Artificial Intelligence System for Predicting Prostate Cancer Lesions from Shear Wave Elastography Measurements

Area of Science:

• Medical imaging
• Artificial intelligence
• Oncology

Background:

• Prostate cancer diagnosis relies on biopsies, which can be invasive.
• Shear wave elastography (SWE) offers a non-invasive method to assess tissue stiffness.
• Integrating AI with SWE may enhance diagnostic accuracy for prostate cancer.

Purpose of the Study:

• To develop an artificial intelligence (AI) system for prostate cancer prediction.
• To evaluate the efficacy of bi-dimensional shear wave elastography (BDSWE) in conjunction with AI.
• To compare AI-driven BDSWE analysis with histopathological examination of prostate biopsy specimens.

Main Methods:

• A prospective study involving 356 patients undergoing prostate biopsy for cancer suspicion.
• Bi-dimensional shear wave ultrasonography was performed, followed by systematic transrectal prostate biopsy.
• Three machine learning algorithms (logistic regression, decision tree, dense neural network) were applied to elastographic data from 223 confirmed prostate cancer cases.

Main Results:

• The dense neural network achieved an Area Under the Curve (AUC) of 0.94.
• An ensemble model combining three machine learning algorithms reached a 98% accuracy rate.
• Upsampling the neural network's training set slightly improved its predictive performance.

Conclusions:

• Bi-dimensional shear wave elastography shows significant potential for predicting prostate cancer lesions.
• AI and machine learning algorithms substantially enhance the diagnostic utility of BDSWE.
• This AI-powered approach could serve as a valuable tool in prostate cancer diagnostics.