A Machine Learning Approach to Cancer Detection and Localization Using Super Resolution Ultrasound Imaging

In this work, we present preliminary results on cancer detection and localization using super-resolution ultrasound imaging (SRUI) data. SRUI enables the visualization of the organ's physiology at micro-vascular level, which is of course related to the development of several major diseases, e.g., cancer. Our analysis serves as a feasibility study in our attempt to identify cancer regions using our SRUI algorithm on real prostate data. Despite the limited number of patients, we demonstrate that discrimination between healthy and tumorous regions using a popular anomaly detection technique, i.e., One-class Support Vector Machine, on SRUI data is indeed feasible. The problem is modeled as a binary classification task and relevant evaluation metrics are used for the evaluation of the method's performance.

Automated summary

In this study, we present preliminary results on cancer detection and localization using super-resolution ultrasound imaging data. Our analysis demonstrates the feasibility of discriminating between healthy and tumorous regions using SRUI data, despite the limited number of patients.