A Scoping Review of Clinical Diagnosis, Classification and Treatment of Patients in Huntington’s Disease: An Artificial Intelligence Approach

Background and aims: Huntington’s disease (HD) is a hereditary disorder arising from neurodegenerationindividualized by chorea, dystonia, impaired gait and dementia. Degeneration occurswhen there is mutation of HTT gene which encodes the protein is called Huntingtin. Also, there isconsiderable domain of genes associated with Huntington’s disease in addition to the mentionedone. Nowadays, numerous methods are utilized to detect these genes, prominent examples areArtificial Intelligence methods. We aim to clarify AI-based impact on delineation of Huntington’sdisease.Method: Four databases including PubMed, Scopus, WOS and Embase were searched accordingto terms related to diverse algorithms of Artificial Intelligence. All aspects of Machine Learningin detecting, classification and analyzing of genetic factors and also, verification of previousstudies were considered in search. ۱۶۵ results for PubMed, ۱۵۹ for WOS, ۳۴۷ for Scopus, ۴۹۰for Embase and then ۱۱۶۱ from all databases were found. After duplication, ۷۰۷ articles left forscreening. Three independent reviewers screened based on title and abstract to elucidate if the fulltext is related or not. Studies, in which the interaction of three features above completely werepresented, included and then review, animal and nonoriginal studies were excluded. Finally, wedid hand-searching to recognize missing studies in database searches.Results: The paper search obtained ۱۱۶۱ studies. After screening titles and abstracts, full-textscreening where done for ۸۷ studies, then ۵۲ studies yielded based on the eligibility criteria. ۲۲studies exclusively utilized single AI technique to detect Huntington’s disease and it’s relatedaspects, ۱۷ used for classification, ۹ for predict and two used for treatment. Plenty of MachineLearning algorithms were used, artificial neural networks for ۶ and support vector machine for۸ studies, individually were used. Also, researchers used multiple Machine Learning methods in۳ studies. Twenty of ۵۲ included papers conforming to Machine Learning were investigating geneticfactors on Huntington’s disease. In two of ۱۲, studies were directly following the extractionof genes associated with various traits of disease. In the first one, BioDCV system using supportvector machine identifies top-ranking genes, finally two ARFGEF۲ and GOLGA۸G genes werechosen as an up-regulated from ۲۰ recognized genes. On the other hand, ۴ algorithms utilizedincluding decision tree (accuracy=۹۰.۷۹%) noticed EPHX۱, ALDH۱A۱, and GLI۱ (EPHX۱ asthe most efficient), Rule induction (accuracy=۸۹.۴۹%) identified EPHX۱, OTP, and ITPKB (OTPas the most efficient), Random forest (accuracy=۹۰.۴۵%) identified ۴۹ genes (KLHDC۵ as themost efficient) and Generalized linear model )accuracy=۹۷.۴۶%) identified ۵۳ genes(OTP as themost efficient one).Conclusion: This study provides generalized evaluation of AI function on detection, classification,treatment and above all, interaction of Al and diagnostic or therapeutic systems associatedwith Huntington’s disease. Due to the hereditary origin and declining executive function in Huntington,improving genetic tests and accelerating the therapy is fundamental in patients, thus AIcan lead to discriminative performance to achieve these goals.