Model strojnog učenja za klasifikaciju poremećaja iz spektra autizma

undergraduate thesis

• Title (english): Machine learning model for the autism spectrum disorder classification
• Author: Lucija Petkoviček
• Mentor: Jelena Božek (mentor)
• Committee member: Jelena Božek (predsjednik povjerenstva)
• Committee member: Sonja Grgić (član povjerenstva)
• Committee member: Ivana Žeger (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-07-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Poremećaji iz spektra autizma (ASD) smatraju se kompleksnijim izazovom današnje
medicine, a primjenom strojnog učenja pokušavaju se otkriti najbolje metode dijagnostike,
najčešće korištenjem funkcijskih MRI podataka u stanju mirovanja. ABIDE je inicijativa
koja osigurava skup podataka i metode pretprocesiranja. U ovom radu se koristi
Preprocessed Connectomes Project CPAC (Configurable Pipeline for the Analysis of
Connectomes) metoda pretprocesiranja, ekstrakcija matrice povezanosti pomoću dva atlasa,
AAL (Automated Anatomical Labeling) i MIST (Multiresolution Intrinsic Segmentation
Template) s više podjela regija, selekcija značajki RFE (Recursive Feature Elimination)
metodom i SHAP vrijednostima te klasifikacija metodom potpornih vektora. Uspoređene su
točnosti klasifikatora sa i bez selekcije značajki, ovisno o odabiru atlasa. Također su
uspoređene i same značajke rangirane kao najbolje za korišteni model klasifikacije ASD,
ovisno o metodi selekcije značajki i odabiru atlasa.

Abstract (english)

Autism spectrum disorders (ASD) are considered one of the most complex challenges in
modern medicine. By applying machine learning, efforts are being made to discover the best
diagnostic methods, often using resting-state functional MRI data. ABIDE is an initiative
that provides datasets and preprocessing methods. In this study, Preprocessed Connectomes
Project CPAC (Configurable Pipeline for the Analysis of Connectomes) preprocessing
method is used, connectivity matrix extraction using two atlases, AAL (Automated
Anatomical Labeling) and MIST (Multiresolution Intrinsic Segmentation Template) with
multiple region parcelations, feature selection using the RFE (Recursive Feature
Elimination) method and SHAP values, and classification using the support vector machine
method. The accuracies of the classifiers with and without feature selection are compared,
depending on the atlas choice. Additionally, the features ranked as the best for the ASD
classification model are compared based on the feature selection method and atlas choice.

• Keywords: autizam; rs-fMRI; ABIDE; strojno učenje; metoda potpornih vektora; selekcija značajki; RFE; SHAP
• Keywords (english): autism; rs-fMRI; ABIDE; machine learning; support vector machine method; feature selection; RFE; SHAP
• Language: croatian
• URN:NBN: urn:nbn:hr:168:483622
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 21:50:54