Domain Specific, Semi-Supervised Transfer Learning for Medical Imaging

TL;DR

This paper introduces MAKNet, a lightweight semi-supervised transfer learning model with mixed asymmetric kernels, designed for medical imaging, achieving high accuracy with fewer parameters and emphasizing domain-specific knowledge.

Contribution

The paper proposes MAKNet, a novel lightweight architecture with mixed asymmetric kernels, combined with semi-supervised learning for improved transfer learning in medical imaging.

Findings

MAKNet reduces parameters by 60-70% compared to popular architectures.

Semi-supervised training with pseudo-labels enhances transfer learning.

Domain-specific knowledge improves classification performance.

Abstract

Limited availability of annotated medical imaging data poses a challenge for deep learning algorithms. Although transfer learning minimizes this hurdle in general, knowledge transfer across disparate domains is shown to be less effective. On the other hand, smaller architectures were found to be more compelling in learning better features. Consequently, we propose a lightweight architecture that uses mixed asymmetric kernels (MAKNet) to reduce the number of parameters significantly. Additionally, we train the proposed architecture using semi-supervised learning to provide pseudo-labels for a large medical dataset to assist with transfer learning. The proposed MAKNet provides better classification performance with $60 - 70\%$ less parameters than popular architectures. Experimental results also highlight the importance of domain-specific knowledge for effective transfer learning.