Iterative Misclassification Error Training (IMET): An Optimized Neural Network Training Technique for Image Classification

Deep learning models have shown strong performance in medical image diagnostics. The scarcity of data for certain medical conditions, coupled with the presence of noisy, mislabeled, or non-generalizable images, poses a significant challenge to model performance. Several data-efficient training strategies have been proposed to address these constraints. However, developing a generalizable difficulty ranking mechanism that works across diverse domains, datasets, and models while reducing the computational tasks still remains challenging. In this research, we propose Iterative Misclassification Error Training (IMET), a novel training technique to optimize and improve the performance of deep learning models. The IMET approach is aimed to identify misclassified samples in order to streamline the training process, while prioritizing the model’s attention to edge case scenarios and rare outcomes. The paper evaluates IMET’s performance on benchmark medical image classification datasets against standard benchmark ResNet architectures. The IMET technique achieved accuracies of 80.3% and 90.2% on the OCTMNIST and PneumoniaMNIST datasets, respectively, in comparison to 77.6% and 88.6% obtained by the benchmark models. Additionally, the IMET technique outperformed the benchmark models with both a significantly lower parameter count as well as a lower number of training samples. These results demonstrate IMET’s potential for enhancing model accuracy and performance in medical image analysis.