Classification of Autistic Patients and Control Via Utilizing Dictionary of Functional Modes as Brain Atlas

سال انتشار: 1400
نوع سند: مقاله کنفرانسی
زبان: انگلیسی
مشاهده: 184

نسخه کامل این مقاله ارائه نشده است و در دسترس نمی باشد

این مقاله در بخشهای موضوعی زیر دسته بندی شده است:

استخراج به نرم افزارهای پژوهشی:

لینک ثابت به این مقاله:

شناسه ملی سند علمی:

IBIS10_003

تاریخ نمایه سازی: 5 تیر 1401

چکیده مقاله:

Autism spectrum disorder is a set of neurodevelopmental disorders difficult to diagnose especially inchildhood when the symptoms emerge. The early diagnosis of the disease can prevent many complicationsin patients in future years. Lacking suitable and applicable biomarkers and indices for the diagnosis, on theone hand, the similarity of these disorders to other neuropsychiatric diseases in another hand, made the earlydiagnosis almost overbearing. Functional Magnetic Resonance Imaging is one of the methods to diagnoseneurological disorders by detecting changes associated with blood flow in the brain. Images from this methodare high dimensional data which makes the training process of machine learning models challenging. One ofthe solutions is to extract essential features from images to ease the process. To this end, features from theseimages can be extracted using brain atlases for regions of interest. We used Dictionary of Functional Modesas an atlas for feature extraction followed by training a logistic regression model to classify images of autisticindividuals and control cases. We leveraged datasets from the Autism Brain Imaging Data Exchange databasecontaining images from autism and control individuals. The dataset has many variants due to variations indata collection from different imaging centers. Since these take-ups could obscure the accuracy of ourmachine learning model, we implicated the Combat method to remove these unwanted side-effects.Exploiting these methods and the atlas for feature extraction resulted in a significant increase in accuracy ofour logistic regression classifier (۷۱%) which is more optimized than previous methods such as asd-diagnetneural network. Our method can be applicable to classify other neurological disorders with regard to the brainregion of interest. This is heartwarming for the future of precision medicine since it has the ability toinvestigate potential biomarkers of such complex disorders.

نویسندگان

Karim Rahimian

Biophysics Department, Faculty of Biological Sciences, Tarbiat Modares University, Tehran, Iran

Mohsen Khodarahmi

Department of Radiology, Shahid Madani Hospital, Karaj, Iran

Parimah Emadi Safavi

Biophysics Department, Faculty of Biological Sciences, Tarbiat Modares University, Tehran, Iran

Seyed Shahriar Arab

Biophysics Department, Faculty of Biological Sciences, Tarbiat Modares University, Tehran, Iran

Javad Zahiri

Department of Neuroscience, University of California, San Diego, La Jolla, CA, United States