You are prohibited from using or uploading content you accessed through this website into external applications, bots, software, or websites, including those using artificial intelligence technologies and infrastructure, including deep learning, machine learning and large language models and generative AI.


To evaluate the usefulness of a computer-aided diagnosis system (CAD) in evaluating the severity of benign prostatic hyperplasia (BPH) using a super-ellipse model to characterize changes in prostate contours.


We prospectively recruited 60 patients who were scanned with T2-weighted MRI (T2WI). A super-ellipse model was used to obtain structural features of their prostates. The super-ellipse shape can be characterized by parameter vector p= (lx, ly, r, sy, sq, xy, t, b) (Figure), and expressed by (x/ax)2/ɛ+(y/ay)2/ɛ=1. We used a learning algorithm in a support vector machine to learn the features of the prostate, and the BPH severity data, which included International Prostate Symptom Score (IPSS), IPSS quality of life (QOL), Over-Active Bladder Symptom Score (OABSS), maximum flow rate (ml/s), and residual urine volume (ml). We then analyzed the predictive effect of the learning model, from the structure of the prostate to BPH severity.


To train the learning model, we used data from the T2WI and features of 20 patients, included 10 with severe BPH and 10 with normal prostates. The BPH severity of the other 40 patients were predicted from their T2WI, using the learning model; results were accuracy: 87.5%, precision: 84.6%, recall: 92.0% and F1 score: 88.0%. The area under curve for predicting BPH severity with the learned model was 0.98 (P<0.0001). However, classifying BPH severity was somewhat difficult due to similar features of prostates with different sizes.


The present learning model can potentially predict BPH severity using the structural data from patients’ T2WI. However, scale information would be useful to predict with more high accuracy in addition to the learned model.

Source of Funding: