Pelvic Uretero Junction Obstruction Detection Using Machine Learning Algorithm

Uretero-Pelvic Junction (UPJ) obstruction, if undetected, can lead to severe kidney dysfunction, underscoring the need for early and reliable diagnostic solutions. This research aims to develop an effective, non-invasive diagnostic tool by integrating real-time ultrasound imaging with advanced machine learning techniques. Pre-processed ultrasound images are analyzed using MobileNetV2 for feature extraction, followed by classification through a Support Vector Machine (SVM). SVM was selected for its robustness with limited training data, its ability to operate effectively in high-dimensional spaces, and its interpretability—an essential factor in medical diagnostics. By combining the strengths of MobileNetV2 and SVM, the model ensures a balance between computational efficiency and diagnostic accuracy, addressing challenges like limited datasets and the complexity of medical imaging. The purpose is to enable precise detection of UPJ obstruction while minimizing the limitations of traditional diagnostic methods. The proposed model demonstrates significant accuracy, supported by evaluation techniques such as confusion matrices and t-SNE. This approach underscores the transformative potential of combining machine learning with medical imaging, offering a pathway to enhanced patient outcomes through timely and reliable intervention.