About this Research Topic
Using computer-based techniques especially artificial intelligence (AI) methods became common in the healthcare field. Many physicians get help from these technologies. AI in healthcare is used to improve diagnosis precision. It also increases the physicians’ capacity to perform a large number of diagnoses in the shortest time. The accuracy of the used AI is getting higher from day to day, especially in image-based diagnosis. This includes all types of cancer detection (e.g. lung, brain, breast, or skin) and using all types of medical images or a combination of images and textual data. Many techniques and technologies are proposed, including machine learning and deep learning algorithms, Case-Based Reasoning (CBR) systems, and embedded machine learning tools.
Most of the proposed AI computer vision techniques for healthcare are based on pre-trained Convolutional Neural Nets (CNN). It is true that recent CNNs perform very well and have high accuracies, but they present several limits. The first limit is the huge volume of datasets that they need for training. This in turn, lead to a need for complex and often expensive hardware material to perform the training in a reasonable time. Many solutions are being proposed to overcome this disadvantage, such as the use of optimization approaches while constructing the model. Another limitation of CNN is the lack of interpretability, which is even more critical for healthcare. To improve the confidence of patients and doctors in the technology, the system must be transparent and should be able to return an explanation of the returning results. This could be realized by adding a CBR system for the classification step, or another explainability/interpretability solution. Patients' health records are confidential, and many hospitals and healthcare services are almost forced to share data (often anonymized) with the AI company they work with. However, anonymization is not enough to guarantee confidentiality. The data privacy issue can be solved if the health institutions have embedded hardware AI tools that permit them to realize all the operations an AI software product can offer. Other techniques that integrate homomorphic encryption allow ensuring data privacy.
This Research Topic calls for scientific contributions as well as industrial experiences in applying computer vision to healthcare. Topics of interest include:
● Machine/Deep learning for cancer detection and prediction based on medical images.
● Machine/Deep learning for medical images classification
● Case-Based-Reasoning for automatic diagnosis with medical images.
● Medical signal and image processing techniques
● Computer-assisted diagnosis including images.
● Precision medicine based on images or videos.
● Models for human-device interaction for medicine
● Telemedicine
Most of the proposed AI computer vision techniques for healthcare are based on pre-trained Convolutional Neural Nets (CNN). It is true that recent CNNs perform very well and have high accuracies, but they present several limits. The first limit is the huge volume of datasets that they need for training. This in turn, lead to a need for complex and often expensive hardware material to perform the training in a reasonable time. Many solutions are being proposed to overcome this disadvantage, such as the use of optimization approaches while constructing the model. Another limitation of CNN is the lack of interpretability, which is even more critical for healthcare. To improve the confidence of patients and doctors in the technology, the system must be transparent and should be able to return an explanation of the returning results. This could be realized by adding a CBR system for the classification step, or another explainability/interpretability solution. Patients' health records are confidential, and many hospitals and healthcare services are almost forced to share data (often anonymized) with the AI company they work with. However, anonymization is not enough to guarantee confidentiality. The data privacy issue can be solved if the health institutions have embedded hardware AI tools that permit them to realize all the operations an AI software product can offer. Other techniques that integrate homomorphic encryption allow ensuring data privacy.
This Research Topic calls for scientific contributions as well as industrial experiences in applying computer vision to healthcare. Topics of interest include:
● Machine/Deep learning for cancer detection and prediction based on medical images.
● Machine/Deep learning for medical images classification
● Case-Based-Reasoning for automatic diagnosis with medical images.
● Medical signal and image processing techniques
● Computer-assisted diagnosis including images.
● Precision medicine based on images or videos.
● Models for human-device interaction for medicine
● Telemedicine
Keywords: Medical imaging, Automatic diagnosis, Artificial Intelligence, Computer vision, Anomaly detection
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
