End-to-end light-weighted deep-learning model for abnormality classification in kidney CT images

Kidney disease is a major health problem that affects millions of people around the world. Human kidney problems can be diagnosed with the help of computed tomography (CT), which creates cross-sectional slices of the organ. A deep end-to-end convolutional neural network (CNN) model is proposed to help radiologists detect and characterize kidney problems in CT scans of patients. This has the potential to improve diagnostic accuracy and efficiency, which in turn benefits patient care. Our strategy involves teaching a suggested deep end-to-end CNN to distinguish between healthy and diseased kidneys. The recommended CNN is trained using a standard CT image library that has been annotated to show kidney stones, cysts, and tumors. The model can then be used to detect kidney abnormalities in fresh CT scans, which may enhance the effectiveness and speed with which diagnoses are made. A total of 1812 pictures were used, each one a unique cross-sectional CT scan of the patient. Our model has a detection rate of 99.17% in CT scan validation tests. We employed a different dataset with a total of 5077 normal samples, 3709 cyst samples, 1377 stone samples, and 2283 tumor samples. In tests, our model proved to be 99.68% accurate. The suggested framework has been validated by applying it to the clinical dataset, resulting in 99% accuracy in predictions. As low-cost and portable CT scanners become more commonplace, the described concept may soon be employed outside of a hospital environment, at the point of treatment, or even in the patient's own home.