Handling Data Heterogeneity in Medical Image Classification: Enhancing Accuracy Through Advanced Machine Learning Techniques - دانشکده فنی و مهندسی
Handling Data Heterogeneity in Medical Image Classification: Enhancing Accuracy Through Advanced Machine Learning Techniques
نوع: Type: Thesis
مقطع: Segment: masters
عنوان: Title: Handling Data Heterogeneity in Medical Image Classification: Enhancing Accuracy Through Advanced Machine Learning Techniques
ارائه دهنده: Provider: Alireza Maleki
اساتید راهنما: Supervisors: Prof. Hassan Khotanlou
اساتید مشاور: Advisory Professors:
اساتید ممتحن یا داور: Examining professors or referees: Dr. Muharram Mansoorizade - Dr.Reza Mohammadi
زمان و تاریخ ارائه: Time and date of presentation: 2025
مکان ارائه: Place of presentation: آمفی تئاتر
چکیده: Abstract: In the field of medical imaging, data heterogeneity is one of the main challenges in developing machine learning models for accurate disease diagnosis. Differences in imaging methods, protocols, and patient demographics make data from different institutions inconsistent; this reduces the generalizability of training models and challenges their performance in the face of new data. Furthermore, reduced diagnostic accuracy can lead to inequities in health care delivery. On the other hand, the technical and computational complexity of managing multidimensional images and privacy concerns create further problems. According to the studies conducted, in order to deal with these challenges, various methods have been investigated, which can be referred to as the categories of federated learning (horizontal and vertical), aggregation-based methods (such as FedAvg, SplitAVG and FedSGD), transfer-based methods (such as cyclic weight transfer and SplitNN), personalized federated methods, self-supervised learning, vision transformer-based approaches and adversarial learning. Each of these categories has its own advantages and disadvantages; for example, horizontal federated learning has the advantage of being easy to share model updates, while vertical federated learning has the ability to provide more comprehensive models by combining multi-modal data, but it has privacy concerns. This study analyzes the advantages and disadvantages of each method and provides guidance for choosing the appropriate method in developing accurate diagnostic models in heterogeneous medical imaging environments.