Enhanced Malignant Lymphoma Classification Using an Explainable Dilated MobileNetV2 and Convolutional-Recurrent Neural Network

Lymphoma, a common malignancy in the human lymphatic system, necessitates accurate and early detection to improve patient outcomes. This research explores a novel approach to classifying malignant lymphoma using an interpretable architecture combining Dilated MobileNetV2 and a Convolutional-Recurrent Neural Network (CRNN). The proposed model integrates the efficiency of dilated convolutions for feature extraction with the temporal analysis capability of recurrent networks. By employing interpretable techniques, the model provides insights into the decision-making process, enhancing its clinical applicability. Experimental evaluations were performed on publicly available lymphoma datasets, leveraging image pre-processing, feature normalization, and class balancing. The results indicate that the proposed model outperforms conventional architectures, achieving high classification accuracy while maintaining interpretability. This research facilitates the implementation of deep learning solutions in practical diagnostic environments, benefiting both pathologists and patients.