Clinically Significant Prostate Cancer Prediction Using Multimodal Deep Learning with Prostate-Specific Antigen...

Area of Science:

• Oncology
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Prostate cancer (PCa) exhibits significant clinical heterogeneity.
• Accurate prediction of clinically significant PCa is crucial, especially in patients with low-to-intermediate prostate-specific antigen (PSA) levels.
• Early identification of aggressive PCa aids in timely and personalized treatment strategies.

Purpose of the Study:

• To evaluate the predictive accuracy of a multimodal deep learning approach for clinically significant PCa.
• To assess the model's performance in patients with PSA levels less than or equal to 20 ng/mL.
• To demonstrate the utility of AI in improving pre-biopsy risk stratification for PCa.

Main Methods:

• A cohort of 178 patients undergoing ultrasound-guided prostate biopsy was analyzed.
• Multimodal medical data, including imaging, was integrated into a deep learning model.
• Receiver operating characteristic curves and area under the curve (AUC) were used to assess predictive accuracy.

Main Results:

• The multimodal deep learning model achieved an AUC of 0.878 for predicting PCa in all patients.
• In patients with PSA ≤ 20 ng/mL, the model demonstrated an AUC of 0.862, showing robust performance.
• Analysis of representative false-negative and false-positive cases provided insights into model behavior.

Conclusions:

• Multimodal deep learning effectively predicts clinically significant PCa, particularly in patients with PSA ≤ 20 ng/mL.
• This AI-driven approach can assist physicians in treatment strategy determination before biopsy.
• The study supports the integration of deep learning for personalized PCa management workflows.