Advancements in Chronic Kidney Disease Diagnosis: Leveraging EfficientNetB3 for Accurate CT Image Classification

This study proposes a novel technique for automatic classification of kidney ailments from CT images using EfficientNetB3 transfer learning model. The EfficientNetB3 architecture has been shown to outperform other designs on a broad range of image classification tasks in recent papers which has motivated the present investigation for classification of renal diseases. The proposed method aims to classify renal ailment images into four classes which can provide valuable insight for medical diagnosis and planning of treatment. The EfficientNetB3 model is trained and tested on a dataset consisting of 12,446 CT scans and has indicated remarkable performance achieving an accuracy of 99%. The training took 20 epochs and 30 samples were processed per batch per epoch. The default learning rate was used for the training. The high accuracy which the proposed approach has achieved suggests that EfficientNetB3 has potential for accurate classification of kidney disorders and therefore, it promises to be useful as a tool for healthcare practitioners to efficiently identify and address renal ailments. The results of this study contribute to the growing body of research in medical image processing and are expected to play a pivotal role in laying the foundation for development of more sophisticated diagnostic tools in nephrology.