Title:
Multi-Stage Transfer Learning with Modified Classification Loss Function for Small Functional Neuroimaging Dataset
Author:
Lee Lok Hua
Supervisor:
Department:
Electrical & Electronics Engineering
Year:
2026
URL:
Abstract:
Small datasets is a prominent challenge in training a Convolutional Neural Network (CNN) to perform well in neuroimaging classification task. Transfer learning appears to be a promising approach, but ...
View more
Multimedia Content:
01
Methodology
1. Functional Near-Infrared Spectroscopy (fNIRS) Data Acquisition
The study employed functional Near-Infrared Spectroscopy (fNIRS) data collected from Major Depressive Disorder (MDD) patients to in...
View more
02
Citation Analysis
1. NeuroImage
Scope: Neuroimaging, brain mapping, cognitive neuroscience, neuroinformatics, and imaging analysis.
Relevance: Directly related to functional neuroimaging, fNIRS analysis, and AI-ass...
View more
03
Impact
1. Technical Impact
This research introduces a modified Multi-Stage Transfer Learning (MSTL) framework specifically designed for small functional neuroimaging datasets, addressing one of the major ...
View more
04
Benefits to Society
1. Environmental Benefits
Although this research is not directly related to environmental sustainability, the adoption of AI-assisted clinical decision support systems can reduce unnecessary diagno...
View more
05
Applications
1. Antidepressant Response Prediction
Benefit: Improves early identification of treatment responders and non-responders.
Implementation: The Deep Learning framework can analyse fNIRS brain activat...
View more
06
Infographic
Summary of this Thesis
07
Patents / Products
Product Name:
Modified MSTL Framework with SPD-Regularized Loss for Small Functional Neuroimaging Datasets
Significance:
Provides a specialized transfer-learning framework for neuroimaging tasks ...
View more
08
Experts
1. Prof. Dr Tang Tong Boon
Institution: Universiti Teknologi PETRONAS (UTP), Malaysia
Contribution: Main supervisor and principal researcher guiding AI-based neuroimaging prediction framework deve...
View more