Predikcija dobi iz slika magnetske rezonancije mozga temeljena na prijenosnom učenju

master's thesis

• Title (english): Age prediction from brain magnetic resonance images based on transfer learning
• Author: Andrija Banić
• Mentor: Jelena Božek (mentor)
• Committee member: Jelena Božek (predsjednik povjerenstva)
• Committee member: Sonja Grgić (član povjerenstva)
• Committee member: Dario Bojanjac (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-09-13, Croatia

Abstract

Predviđanje dobi mozga putem magnetske rezonancije postalo je ključna metoda za procjenu neurološkog zdravlja. U ovom radu korišteno je prijenosno učenje s pretreniranim konvolucijskim neuronskim mrežama, prvotno treniranim na LifespanCN bazi podataka, za predviđanje dobi mozga. Model je testiran na IXI i ADNI bazama podataka, uspješno razlikujući kontrolne skupine od osoba s Alzheimerovom bolešću i blagim kognitivnim oštećenjem. Iako su rezultati pokazali potencijal modela kao biomarkera u kliničkoj dijagnostici, problemi su uočeni kod mlađih ispitanika, posebno na ABIDE datasetu za istraživanje autizma. Ova tehnologija ima obećavajuće primjene u kliničkoj dijagnostici i istraživanjima, no daljnja poboljšanja su nužna za točnije
predviđanje u različitim dobno-specifičnim skupinama. Bilo kako bilo, integracija umjetne inteligencije u medicinska istraživanja otvara nove perspektive za dijagnostiku i liječenje složenih bolesti. Ovo dinamično rastuće područje sigurno će donijeti značajan napredak za čovječanstvo, ali za puni potencijal potrebno je nastaviti s dubljim istraživanjem i eksperimentiranjem u možda još neotkrivenim pravcima.

Abstract (english)

Prediction of brain age using magnetic resonance imaging has become a crucial method for assessing neurological health and age-related brain changes. This study employed transfer learning with pretrained convolutional neural networks, initially trained on the LifespanCN dataset, for brain age prediction. The model’s performance was evaluated on the IXI and ADNI datasets, which include healthy controls and individuals with Alzheimer’s disease or mild cognitive impairment. The model successfully differentiated control groups from those with neurological conditions, demonstrating its potential as a biomarker in clinical diagnostics. However, challenges were noted in accurately predicting brain age for younger individuals, as seen in suboptimal results on the ABIDE dataset for autism research. Nevertheless, the integration of artificial intelligence into medical research opens new perspectives for diagnosing and treating complex diseases. This dynamically growing field is sure to bring significant advancements for humanity, but realizing its full potential requires further in-depth research and experimentation, possibly in yet undiscovered directions.

• Keywords: predviđanje dobi; magnetska rezonancija; prijenosno učenje; strojno učenje; duboko učenje; medicinsko oslikavanje
• Keywords (english): age prediction; brain MRI; transfer learning; machine learning; deep learning; neuroimaging
• Language: croatian
• URN:NBN: urn:nbn:hr:168:150390
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 20:01:50