer, Cham) vol 12350. 10.1007/978-3-030-58558-7_43

  • CrossRef
  • Google Scholar

• 145

  ZappettiD.JeongS. H.ShintakeJ.FloreanoD. (2020). Phase changing materials-based variable-stiffness tensegrity structures. Soft Robot.7 (3), 362–369. 10.1089/soro.2019.0091

  • CrossRef
  • Google Scholar

• 146

  ZengG.YangX.LiJ., (2017). “3D U-net with multi-level deep supervision: Fully automatic segmentation of proximal femur in 3D MR images,” in Proceedings of the International Workshop on Machine Learning in Medical Imaging, Cham (Springer).

  • Google Scholar

• 147

  ZhangD.BurkesW. L.SchoenerC. A.GrunlanM. A. (2012). Porous inorganic–organic shape memory polymers. Polymer53 (14), 2935–2941. 10.1016/j.polymer.2012.04.053

  • CrossRef
  • Google Scholar

• 148

  ZhangD.GeorgeO. J.PetersenK. M.Jimenez-VergaraA. C.HahnM. S.GrunlanM. A. (2014). A bioactive “self-fitting” shape memory polymer scaffold with potential to treat cranio-maxillo facial bone defects. Acta biomater.10 (11), 4597–4605. 10.1016/j.actbio.2014.07.020

  • CrossRef
  • Google Scholar

• 149

  ZhangY. S.AndreaA.SimoneB.Su-RyonS.KaiZ.ZahraG. M.et al (2016). Bioprinting 3D microfibrous scaffolds for engineering endothelialized myocardium and heart-on-a-chip. Biomaterials110, 45–59. 10.1016/j.biomaterials.2016.09.003

  • CrossRef
  • Google Scholar

• 150

  ZhangZ.MaZ.ZhangY.ChenF.ZhouY.AnQ. (2018). Dehydrothermally crosslinked collagen/hydroxyapatite composite for enhanced in vivo bone repair. Colloids Surfaces B Biointerfaces163, 394–401. 10.1016/j.colsurfb.2018.01.011

  • CrossRef
  • Google Scholar

• 151

  ZhangR.WuS.WuW.GaoH.ZhouZ. (2019). Computer-assisted needle trajectory planning and mathematical modeling for liver tumor thermal ablation: A review. Math. Biosci. Eng.16 (5), 4846–4872. PMID: 31499693. 10.3934/mbe.2019244

  • CrossRef
  • Google Scholar

• 152

  ZhangY. (2018). Springer, Singapore. Jointly published with People’s Medical Publishing House, Beijing, China (eBook). 10.1007/978-981-10-6044-1

  • CrossRef
  • Google Scholar

• 153

  ZhengG.YuW. (2017). Statistical shape and deformation models based 2D–3D reconstruction. Stat. Shape Deformation Analysis, 329–349. 10.1016/b978-0-12-810493-4.00015-8

  • CrossRef
  • Google Scholar

• 154

  ZhengG.TianW.ZhuangX. (2018). “Biomechanical optimization-based planning of periacetabular osteotomy: Artificial intelligence and smart image-guided technology for orthopaedics,” in Intelligent orthopaedics1093, 157–168. 10.1007/978-981-13-1396-7_13

  • CrossRef
  • Google Scholar

• 155

  ZhengW.WangY.ZhangF.LiC.LiuY.LengJ. (2018). Development of shape memory polymers micro/nanofiber membranes in biomedical applications. Sci. Sin. -Tech.48 (8), 811–826. 10.1360/n092018-00126

  • CrossRef
  • Google Scholar

• 156

  ZhengW.DongX.HeY.LiuY.LengJ. (2021). Development of bioactive polymers and their composite materials in bone tissue engineering. J. Harbin Inst. Technol.53 (8), 1. 10.11918/202007079

  • CrossRef
  • Google Scholar

• 157

  ZhouX. Y.GuoY.ShenM.YangG. Z. (2020). Application of artificial intelligence in surgery. Front. Med.14 (4), 417–430. 10.1007/s11684-020-0770-0

  • CrossRef
  • Google Scholar

• 158

  ZhouZ.SiddiqueeM.TajbakhshN.LiangJ. (2020). UNet++: Redesigning skip connections to exploit multiscale features in image segmentation. IEEE Trans. Med. Imaging39 (6), 1856–1867. 10.1109/tmi.2019.2959609

  • CrossRef
  • Google Scholar

• 159

  ZimmerbiometThe Smart Knee® by combining the power of Persona® the Personalized Knee® with CANARY caturio te with CHIRP system, Persona IQ is smart, connected and simple. Available at: https://www.zimmerbiomet.com/en/products-and-solutions/specialties/knee/persona-iq.html [Accessed August 20,2021]

  • Google Scholar