LMFLOSS with MODEL SOUPS:A Novel Approach for Improving Imbalanced Medical Image Classification

Research Supervisor
Journal Title
Journal Title
Journal ISSN
Volume Title
Automatic medical image classification is a very important field where the use of AI has the potential to have a real social impact. However, there are still many challenges that act as obstacles to making practically effective solutions. One of those is the fact that most of the medical imaging datasets have a class imbalance problem. This leads to the fact that existing AI techniques, particularly neural network-based deep-learning methodologies, often perform poorly in such scenarios. Thus this makes this area an interesting and active research focus for researchers. In this study, we propose a novel loss function to train neural network models to mitigate this critical issue in this important field. Through rigorous experiments on three independently collected datasets of three different medical imaging domains, we empirically show that our proposed loss function consistently performs well with an improvement between 2%-10% macro f1 when compared to the baseline models. In order to further improve the performance of the imbalanced datasets, we incorporate a state of art method called ‘Model Soups’ into our study. For two of the three datasets, model soups showcase further improvement of macro f1 score of up to 2% when compared to the best-performing individual models. We hope that our work will precipitate new research toward a more generalized approach to medical image classification.
Department Name
Electrical and Computer Engineering
North South University
Printed Thesis