iProstate, a mathematical predictive model-based, 3D-rendering tool to visualize the location and extent of prostate cancer

✍ Scribed by Juan Rodriguez; Jose-Maria Sabater; Jose-Luis Ruiz; Ana Soto; Nicolas Garcia

• Publisher: Wiley (Robotic Publications)
• Year: 2011
• Tongue: English
• Weight: 993 KB
• Volume: 7
• Category: Article
• ISSN: 1478-5951
• DOI: 10.1002/rcs.372

✦ Synopsis

Abstract

Background

Prostate cancer is the most common non‐cutaneous cancer in men. Radical prostatectomy (RP) is a mainstay of treatment for organ‐confined prostate cancer. Preoperatively, the process of planning RP is based on local stage and precisely defines the location and size of disease within the prostate. Nowadays, there is no technology that can accurately map prostate cancer within the gland. Hence, urologists rely mainly on information provided by histopathological examination of biopsy cores, but this information does not accurately locate or stage prostate cancer. The purpose of this study was to provide the surgeon with a 3D visualization tool capable of showing the location and extent of tumour within the prostate.

Methods

To perform this task, an application named iProstate, which makes use of four different mathematical predictive models that use biopsy cores information, was developed. These predictive models were trained with 277 clinical reports from patients who had undergone radical prostatectomy. Two sets of data from the patient reports were used to train the predictive models, the first containing the lengths of the biopsy cores and the tumour percentages of the biopsy cores information, and the second containing the lengths of the biopsy cores, tumour percentages of the biopsy cores, Gleason score, PSA and gland volume information.

Results

Multilayer Perceptron was the predictive model that scored the better results, giving a better approximation in the prediction of location and extent of tumour in the prostatic gland. Gleason score, PSA and gland volume proved to be important variables for the training of the predictive model.

Conclusions

The application was able to perform predictions of the location and extent of tumour that were very close to the real location and extent of tumour observed in the whole mount radical prostatectomy specimens, therefore its implementation in clinical assessment and follow‐up should be considered. Copyright © 2011 John Wiley & Sons, Ltd.