Kidney Stone Detection using Deep Learning Model
Neha Panchal, Meenaxi M Raikar, Vishwanath P. Baligar
Abstract
The identification of kidney stones, a common medical condition, has prominently improved with the application of deep learning and medical imaging technologies. The proposed system utilizes the You Only Look Once (YOLOv8) algorithm to increase the accuracy and effectiveness of kidney stone detection in images obtained from Computed Tomography (CT) scans. Traditional investigative methods, although effective, can be time-intensive and susceptible to human error. By leveraging the YOLOv8 model, healthcare professionals can analyze images more swiftly and accurately, identiflying the presence and location of kidney stones with minimal oversight. The proposed system is trained on a dataset of 3405 images, incorporating data augmentation techniques such as Blur, MedianBlur, ToGray, and CLAHE to simulate real-world scenarios. The YOLOv8 architecture, known for its speed and precision, is optimized with a training process that achieved a trained mAP50 accuracy of 76.1% and a precision of 82.3%. These results emphasize the potential of advanced AI algorithms to transform medical diagnostics, enabling more accurate, efficient, and automated detection systems. This study highlights the capability of YOLOv8 to reduce diagnostic burdens on radiologists and improve patient outcomes through enhanced detection accuracy and efficiency.