Abstract
Background and aim:
Neurodegenerative disorders (e.g., Alzheimer’s, Parkinson’s) lead to neuronal loss; neurocognitive disorders (e.g., delirium, dementia) show cognitive decline. Early detection is crucial for effective management. Machine learning aids in more precise disease identification, potentially transforming healthcare. This comprehensive systematic review discusses how machine learning (ML), can enhance early detection of these disorders, surpassing traditional diagnostics’ constraints.
Methods:
In this review, databases were examined up to August 15th, 2023, for ML data on neurodegenerative and neurocognitive diseases using PubMed, Scopus, Google Scholar, and Web of Science. Two investigators used the RAYYAN intelligence tool for systematic reviews to conduct the screening. Six blinded reviewers reviewed titles/abstracts. Cochrane risk of bias tool was used for quality assessment.
Results:
Our search found 7,069 research studies, of which 1,365 items were duplicates and thus removed. Four thousand three hundred and thirty four studies were screened, and 108 articles met the criteria for inclusion after preprocessing. Twelve ML algorithms were observed for dementia, showing promise in early detection. Eighteen ML algorithms were identified for Parkinson’s, each effective in detection and diagnosis. Studies emphasized that ML algorithms are necessary for Alzheimer’s to be successful. Fourteen ML algorithms were discovered for mild cognitive impairment, with LASSO logistic regression being the only one with unpromising results.
Conclusion:
This review emphasizes the pressing necessity of integrating verified digital health resources into conventional medical practice. This integration may signify a new era in the early detection of neurodegenerative and neurocognitive illnesses, potentially changing the course of these conditions for millions globally. This study showcases specific and statistically significant findings to illustrate the progress in the area and the prospective influence of these advancements on the global management of neurocognitive and neurodegenerative illnesses.
Introduction
Machine learning (ML) describes circumstances in which machines can mimic human minds in learning and analysis and thus be used to solve problems (1). Recent advances in ML have produced a computational framework by integrating a multitude of patient data and providing unique risk assessments and recommendations to each patient, which has the potential to revolutionize clinical decision-making (2) fundamentally.
Helping with diagnosis is one of the most significant uses of machine learning in this field. The promise of machine learning-based disease diagnosis (MLBDD), which is affordable and time-effective, is demonstrated by numerous researchers and practitioners (2). To identify chronic kidney disease, Ma et al. (2020) suggested a heterogeneous modified artificial neural network (HMANN) model that obtained an accuracy of 87–99% (3). To improve the diagnosis of COVID-19, Apostolopoulos and Mpesiana (2020) used a CNN-based Xception model on an imbalanced dataset of 284 COVID-19 and 967 non-COVID-19 patient chest X-ray images and achieved 89.6% accuracy in diagnosis (4). Regarding the diagnosis of diabetes, Yahyaoui et al. (2019) showed that the machine-learning RF technique works with an accuracy of 83.67% (5). The examples demonstrate how machine learning algorithms can provide more accurate and reliable disease diagnosis than other diagnostic techniques.
Neurodegenerative disorders are characterized by a gradual loss of neurons, often leading to death. The term covers a wide range of clinical diseases and progressive dementing conditions, including Alzheimer’s disease (AD), Parkinson’s disease (PD), and a number of other neurological disorders (6). Neurocognitive disorders, including delirium, mild cognitive impairment and dementia, are characterized by a decrease in cognitive functioning from a previously attained level (7). Many of these diseases are incurable and sometimes fatal, but early detection can significantly improve the ability to control them.
AD is the most prevalent form of dementia. Patients with AD have trouble remembering things, which limits their ability to learn. Due to the slow progression of AD and the difficulty of current diagnostic techniques in identifying it in its early stages, early diagnosis of the disease is crucial.
PD is a progressive and chronic neurodegenerative disease. The overall validity of PD’s clinical diagnosis, particularly in the early stages of the disease, is unsatisfactory (8).
Delirium is acute brain dysfunction that causes cognitive impairment and shifting attention. Numerous symptoms, such as significant psychomotor agitation, a low level of consciousness, or both, may be present. Traditionally, one or more physicians’ evaluations have been used to diagnose delirium clinically. However, this method of diagnosis might contain flaws because of the disease’s unstable nature (9).
As evident, standard clinical diagnostic techniques for neurodegenerative and neurocognitive diseases have flaws, which make it difficult and occasionally impossible to diagnose the disease, especially in its early stages. On the other side, machine learning algorithms can be highly accurate when it comes to diagnosing a variety of diseases. Recently, many studies have been conducted on the efficacy of ML algorithms as a quick and reliable alternative diagnostic method. Therefore, in this article, we aimed to systematically assess different uses of ML algorithms in detecting neurodegenerative and neurocognitive disorders early.
Methods
This systematic review study was conducted as stated by Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA2020) principles (10). This review has been registered on The Open Science Framework (OSF) (registration DOI https://osf.io/rtsyk/).
Information sources, search strategy
A comprehensive search of several databases was conducted from inception to August 15th, 2023. The databases included PubMed/MEDLINE, Scopus, Google Scholar and Web of Science. As seen in table 1, the search for AI algorithms used for detecting and screening neurodegenerative and neurocognitive diseases involved a controlled vocabulary supplemented with keywords in each database. Table 1 demonstrates the specific search syntax used for each database involved.
Table 1
| Database | Search strategy |
|---|---|
| PubMed | “deep learning”[Title/Abstract] OR “support vector machine”Title/Abstract] OR “Machine learning”[Title/Abstract] OR “supervised machine learning”[Title/ Abstract] OR “unsupervised machine learning”[Title/Abstract] OR “Machine learning”[MeSH Terms] OR “supervised machine learning”[MeSH Terms| OR “unsupervised machine learning”[MeSH Terms]) AND (“neurodegenerative disorders”[Title/Abstract] OR “neurodegenerative disease”[Title/Abstract] OR “neurodegenerative conditions”[Title/Abstract] OR “ALS”[Title/ Abstract] OR “Huntington’s disease”[Title/ Abstract] OR “Tauopathies”[Title/Abstract] OR “neurofibrillary tangles”|Title/Abstract] OR “myelitis”[Title/Abstract] OR “paraneoplastic polyneuropathy”|Title/ Abstract] OR “paraneoplastic cerebellar degeneration”|[Title/Abstract] OR “Tourette syndrome”|Title/Abstract] OR “neurofibromatoses”[Title/Abstract] OR “Encephalopathy”|Title/Abstract] OR “neuropathy”[Title/Abstract] OR “brain degeneration”[Title/Abstract] OR “CNS neurodegenerative disease”[Title/Abstract] OR “CNS degenerative disorder”[Title/ Abstract] OR “neurodegenerative diseases”[MeSH Terms] OR “neurocognitive diseases”[Title/Abstract] OR “neurocognitive disorders”[Title/ Abstract] OR “neurocognitive conditions”[Title/Abstract] OR “mild cognitive impairment”[Title/Abstract] OR “organic brain disorder” Title/Abstract] OR “acquired cognitive dysfunction”[Title/ Abstract] OR “dementia”|Title/Abstract] OR “delirium”|[Title/Abstract] OR “mild neurocognitive disorders”[Title/Abstract] OR “major neurocognitive disorders”[Title/ Abstract] OR “neurocognitive disorders”|MeSH Terms]) |
| WOS | ((TS = (“machine learning”)) OR TS = (“allgobrithem”)) OR TS = (“artificial inteligency”)| And ((TS = (detection)) OR TS = (“early detection”)) OR TS = (Diagnosis)) OR TS = (identification)) OR TS# (recognation)) OR TS = (observation)) OR TS = (“early diagnosis”) And ((((((((((((((((TS = (“Neurodegenerative disease*”)) OR TS-(ALS)) OR TS= (“Huntington”)) OR TS = (“Tauopathies and the subclassifications”)) OR TS= (Tauopathie*)) OR TS = (“Neurofibrillary tangle*”)) OR TS = (Myelitis)) OR TS= (“Paraneoplastic polyneuropathy”)) OR TS = (“Paraneoplastic cerebellar degeneration*”) OR TS = (“Tourette syndrome”)) OR TS = (Neurofibromatosis)) OR TS = (Encephalopathy)) OR TS = (Neuropathy)) OR TS = (“neurocognitive disorder* “)) OR TS = (“senile dementia”)) OR TS = (“Creutzfeldt-Jakob disease”)) OR TS = (“Diffuse Lewy body disease”)) OR TS = (“Multiple sclerosis”)) OR TS- (“Normal pressure hydrocephalus”)) OR TS = (“Pick disease*’)) OR TS = (amnesia)) OR TS = (“cognitive dysfunction”)) OR TS = (“consciousness disorder* “)) OR TS-(delirium)) OR TS = (dyslexia) |
| Scopus | TITLE-ABS-KEY (detection) OR TITLE-ABS-KEY (“early detection”) OR TITLE-ABS-KEY (diagnosis) OR TITLE-ABS-KEY (identification) OR TITLE-ABS-KEY (recognation) OR TITLE-ABS-KEY (observation) OR TITLE-ABS-KEY (“early diagnosis”)) AND (TITLE-ABS-KEY (“machine learning”) OR TITLE-ABS-KEY (“allgohrithem”) OR TITLE-ABS-KEY (“artificial inteligency”)) AND (TITLE-ABS-KEY (“Neurodegenerative disease”) OR TITLE-ABS-KEY (als) OR TITLE-ABS-KEY (“Huntington”) OR TITLE-ABS-KEY (“Tauopathies and the subclassifications”) OR TITLE-ABS-KEY (tauopathie) OR TITLE-ABS-KEY (“Neurofibrillary tangle”) OR TITLE-ABS-KEY (myelitis) OR TITLE-ABS-KEY (“Paraneoplastic polyneuropathy”) OR TITLE-ABS-KEY (“Paraneoplastic cerebellar degeneration “) OR TITLE-ABS-KEY (“Tourette syndrome”) OR TITLE-ABS-KEY (neurofibromatosis) OR TITLE-ABS-KEY (encephalopathy) OR TITLE-ABS-KEY (neuropathy) OR TITLE-ABS-KEY (“neurocognitive disorder “) OR TITLE-ABS-KEY (“senile dementia”) OR TITLE-ABS-KEY (“Creutzfeldt-Jakob disease”) OR TITLE-ABS-KEY (“Diffuse Lewy body disease”) OR TITLE-ABS-KEY (“Multiple sclerosis”) OR TITLE-ABS-KEY (“Normal pressure hydrocephalus”) OR TITLE-ABS-KEY (“Pick disease”) OR TITLE-ABS-KEY (amnesia) OR TITLE-ABS-KEY (“cognitive dysfunction”) OR TITLE-ABS-KEY (“consciousness disorder “) OR TITLE-ABS-KEY (delirium) OR TITLE-ABS-KEY (dyslexia)) |
Search strategies and databases used in the study.
Data screening and eligibility criteria
We used the RAYYAN intelligent tool for systematic reviews to screen the search results (11). Titles and abstracts from 7,069 articles obtained from our search strategy were independently and blindly screened by six reviewers (Zh.M., Sh.K., H.D., A.A., H.B., M.Y.). The duplicate records were removed using the same tool. The conflicts were resolved by a seventh reviewer (Sh.K.) using RAYYAN’s compute rating feature.
Inclusion criteria
The study was conducted on this specified list of neurodegenerative and neurocognitive diseases, and the search keywords included items below:
Huntington
Tauopathies and the subclassifications
Neurofibrillary tangles
Myelitis
Paraneoplastic polyneuropathy
Paraneoplastic cerebellar degeneration
Tourette syndrome
Neurofibromatoses
Encephalopathy
Neuropathy
ALS
Alzheimer’s disease (AD)
Mild cognitive impairment (MCI)
Parkinson’s disease (PD)
Frontotemporal dementia (FTD)
Lewy Body’s disease (LBD)
Progressive supranuclear palsy (PSP)
Corticobasal degeneration (CBD)
Wernicke-Korsakoff syndrome
Normal pressure hydrocephalus (NPH)
Prion diseases, such as Creutzfeldt-Jakob disease
Vascular dementia
Studies that were not available as open access were in languages other than English were conducted on animals, and were published as book chapters, Conference papers were excluded.
Quality assessment of included studies
Two assessors (MY and HD) evaluated each study separately based on the Cochrane risk of bias tool, evaluating all included studies (12). With a focus on six domains—sequence generation, allocation concealment, blinding, incomplete data, and selective reporting—the Cochrane risk of bias tool is a widely used and standard tool that contains all the necessary questions to evaluate methodological quality and bias risk. The two assessors settled other biases and disagreements through discussion and consensus.
Results
Study selection
Our search strategies in four databases yielded 7,069 studies, 1,365 were eliminated as duplicates. At least two individuals screened each of 4,334 remaining studies through title and abstract. Unrelated studies whose full text was unavailable, did not meet our inclusion criteria, and were not in English were excluded. At last, 108 studies were included for interpretation. Figure 1 depicts the study selection procedure.
Figure 1
Study characteristics
The included studies were published between 2015 and 2023. A study was carried out in Africa, another in Australia, 17 in Europe, 29 in America, and the remaining in Asia.
Findings
In the included studies, 3,723,329 participants were examined. Thirty-four studies on AD, 14 on PD, 13 on MCI, 10 on dementia, 7 on MS and the remaining studies were carried out on other neurodegenerative and neurocognitive disorders.
Dementia
In 10 studies conducted on dementia, 12 ML algorithms were used: XGBoost classification, Binary logistic regression (LR), A logistic model tree classifier combined with information gain feature selection, 3D convolutional neural networks (3D CNN), k-NearestNeighbor (kNN), support vector machine (SVM), random forest (RF), parallel recurrent convolutional neural network (PRCNN), support vector machine classifiers (SVC), support vector regression (SVR), partial least squares regression (PLSR) and Deep Neural Network (DNN), All of which showing promising results in early detection and screening of the disease. Table 2 summarizes our included studies.
Table 2
| Author | Year | Country | Aim of study | Population | Type of pathology | Used ML algorithm | Outcome | Conclusion |
|---|---|---|---|---|---|---|---|---|
| Raymond Gao et al. (28) | 2023 | USA | The present study aims to construct Polygenic Risk Scores (PRSs) for Alzheimer’s disease (AD) risk and Age at Onset (AAO). Additionally, it seeks to develop Machine Learning models for predicting AD risk and explore feature importance, including PRSs, conventional risk factors, and ICD-10 codes extracted from Electronic Health Records (EHRs). | 457,936 participants | Alzheimer’s disease | XGBoost models | The study’s primary finding highlights the greater significance of Polygenic Risk Scores (PRSs) derived from Alzheimer’s disease (AD) risk and Age at Onset (AAO) compared to age alone in predicting AD. Additionally, the Machine Learning model identifies key predictors from Electronic Health Records (EHRs), such as urinary tract infection, syncope and collapse, chest pain, disorientation, and hypercholesterolemia, as crucial factors in the development of AD. | In conclusion, to tackle the broader issue of AD early detection, this study not only discovered critical traits for developing AD but also developed powerful explainable ML models. |
| Fayemiwo et al.(29) | 2023 | National Health and Aging Trends Study database | In this study, nine distinct experiments were conducted to determine which responses (either SP’s or proxy’s) in the “word-delay,” “tell-words-you-can-recall,” and “immediate-word-recall” tasks are essential in the prediction of dementia cases, and to what extent the combination of these two responses is helpful in the prediction of dementia. | National Health and Aging Trends Study (NHATS) was drawn from a nationally representative survey of Medicare recipients between the ages of 65 and older. | Alzheimer’s disease | Four ML algorithms (K-nearest neighbors (KNN), decision tree, random forest, and artificial neural networks (ANN)) were used | In the first scenario of experiments using “word-delay” cognitive assessment, the highest sensitivity (0.60) was obtained from combining the responses from both SP and proxies trained KNN, random forest, and ANN models. Also, in the second scenario of experiments using the “tell-words-you-can-recall” cognitive assessment, the highest sensitivity (0.60) was obtained by combining the responses from both SP and proxies trained KNN model. From the third set of experiments performed in this study on the use of “Word-recall” cognitive assessment, it was equally discovered that the use of combined responses from both SP and proxies trained models gave the highest sensitivity of 1.00 (as obtained from all the four models). | t can be concluded that the combination of responses in a word recall task as obtained from the SP and proxies in the dementia study (based on the NHATS dataset) is clinically useful in predicting dementia cases. Also, the use of “word-delay” and “tell-words-you-can-recall” cannot reliably predict dementia as they resulted in poor performances in all the developed models, as shown in all the experiments. |
| Bhandari et al.(30) | 2023 | India | This study employed explainable artificial intelligence (XAI) techniques to identify the significant set of gene features contributing to diagnosis and integrated gene expression data from various sources to diagnose Parkinson’s disease (PD) using Machine Learning (ML) based methods. | Parkinson’s disease | Based on the findings, it may be helpful to employ XAI when making early treatment decisions for Parkinson’s disease. | The study showcased a robust blood-based gene expression classification for Parkinson’s disease (PD) and healthy controls. By integrating PD datasets from various studies, the analysis gained reliability. Support Vector Machine (SVM) consistently outperformed other machine learning methods, and combining LASSO feature selection with Logistic Regression (LR) and SVM yielded the highest diagnostic accuracy. | ||
| Li et al. (31) | 2023 | USA | This research aims to explore machine learning techniques for early detection of Alzheimer’s disease (AD) and related dementias (ADRD) utilizing actual electronic health records (EHRs). | A total of 23,835 ADRD and 1,038,643 control patients | Alzheimer’s disease and related dementias | Various machine learning algorithms, including random forest and support vector machine, were initially tested on a smaller subset of the entire population. For subsequent experiments, logistic regression was chosen as the baseline, and Gradient Boosted Trees (GBTs), exhibiting the best performance in small-scale experiments, was selected for further analysis. | The best outcomes were obtained by the gradient boosting tree (GBT) models that were trained using the data-driven methodology and also a number of important clinical and sociodemographic factors were identified. | The study examined multiple cohorts of individuals with ADRD and associated dementias, identifying significant similarities and variability in predictions. The models that are being presented help identify ADRD early on and direct research and clinical trial recruitment. |
| Ostertag et al. (32) | 2023 | France | This study introduces a machine learning approach utilizing multimodal data (brain MRI and clinical information) from initial medical visits to predict long-term cognitive decline in patients. | 229 subjects for training, 76 subjects for validation, and 76 subjects for testing | Alzheimer’s disease and Parkinson’s disease | This study introduces an adaptable deep neural network architecture for making long-term prognoses on the progression of neurological diseases, identifying high-risk individuals. | The model demonstrates effective long-term predictions of cognitive decline from any pair of early visits, even without a fixed time delay between them. | The model effectively predicts long-term cognitive decline with only two visits, accommodating irregular intervals. It also demonstrates successful knowledge transfer from Alzheimer’s to Parkinson’s, making it applicable to less studied diseases. |
| Ponce de Leon-Sanchez et al. (33) | 2023 | Mexico | The paper introduces a deep learning model, utilizing an artificial neural network with a single hidden layer, for predicting the diagnosis of multiple sclerosis. | 99 with MS and 45 healthy controls | Multiple Sclerosis | K-Neighbors (KN), Gaussian Naive Bayes (GNB), C-Support Vector (CSV) Decision Tree (DT). Recursive Feature Elimination with Cross-Validation (RFECV) Deep Learning models Neural Networks | Feature selection was optimized based on accuracy, with the model achieving the highest accuracy using 35 features. The remaining 39 features were excluded, enhancing the efficiency of all compared classifiers. | Researchers propose an ANN model using 35 genetic features for MS diagnosis, outperforming conventional methods with high accuracy. The study underscores the potential clinical application of the ANN model in predicting MS based on genetic features, improving accuracy and enabling the emergence of new preventive treatments. |
| Russo et al. (34) | 2023 | Italy | The goal of the study was to develop a gait pattern involving particular spatial and temporal metrics that could be used to consistently differentiate between patients with Parkinson’s disease (PD) and those without mild cognitive impairment (MCI) through the use of supervised machine learning. | 80 participants | Mild Cognitive Impairment and Parkinson’s Disease | Decision Tree (DT) Random Forest (RF) Naïve Bayes (NB) Support Vector Machine (SVM) K-Nearest Neighbor (KNN) | SVM and RF showed the best performance and detected MCI with an accuracy of over 80.0%. | The study demonstrates a robust relationship between gait dysfunction and Parkinson’s disease (PD)-related mild cognitive impairment (MCI). Notably, even on an independent dataset, selected gait parameters work well in machine learning methods for PD-MCI detection. By selecting homogeneous individuals, testing on an external patient group, and expanding the sample size, the research overcame earlier constraints to support these gait features as potential surrogate biomarkers for cognitive impairment in Parkinson’s disease (PD). |
| Syam et al. (35) | 2023 | India | The aim of this research was to propose a machine learning-based framework for accurate detection of Parkinson’s Disease (PD), Huntington’s Disease (HD), and Amyotrophic Lateral Sclerosis (ALS) from gait signals in both binary and multi-class detection environments. | ? | Corticobasal Syndrome (CS), Huntington’s Disease (HD), Dementia, Amyotrophic Lateral Sclerosis (ALS), Progressive supranuclear palsy (PSP) and Parkinson’s Disease (PD) | The study proposes an ensemble framework named Ultaboost, utilizing Naïve Bayes and Logistic Regression, empowered by adaptive boosting principles such as Adaboost. Tested on prominent gait signal features obtained through feature selection techniques (IFS, ILFS, SFS), the ensemble framework addresses class imbalance with SMOTE. | In a multi-class environment, Infinite Feature Selection outperforms Infinite Latent Feature and Sigmis feature selection in detecting Parkinson’s and Huntington’s Disease from gait signal features | Using the UltraBoost ensemble framework, the paper presents a machine learning system that uses Naive Bayes and Logistic Regression to accurately detect Parkinson’s disease (PD), Huntington’s disease (HD), Amyotrophic Lateral Sclerosis (ALS), and Controls in binary and multi-class scenarios. Interestingly, the approach, which focuses on a small number of gait factors, performs well in binary classifications but has certain difficulties in multi-class environments, which are mostly related to class imbalance. |
| Tan et al. (36) | 2023 | Singapore | This study aims to develop a reliable machine learning (ML) model using socio-demographics, vascular risk factors, and structural neuroimaging markers for early diagnosis of cognitive impairment in a multi-ethnic Asian population. | 911 participants | Cognitive Impairment | Logistic regression (LR), support vector machine (SVM), and gradient boosting machine (GBM). | The ensemble model demonstrated strong performance and it outperformed individual classifiers. Important predictors of cognitive impairment included age, ethnicity, highest education attainment, and neuroimaging markers. | The study demonstrates how machine learning techniques may be used to combine many data domains for precise early detection of cognitive impairment. In a population-based context, the model is scalable and makes use of characteristics that are easily accessible for the purpose of screening people who are at high risk of developing dementia. |
| Tayyab et al. (37) | 2023 | Canada | Using machine learning algorithms that can handle uncertain labels improves predictions when a substantial number of subjects have unknown outcomes in the dataset. | 142 participants | Multiple sclerosis | Random Forest The study utilized three approaches, including a classic Random Forest (RF), to handle uncertain data points. | The Probabilistic Random Forest outperformed traditional Random Forest models, achieving the highest AUC. | In datasets with a significant number of subjects having unknown outcomes, employing machine learning algorithms that can model label uncertainty enhances predictive performance. |
| Tena et al. (38) | 2023 | Spain | The primary objective of this paper is to introduce a novel methodology for the early automated diagnosis of this dysfunction, surpassing the timing capabilities of clinicians. | 45 ALS participants and 18 control subjects | Amyotrophic lateral sclerosis (ALS) | Five supervised classification models (RF, LR, LDA, NN, SVM) were implemented in R with standardized features using 10-fold cross-validation. | The Random Forest model achieved high accuracy, sensitivity, and specificity for classifying bulbar vs. control participants. Due to uncertainty in ALS patients without bulbar involvement, a semi-supervised SVM was used, resulting in improved performance. The model outperformed clinicians and existing methods, showcasing its efficacy in diagnosing bulbar dysfunction. | The obtained outcomes highlight the efficacy and feasibility of the approach suggested in this research. This strategy may lead to the development of an affordable and user-friendly instrument for the early identification and tracking of bulbar dysfunction in the early phases of the illness. |
| Mueller et al. (39) | 2023 | USA | The objective was to use electronic health records to find a risk estimation model for common delirium in patients being moved from emergency departments to inpatient units that would be therapeutically useful. | 8,057 positive Delirium of total 28,531 participants | Delirium | Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Gradient Boosting Machine (GBM), Support Vector Machine (SVM), and K Nearest Neighbor (KNN). | Gradient Boosting Machine (GBM) had the best performance. | These findings will help design management or preventative measures. |
| Swarnalatha et al. (40) | 2023 | UAE | The current study aims to apply a novel deep feature that offers the best solution for EEG signal analysis and severity determination. | Alzheimer’s disease | A novel sandpiper-based recurrent neural system (SbRNS) has been developed to predict Alzheimer’s disease early stage. | The scheme that proposed by study had better performance that other same schemes. | Using EEG signals, a new SbRNS in MATLAB efficiently detects AD severity, exceeding traditional methods in terms of accuracy and performance. | |
| Ahmed et al. (41) | 2022 | Egypt | The study suggests utilizing various modalities of Alzheimer’s disease brain images for the early identification of the disease in this publication. | 300 individuals | Alzheimer’s disease (AD) | XGB CNN | The study achieved high accuracy, specificity and sensitivity in both early and late fusion. | The proposed model utilizes Laplacian Re-Decomposition for image fusion, combining data from MRI and PET modalities with XGBoost (XGB) to enhance early Alzheimer’s disease diagnosis, demonstrating superior performance compared to Naive Bayes (NB), Decision Trees (DT), Support Vector Machine (SVM), and Random Forest (RF) methods. |
| Ahmed et al. (42) | 2022 | Egypt | This paper’s primary goal is to detect and diagnose AD using SNP biomarkers that have a high degree of early classification accuracy. | 1,569 subjects from 2 databases | Alzheimer’s disease | Boruta FS algorithm Gradient Boosting | The suggested method can be preferred for AD early detection. | |
| García-Gutierrez et al. (43) | 2022 | Spain | This paper has presented the design and implementation of a machine learning–based framework for the automatic diagnosis, especially, of neurodegenerative diseases. | 329 patients | Alzheimer’s disease | EG or Bayesian classifiers | The results showed the potential of the framework for early and automated diagnosis with neuroimages and neurocognitive assessments from patients with Alzheimer’s disease (AD) and frontotemporal dementia (FTD). | The tool is presented from a XAI perspective to assist clinicians in diagnosis, as it offers all the necessary processes for analyzing these datasets, including feature selection via evolution, data preprocessing, and illness modeling. |
| Kavitha et al. (44) | 2022 | India, Iraq and Colombia | In this research, individuals affected by Alzheimer’s Disease are identified, and the objective is to detect individuals who may potentially have Alzheimer’s at an early stage. | MRI data from 150 patients aged | Alzheimer’s disease | Several techniques include Decision Tree, Random Forest, Support Vector Machine, Gradient Boosting, and Voting classifiers. | The suggested method shows excellent results, with a high average accuracy in validation. | Addressing Alzheimer’s involves risk reduction, early intervention, and accurate diagnosis. Future work will focus on improving detection techniques by extracting new features, eliminating irrelevant ones, and integrating metrics like MMSE and Education for enhanced accuracy. |
| Kumar et al. (45) | 2022 | India | The primary individual speech characteristics for dementia recognition are examined in this study. The primary contribution is the discovery of a small group of speech characteristics that improve the process of identifying dementia. The study also uses deep learning (DL) and machine learning (ML) models for efficient recognition. | A total of 442 subjects | Dementia | The study explores deep learning models, including artificial neural networks (ANN) CNN, RNN (GRU and LSTM), and a parallel recurrent convolutional neural network (PRCNN) for dementia recognition based on speech features. | Machine learning models are better than Deep learning models in diagnosing Dementia using speech characteristics. | The proposed approach shows promising results when compared to existing works on dementia recognition through speech analysis. |
| Li et al. (46) | 2022 | USA | In order to create feature subsets more effectively, the study presents the Ontology-guided Attribute Partitioning (OAP) approach, which considers domain-specific associations between features. These more effectively partitioned feature subsets are used in the study to create OAP-Ensemble Learning (OAP-EL), an ensemble learning framework. | 276 very preterm infants | Cognitive Deficits | Ontology-guided Attribute Partitioning ensemble learning (OAP-EL) model | The proposed machine learning model had proficient results in early prediction of cognitive deficits at 2 years corrected age in very preterm infants. | The study introduced an advanced ensemble learning model for early predicting cognitive deficits in preterm infants, outperforming traditional methods. Future work includes exploring ontology-aided machine learning to understand brain features better. |
| Liu et al. (47) | 2022 | USA | The aim was to create and evaluate a precise deep-learning model for detecting adult hospitalized patients’ new-onset delirium. | A total of 331,489 CAM (confusion assessment method) assessments from 34,035 patients with 39,567 encounters were included in the final dataset. | Delirium | Logistic regression, random forest, support vector machine, and LightGBM models were developed with the training set and evaluated using 1,000-round bootstrapping on the testing dataset. | The LightGBM model showed the best performance, and by combining the LightGBM model with the LSTM, the model significantly improved performance. 20 features were identified using mean absolute SHAP values. | By combining the temporal trend-capturing abilities of LSTM with the LightGBM model’s predictive power, the model enhances the accuracy of predicting new-onset delirium. |
| Mehra et al. (48) | 2022 | India | The primary objective of this research is to precisely categorize people as either healthy or have Parkinson’s disease to create an efficient machine learning-based healthcare model. The most essential features for categorization are extracted with the primary goal in mind. | 73 healthy subjects and 93 PD | Parkinson’s disease | The study utilizes a step regression-based approach for feature selection to enhance the classification of Parkinson’s disease (PD). The model is applied to three publicly accessible Parkinson’s datasets from diverse studies on Psyionet, all featuring Vertical Ground Reaction Force (VGRF) recordings from eight sensors under each foot. | The proposed model, incorporating effective pre-processing, feature extraction, and feature selection methods, achieved high accuracy when applied to three datasets. | This study introduces a superior machine learning-based Parkinson’s disease diagnosis model using wearable sensors. |
| Nelson et al. (49) | 2022 | USA | The method involves incorporating domain knowledge into clinical classifications by embedding individual patient data into a biomedical knowledge graph. | The study involved EHR data from 2,180,882 patients, patients with confirmed MS diagnoses 5,752 and control group (non-MS) consisted of 2,175,130. | Multiple Sclerosis (MS) | The Page rank algorithm was modified to embed millions of deidentified EHRs into a biomedical knowledge graph (SPOKE). This resulted in high-dimensional, knowledge-guided patient health signatures (i.e., SPOKEsigs) that were subsequently used as features in a random forest environment to classify patients at risk of developing chronic diseases. | The model successfully predicted the disease status of 5,752 subjects 3 years before their multiple sclerosis (MS) diagnosis | SPOKEsigs utilize Electronic Health Record (EHR) data to characterize patients clinically and biologically. The study demonstrates a clinical use case for detecting multiple sclerosis (MS) up to 5 years before the documented diagnosis, highlighting the distinctive biological features of the prodromal MS state. |
| Penfold et al. (50) | 2022 | USA | The goal was to create a natural language processing system and prediction model to identify Mild Cognitive Impairment (MCI) from clinical text. | 4,185 patients from two datasets | Mild cognitive impairment (MCI) and | The study used a LASSO logistic regression approach to create a prediction model for MCI identification, incorporating NLP-derived concepts and demographic variables. | The prediction model showed modest performance in the validation dataset. Using a cutoff of 0.60, the classifier demonstrated low sensitivity and high specificity. | Though with low sensitivity, the model demonstrated a high negative predictive value, essential for population-based screening. It is comparable to widely used clinical tests. |
| Qiu et al. (51) | 2022 | Singapore | The study aimed to create a deep learning model, specifically a graph convolutional and recurrent neural network (graph-CNN-RNN), utilizing a series of brain structural MRI data to predict AD conversion based on age before clinical diagnosis. | 2,489 subjects from two datasets. | Alzheimer’s disease | This study employed a deep learning longitudinal model, graph convolutional and recurrent neural network. (graph-CNN-RNN) | The graph-CNN-RNN accurately predicted the conversion of Alzheimer’s disease up to 4 years ahead of time with high reliability, and it consistently delivered accurate diagnoses of the condition at all periods. | The graph-CNN-RNN offered a detailed quantitative trajectory of brain morphology, from the early prognosis to the advanced stages of Alzheimer’s disease (AD). |
| Revathi et al. (52) | 2022 | India | The study aims to enhance prediction techniques and assess the cognitive function of individuals with potential dementia using the Cognitive Ability Test (CAT). | 2,361 patients | This study utilized Support Vector Machine (SVM), Random Forest algorithm, and Multinomial Logistic Regression algorithm as machine learning techniques. | The first-stage classifier, using a Support Vector Machine (SVM), achieves a prediction accuracy of 86%, while the Random Forest classifier attains an accuracy of 71%. In the second stage, the Multinomial Logistic Regression algorithm for cognitive assessment achieves an accuracy of 89%. | The proposed work facilitates early prediction of individuals at risk of Alzheimer’s Disease through the analysis of clinical data. | |
| Riad et al. (53) | 2022 | France and Belgium | The study aimed to predict clinical performance in Huntington’s Disease (HD), an inherited neurodegenerative disease, using machine learning applied to brief speech recordings. | 103 individuals | Huntington’s disease | The study utilized the auto-machine-learning system, auto-sklearn, to predict clinical variables based on speech features. Auto-sklearn employs Bayesian optimization algorithms to identify the model with the optimal cross-validated performance on the training set. | Combining speech features with demographic information enabled the prediction of individual cognitive, motor, and functional scores. | In conclusion, this pioneering machine learning model and speech analysis accurately estimated classical scale scores for both pre-HD individuals and HD participants. |
| Schumann et al. (54) | 2022 | Germany | The study aimed to identify the optimal method for assessing fall risk by analyzing 11 gait datasets. It employed a new feature selection ensemble (FS-Ensemble) and four classification models—Gaussian Naive Bayes, Decision Tree, k-nearest Neighbor, and Support Vector Machine. | 1,240 participants | Multiple Sclerosis | Gaussian Naive Bayes Decision Tree k-Nearest Neighbor Support Vector Machine (SVM) | The Gaussian Naive Bayes emerged as the most effective classification model for detecting falls across nearly all datasets. | The FS-Ensemble proved beneficial in enhancing classification models and is a suitable technique for reducing datasets with numerous features. Subsequent research focusing on additional risk factors, such as fear of falling, may offer further insights. |
| Sun et al. (55) | 2022 | China | A novel computation framework is introduced for predicting Multiple Sclerosis-associated miRNAs. The approach utilizes a network representation model to learn miRNA feature representations and employs a deep learning-based model for predicting Multiple Sclerosis-associated miRNAs. | 102 MS-related miRNAs | Multiple Sclerosis | A convolutional neuron network (CNN)-based model is developed to integrate miRNA features and predict multiple sclerosis (MS)-related miRNAs. The workflow comprises a feature encoder, backpropagation training with dropout, and a Gaussian Naive Bayes (GaussianNB) classifier. | The assessment demonstrates that the proposed model accurately predicts miRNAs linked to Multiple Sclerosis, surpassing several existing methods by a significant margin. | The evaluation confirms that the proposed model outperforms existing methods and accurately predicts Multiple Sclerosis-related miRNAs. |
| Tufail et al. (56) | 2022 | China, Pakistan, Saudi Arabia, Canada | The objective of this study is to employ Convolutional Neural Network (CNN) architectures in both 2D and 3D domains, utilizing positron emission tomography neuroimaging, for the classification of early stages of Alzheimer’s Disease (AD) into AD, Mild Cognitive Impairment (MCI), and Normal Control (NC) classes. | AD 94, MCI 97 and NC 102 | Alzheimer’s Disease | The study applied to transfer and non-transfer learning using 2D and 3D CNNs for binary. Custom 3D CNN architectures were used, and a transfer learning model based on Xception addressed MCI and AD classification. | 3D-CNN architecture had the best performance. Data augmentation also contributed to superior performance in the multiclass classification task. | The outcomes support using deep learning models in the early diagnosis of Alzheimer’s disease. |
| Wang et al. (57) | 2022 | USA | This study introduces a novel Attentive All-level Fusion (AANet) system designed to integrate multi-level and multi-modality patient data (3D brain images, demographics, genetics, and blood biomarkers) into a deep-learning framework for early Alzheimer’s disease diagnosis. | 11,333 valid MRI samples. | Alzheimer’s Disease | AANet incorporates a **Feature Pyramid Network (FPN)** for MRI image representation extraction and a self-attention fusion method to integrate features from various data modalities. | AANet demonstrated remarkable accuracy, surpassing various state-of-the-art methods. AANet presents an advanced methodological framework for disease diagnosis based on multiple modalities. | In summary, AANet exhibits remarkable potential for early Alzheimer’s Disease (AD) detection through its combination of the Feature Pyramid Network model and self-attention all-level fusion. It also provides a flexible framework for deep learning-based multi-modality and multi-level disease diagnosis. |
| Yu et al. (58) | 2022 | USA | The study aims to demonstrate an Alzheimer’s Disease (AD) diagnosis approach by utilizing the surface-enhanced Raman spectroscopy (SERS) fingerprints of human cerebrospinal fluid (CSF). The approach involves combining SERS with a convolutional neural network (CNN) for biomarker detection, specifically to analyze disease-associated biochemical changes in the CSF. | 30 samples | Alzheimer’s Disease | A one-dimensional Convolutional Neural Network (CNN) was employed to process and classify surface-enhanced Raman scattering (SERS) spectral data. | An excellent correlation coefficient was observed between the test score and the Clinical Dementia Rating (CDR) score, indicating the feasibility of detecting Alzheimer’s Disease biomarkers through the innovative combination of Surface-Enhanced Raman Scattering (SERS) and machine learning. | In double-blind tests, a hybrid system combining Convolutional Neural Network (CNN) and Surface-Enhanced Raman Spectroscopy (SERS) demonstrated exceptional reproducibility, achieving 92% accuracy in diagnosing Alzheimer’s disease. Despite being based on a tiny sample size, the SERS neural network exhibits remarkable accuracy, making a biological test for Alzheimer’s diagnosis feasible. Future studies intend to increase patient sample sizes and investigate the possibilities of applying SERS/AI technology for clinical trial applications and early-stage diagnostics. |
| Zhang et al. (59) | 2022 | China | This study aims to explore potential peripheral blood biomarkers for the early diagnosis of PD. | ANIMALS we recused in this study. | Parkinson’s disease (PD) | SVM (Support Vector Machine) kNN (k-Nearest Neighbors) RF (Random Forest) | The study identified three upregulated genes in the peripheral-blood transcriptome datasets of Parkinson’s disease (PD) patients. Further analyses and validation in animal models revealed that SSR1 (Signal Sequence Receptor Subunit 1) was significantly upregulated in both models and negatively correlated with dopaminergic neuron survival. | In brief, this study identifies potential biomarkers for early PD diagnosis and establishes a potential artificial intelligence model for predicting Parkinson’s disease. |
| Valencia et al. (60) | 2022 | Spain | This study examines a technique for generating synthetic T1-weighted (T1-w) pictures from T2-FLAIR images. The study then evaluates the impact of utilizing original and synthetic T1-weighted images on the efficacy of the established approach for identifying longitudinal Multiple Sclerosis (MS) lesions. | 136 subjects | Multiple Sclerosis | Fully Convolutional Network (FCN) | The proposed method can be useful. | The study demonstrates that synthetic images can effectively compensate for data scarcity or substitute for original images. This proves beneficial in standardizing contrast across diverse acquisitions, particularly in developing new algorithms for detecting T2 lesions in Multiple Sclerosis. |
| Adhikari et al. (61) | 2021 | Nepal | The transcripts of AD patients and control normal people were combined to build a novel dataset on low-resource language, namely Nepali, for this work and also provided baselines for the early identification of AD by utilizing a variety of machine learning (ML) and deep learning (DL) methods on a fresh dataset. | 98 CN subjects and 168 AD patients | Alzheimer’s disease | Decision Tree (DT) K-Nearest Neighbors (KNN) Support Vector Machines (SVM) Naïve Bayes (NB) Random Forest (RF) AdaBoost XGBoost (XGB) CNN | best performing model is the attention-based CNN with domain-specific Word2Vec. | For conclusion, the study’s goal is to identify AD in Nepali speakers as soon as possible. This is a step in the right direction for resolving issues with disease identification and providing inspiration for future studies in this area. The main benefit of this automated system is that it predicts the existence of AD much more quickly. |
| Ahmed H et al. (62) | 2021 | Australia | This study aims to create multiple heterogeneous stacked fusion models by leveraging the strengths of various base learning algorithms, aiming to enhance the generalizability and robustness of machine learning models for Alzheimer’s disease diagnosis. The study combines written and spoken-based datasets to train the stacked fusion models. | 1,598 AD patients and 1,628 healthy controls. | Alzheimer’s disease (AD) | Stacked Fusion Models Hybrid Stacked Fusion Model | Hybrid Stacked Fusion Model had better performance than Stacked Fusion Models. | This study recommends replacing the initial conventional screening test with such models that can be embedded into an online format for a completely automated remote diagnosis in light of the achieved performance and improved generalizability of such fusion models over single classifiers. |
| Etminani et al. (63) | 2021 | Italy, Belgium, Sweden, Switzerland, Slovenia, Germany and Netherlands. | This research aims to create and verify a 3D deep learning model using fuorine 18 fuorodeoxyglucose PET (18F-FDG PET) that can predict the final clinical diagnosis of Alzheimer’s disease (AD), dementia with Lewy bodies (DLB), mild cognitive impairment due to Alzheimer’s disease (MCI-AD), and cognitively normal (CN). The model’s performance will be compared to that of multiple expert nuclear medicine physicians. | 757 patients | dementia with Lewy bodies, Alzheimer’s disease, and mild cognitive impairment | The 3D-CNN model (3D convolutional neural networks) is designed with reference to the architecture of VGG16. | The proposed model could achieve high levels of diagnosis and surpass human readers performance. | The final diagnosis of the most prevalent neurodegenerative diseases could be predicted by a 3D deep learning model with just the brain’s 18F-FDG PET, and it performed as well as human readers’ consensus. |
| Herzog et al. (64) | 2021 | UK | This study introduces a data processing pipeline that may be used with standard hardware. It analyzes structural alterations by using brain asymmetry parameters taken from MRI scans. Based on these parameters, the study uses machine learning to classify pathology. | MRI data of 750 subjects from ADNI | Alzheimer’s Disease | The pipeline tested various machine learning methods, focusing on brain asymmetry features enriched with Bag-of-Features for dementia diagnosis. Supervised learning was explored, including Naïve Bayes, Linear Discriminant, Support Vector Machine, and K-Nearest Neighbor. Transfer learning with AlexNet was also considered. | The introduced model was successful in distinguishing between normal cognitive, early mild cognitive impairment and AD patients. | In addition to providing a viable, affordable option for classifying dementia, the suggested pipeline may also prove helpful in treating other brain degenerative diseases that also cause alterations in brain asymmetry. |
| James et al. (65) | 2021 | USA | Can machine learning algorithms accurately predict 2-year dementia incidence in memory clinic patients, and how do these predictions compare with existing models? | 1,568 dementia patients out of 15,307 samples | Dementia | The study implemented four machine learning algorithms, namely logistic regression (LR), support vector machine (SVM), random forest (RF), and gradient-boosted trees (XGB), for a classification task. | Machine learning algorithms outperformed two existing predictive models in predicting incident dementia within a 2-year timeframe. | It is indicated that machine learning algorithms can effectively predict the occurrence of dementia within 2 years. |
| Kleiman et al. (66) | 2021 | USA | The goal is to find optimized cognitive assessment features for detecting mild impairment improving routine screening. | 1,565 participants from ADNI database | Alzheimer’s Disease | Developed a Multi-Classifier Network (MCN) by integrating multiple optimized random forest classifiers for three-class classification. | The proposed model achieved high classification level, and it is also able to detect in a short time. | The high detection rate and the minimal assessment time of the four identified features may be a practical starting point for developing screening protocols targeting cognitive impairment defined at CDR 0.5 and above. |
| Noh et al. (67) | 2021 | South Korea | This study seeks to employ machine learning (ML) to pinpoint significant features related to gait and physical fitness, aiming to predict a decline in global cognitive function among older adults. | 306 patients | cognitive function decline in older adults | Eight machine learning models, namely SVM, DT, RF, NN, LASSO, EN, MCP, and SCAD, were employed to identify features with the lowest Root Mean Squared Error (RMSE) for both men and women. | Elastic Net selected five optimal features from the LP data for men, while Support Vector Machine selected twenty optimal features from the XI data for women. This approach successfully identified essential features for predicting a potential decline in global cognitive function in older adults. | The study successfully identified crucial features for predicting potential declines in global cognitive function among older adults. The proposed machine learning approach has the potential to inspire future research focused on early detection and prevention of cognitive function decline in the elderly. |
| Roshanzamir et al. (68) | 2021 | Iran | The study aimed to develop transformer-based deep learning models using natural language processing for the early risk assessment of Alzheimer’s disease by analyzing picture description test data. | 170 AD patients and 99 healthy controls | Alzheimer’s disease | Logistic regression Single hidden layer neural network Single-layer bidirectional LSTM Three-layer CNN | The study evaluated models using picture description test transcripts from the Pitt corpus. That result improves 2.48% over the existing state-of-the-art. | Leveraging pre-trained language models enhances Alzheimer’s Disease (AD) prediction by addressing the challenges of limited datasets and diminishing the reliance on expert-defined features. |
| Sánchez-Reyna et al. (69) | 2021 | Mexico | This research introduces a novel methodology for creating a multivariate model that integrates various features to detect Alzheimer’s Disease (AD). | 106 patients | Alzheimer’s disease | Subsequently, a support vector machine model was created to develop and validate the multivariate classification model. | A five-fold cross-validation showed an AUC of 87.63% for model performance, and in an independent blind test with 20 patients not considered during model construction, the final model achieved a perfect AUC of 100%. | The study introduces a methodology using genetic algorithms to select critical features from Alzheimer’s Disease (AD) data, including gene indexes and clinical assessments. These features are utilized to create supervised classification algorithms with SVM architecture. Model efficiency is assessed through cross-validation and a blind test, emphasizing high sensitivity and specificity for early AD detection among subjects with AD, MCI, or CN. |
| Singhania et al. (70) | 2021 | India, Saudi Arabia | The paper introduces a model utilizing biomarkers, including amyloid-beta protein, for detecting, predicting, and preventing Alzheimer’s Disease onset. | 416 individuals for the cross-sectional MRI scan collection, and 373 individuals for the longitudinal MRI scan collection. | Alzheimer’s Disease | A Convolution Neural Network (CNN) based model is developed to predict AD at its early stages. | The findings demonstrated that the suggested model surpassed traditional Machine Learning (ML) methods, including Logistic Regression, Support Vector Machine, Decision Tree Classifier, and K Nearest Neighbor algorithms. | The model highlights the need for more research on biomarkers to improve algorithm accuracy in forecasting Alzheimer’s Disease progression and combines current patient circumstances to give preventive interventions. |
| Syed et al. (71) | 2021 | Pakistan and Australia | This study aims to propose a multimodal system capable of identifying linguistic and paralinguistic traits associated with dementia, serving as an automated screening tool. | ? | Alzheimer’s dementia | Logistic regression (LRC), support vector machine classifiers (SVC), support vector regression (SVR), and partial least squares regression (PLSR) were used for the classification and regression tasks. | The system was evaluated on the Alzheimer’s Dementia Recognition Challenge dataset, achieving a new state-of-the-art performance in classification and matching the current state-of-the-art in regression. | Tested on the ADReSS challenge dataset, the system outperformed domain-knowledge-based features in audio and text and showed higher performance with deep neural embedding. In the ADReSS challenge, model ensembling was essential in setting a new state-of-the-art for classification and matching the existing state-of-the-art for regression. These results significantly advance the automated detection of Alzheimer’s disease. |
| Tsai et al. (72) | 2021 | Taiwan | The study suggests an intelligent assessment method for evaluating executive functions, employing machine learning to create an automated, evidence-based assessment model. Behavioral data is gathered by engaging participants in executive-function tasks within a virtual reality supermarket. | 6 MCI or early AD participants and 6 control healthy participants | Cognitive impairment and Alzheimer’s disease | Logistic Regression, Support Vector Machines, Decision Tree, Random Forest, AdaBoost (Adaptive Boosting) and XGBoost (eXtreme Gradient Boosting) were applied. | The results indicated that the features derived from the Virtual Reality (VR) system strongly correlated with the diagnosis of individuals with MCI or early AD. | A virtual supermarket assessing executive functions was successfully tested on six healthy and six MCI/early AD participants. Trajectory analysis revealed 45 significantly different indices between groups. Machine learning achieved 100% accuracy in distinguishing healthy from MCI/early AD participants. Study limitations include a small sample size and technical challenges in VR computation. Future large-scale clinical trials are essential for validating the machine-learning model. |
| Uehara et al. (73) | 2021 | Japan | This study aims to conduct transcriptome analyses using SSL-RNAs and evaluate the potential of these expression profiles as diagnostic biomarkers for Parkinson’s disease (PD) through the application of machine learning. | 65 PD patients and 65 control subjects | Parkinson’s disease | Extremely Randomized Trees (ERT) | Differential expression analysis identified over 100 genes differentially expressed between patients with PD and healthy controls in both cohorts, with upregulation of genes related to oxidative phosphorylation. Gene ontology analysis highlighted functional processes associated with PD. | The study explored the potential of utilizing SSL-RNA transcriptome for non-invasive differentiation between patients with Parkinson’s disease (PD) and healthy controls through machine learning. |
| Venugopalan et al. (74) | 2021 | USA | The study aims to employ deep learning (DL) for the integrated analysis of magnetic resonance imaging (MRI), single nucleotide polymorphisms (SNPs), and clinical test data, with the goal of classifying patients into distinct categories, including Alzheimer’s Disease (AD), Mild Cognitive Impairment (MCI), and controls (CN). | 2004 patients | Alzheimer’s disease | Deep Learning Models like SVM, random forests, and decision trees. | Results demonstrate that deep models outperform shallow models, including support vector machines, decision trees, random forests, and k-nearest neighbors. | This study explores the potential of Deep Learning (DL) for improving the accuracy of Alzheimer’s Disease (AD) diagnosis and staging assessment using multi-modal data fusion. Key findings include the superiority of DL over shallow models for single-modality AD stage prediction, the effectiveness of a novel DL framework for multi-modality data fusion, and the application of perturbation and clustering-based feature extraction for interpretable DL model insights in AD stage prediction. |
| Wang et al. (75) | 2021 | USA and China | The study aims to develop and validate a deep learning model for detecting evidence of cognitive decline from clinical notes in the Electronic Health Record (EHR). | 3,130 patients from two databases. | Cognitive decline | The study implemented a hierarchical attention-based deep learning structure along with four baseline machine learning algorithms: logistic regression, random forest, support vector machine, and XGBoost. | The deep learning model outperformed baseline models in both datasets. | A deep learning model proved to be accurate in this diagnostic study in identifying cognitive decline from clinical notes before diagnosing mild cognitive impairment (MCI). It performed better than other machine learning models and keyword-based searches, suggesting that it may be able to identify early cognitive decline in electronic health records (EHRs). |
| Zhu et al. (76) | 2021 | USA | This study aims to propose a flexible spatial–temporal solution that can predict the risk of Mild Cognitive Impairment (MCI) conversion to Alzheimer’s Disease (AD) before the onset of clinical symptoms. This is achieved by sequentially recognizing abnormal structural changes from longitudinal magnetic resonance (MR) image sequences. | 151 subjects | Mild Cognitive Impairment (MCI) and Alzheimer’s Disease (AD) | Temporally Structured Support Vector Machine (TS-SVM) model | The early diagnosis method, utilizing only two follow-up MR scans, predicts conversion to Alzheimer’s Disease 12 months ahead of clinical diagnosis with an accuracy of 81.75%. | The paper introduces a novel method for predicting the conversion from Mild Cognitive Impairment (MCI) to Alzheimer’s Disease (AD) using only 2 MR images. The approach employs a Temporally Structural-SVM (TS-SVM) and joint feature selection framework. The model extracts partial MR image sequences at different time points, enforces monotony on SVM outputs, and achieves promising accuracy in classifying MCI converters and non-converters with fewer MR images compared to standard SVM approaches. |
| Mehmood et al. (77) | 2021 | China | The study aims to diagnose Alzheimer’s disease (AD) in its early stages by focusing on the problem with layer-wise transfer learning and brain imaging tissue segmentation. | 85 NC patients, 70 EMCI, 70 LMCI, and 75 AD patients | Alzheimer’s Disease | Convolutional neural networks (CNNs) | The proposed method could distinguish between AD and NC. | Finally, comparing this model with comparable researches showed that it outperformed the most recent models in the field in terms of testing accuracy. |
| Montolío et al. (78) | 2021 | Spain | This study used clinical data and measures of retinal nerve fiber layer (RNFL) thickness using optical coherence tomography (OCT) to improve the diagnosis of multiple sclerosis (MS) and predict the progression of long-term impairment in MS patients. | 104 healthy controls and 108 MS patients | Multiple Sclerosis | Multiple linear regression Support vector machine Decision tree K-nearest neighbors Naïve Bayes Ensemble classifier Long short-term memory | The ensemble classifier performed best in diagnosing MS, while LSTM performed best in the long-term prediction of MS disability course. | This study shows that it is possible to get an early MS diagnosis and predict the course of the disease by utilizing machine learning techniques with both clinical and OCT data. |
| Sudharsan et al. (79) | 2021 | India | This study aims to compare and recommend efficient classification methods for a specific and thoroughly analyzed dataset, with classifiers including SVM, significance vector machine, and RELM. | 214 individuals | Alzheimer’s disease | Import Vector Machine (IVM) Regularized Extreme Learning Machine (RELM) Support vector machine (SVM) | RELM had the best accuracy among classifiers. | With limited data, the study addresses the diagnosis issues associated with Alzheimer’s/MCI, highlighting the efficacy of RELM and investigating ways to improve accuracy using item-measure methodologies. |
| Wang et al. (80) | 2021 | China | Through machine learning techniques and electronic health records, this study aimed to create a helpful tool to identify MS patients early. | Training set with 239 MS and 1,142 controls, and the test set with 23 MS and 92 controls. | Multiple Sclerosis | Extreme Gradient Boosting (XGBoost) Random Forest (RF) Naive Bayes K-nearest-neighbor (KNN) Support Vector Machine (SVM) | XGBoost had the best performance. | Reducing diagnostic delays in MS can be achieved by developing a diagnostic tool for early MS detection based on the XGBoost model and electronic health records. |
| Zeng et al. (81) | 2021 | China | This study propose a data-driven approach to identify potential biomarkers for Alzheimer’s disease (AD) and other poorly understood brain diseases. | 458 subjects for training set and the 278 subjects for validation set. | Alzheimer’s Disease | Convolutional Neural Network (CNN) Ensemble Learning (EL) Genetic Algorithm (GA) | 6 genes were identified in association with AD. | This method adaptively achieves more reliable and efficient candidate biomarkers in a data-driven manner, overcoming the limits related to the impact of subjective factors and dependence on prior knowledge. |
| Peng et al. (82) | 2021 | China | This research aimed to develop radiomics models using different machine learning methods to forecast the evolution of unenhanced Multiple Sclerosis (MS) lesions and identify the best model. | 36 patients with MS | Multiple Sclerosis | three machine learning classifiers, including logistic regression (LR), random forest (RF), and support vector machine (SVM) | The best prediction performance was for the SVM classifier with ReliefF | The outcomes showed that the machine learning model based on radiomics could forecast how MS lesions will change over time. |
| Lee et al. (83) | 2021 | Republic of Korea | An novel strategy for identifying people with early-stage mild cognitive impairment (eMCI) is presented in this study. By concurrently learning functional connections from automatically selected areas of interest (ROIs) for each subject, the technique accounts for individual variability. | 53 eMCI and 48 cognitively normal | MCI | Deep Neural Network(composed of a temporal embedding module, an ROI selection module, and a disease-identification module) Support vector machine (SVM) with FC computed using Pearson correlation. Self-attention conventional CNN LSTM-DG | The proposed model had efficacy In identifying eMCI. | This paper presented a novel framework for the identification of personalized early-stage mild cognitive impairment (eMCI), employing a Graph Convolutional Network (GCN) for relational representation learning and reinforcement learning for automatic ROI selection. The examination validated the efficacy of the approach in identifying pertinent areas documented in neuroscientific investigations on Alzheimer’s disease (AD) and moderate cognitive impairment (MCI). |
| Buegler et al. (84) | 2020 | EU/USA | Using just NMI (Neuro-Muscular Index) digital biomarkers, the comprehensive external validation study aimed to develop prediction models that were both widely applicable and dependable. | 215 subjects for first dataset and 496 subjects for second dataset. | Dementia | XGBoost classification algorithm as the prediction model, using binary logistic regression | The proposed model was able to distinguish healthy subjects from subjects at risk to dementia within 3 years. | Digital biomarker prognostic models serve as beneficial resources for enormous scale population screening because they allow cognitive impairment to be detected early and allow for ongoing patient monitoring. |
| Alkhatib et al. (85) | 2020 | Lebanon | This study provides a detection algorithm that uses the load distribution during gait to categorize participants as either normal or Parkinson’s patients. | 18 normal subjects and 29 PD subjects | Parkinson’s disease | Center of Pressure Path and Load Distribution | Using a linear decision boundary caused achievement of high classification accuracy. | This letter serves as a foundation for developing a portable device for real-time early detection of Parkinson’s disease. Additionally, it can be utilized for assessing the effectiveness of a rehabilitation program. |
| Gill et al. (86) | 2020 | Canada | The purpose of this study is to determine whether baseline mild behavioral impairment (MBI) status, which is utilized for NPS quantification, and brain morphological features are predictive of a follow-up diagnosis in individuals with normal cognition (NC) or MCI, median 40 months later. | 102 individuals with NC and 239 with MCI | Dementia | A logistic model tree classifier combined with information gain feature selection was trained to predict follow-up diagnosis. | The optimal model required MBI total score and left hippocampal volume to classify participant as NC or impaired cognitive. | This study integrates clinical, neuropsychiatric, and MRI data using machine learning to predict future cognitive categories in non-demented older adults, showing improved accuracy with the inclusion of well-described neuropsychiatric symptoms. |
| Cheng et al. (87) | 2020 | China | The aim of this study is to overcome data density issues in understanding the Huntington’s disease (HD) mechanism by strategically reducing dimension size and employing machine learning to identify enriched pathways associated with HD using existing data. | 157 HD and 157 controls | Huntington’s disease | Decision tree Rule induction Random forest Generalized linear model | 66 potential HD-contributing genes were identified by proposed machine learning algorithms. | To further the pathophysiology of HD, the mutant HTT may obstruct the expression and trafficking of several identifiable genes. |
| Pan et al. (88) | 2020 | France | Determining which MCI individuals are at risk of developing AD-type dementia is the primary goal of this study. | 1,005 subjects | Alzheimer’s disease | Multi-view Separable Pyramid Network (MiSePyNet) model was designed and also CNN algorithm was used. | Comparable to other state-of-the-art algorithms, the suggested technique can distinguish AD from Normal Control (NC). In terms of forecasting the course of mild cognitive impairment, the approach can outperform both conventional and deep learning-based algorithms. | The paper introduces MiSePyNet, a novel CNN model designed for AD prediction in the MCI stage and classification among NC subjects using 18F-FDG PET modality. MiSePyNet employs factorized c |