Article

Artificial Intelligence may enhance efficacy of mpMRI in prostate cancer

Author(s):

Researchers examined the impact of deep learning–based AI software upon radiologists reviewing multiparametric MRI scans in the assessment of clinically significant prostate cancer.

A study published in Insights into Imaging showed the potential of artificial intelligence (AI)–based software to facilitate improved sensitivity and specificity rates with multiparametric magnetic resonance imaging (mpMRI) in the assessment of clinically significant prostate cancer (csPCa).1

“Our findings suggest that the ability of radiologists to detect difficult lesions was significantly improved with the help of AI. Given the significant improvement in the detection of MRI-visible csPCa lesions (with adjunctive AI), the AI-aided results have greater value in guiding the localization of biopsy,” maintained study co-author Xiaoying Wang, MD, PhD, and colleagues.

“Our findings suggest that the ability of radiologists to detect difficult lesions was significantly improved with the help of AI. Given the significant improvement in the detection of MRI-visible csPCa lesions (with adjunctive AI), the AI-aided results have greater value in guiding the localization of biopsy,” maintained study co-author Xiaoying Wang, MD, PhD, and colleagues.

For the retrospective multicenter study, researchers examined the impact of a deep learning-based AI software upon 16 radiologists reviewing mpMRI scans for csPCa. The radiologist experience with interpretation of mpMRI ranged from 1 to 5 years. Out of 480 mpMRI images, 180 cases revealed a total of 349 csPCa lesions, according to the study. The study authors noted that each radiologist diagnosed 30 cases twice, one time with AI and one time without AI with a four-week interval between the reviews.

For csPCa lesion level sensitivity, researchers noted an increase from 40.1% without AI to 59% with AI. Adjunctive AI also led to a 14% increase (from 57.7% to 71.7%) in patient level specificity for csPCa while maintaining sensitivity (88.3% without AI to 93.9% with AI), according to the study authors.1

“The results showed that AI software substantially improved the lesion-level and patient-level specificity of the readers while preserving patient-level sensitivity in detecting MRI-visible csPCa,” wrote study co-author Xiaoying Wang, MD, PhD, who is affiliated with the Department of Radiology at Peking University First Hospital in Beijing, China, and colleagues.

The researchers also found that adjunctive AI had a 21.7% higher accuracy rate in identifying true positive lesions (68.2%) in comparison to 46.5 % for radiologist assessment without AI. For true-positive patient diagnosis, the study authors noted that adjunctive AI had a 66.1% detection rate in comparison to 55.8% for radiologists without AI.1

Citing research that demonstrates the ability of radiologists to detect large prostate lesions without AI assistance, Wang and colleagues said they excluded cases involving prominent lesions and advanced cancer.2 While they noted this may have contributed to underestimation of radiologist assessment without AI, the researchers emphasize this also conveys the ability of the AI software to assist in diagnosing more challenging presentations of prostate cancer.

“Our findings suggest that the ability of radiologists to detect difficult lesions was significantly improved with the help of AI. Given the significant improvement in the detection of MRI-visible csPCa lesions (with adjunctive AI), the AI-aided results have greater value in guiding the localization of biopsy,” maintained Wang and colleagues.

Beyond the inherent limitations of a retrospective study, the authors acknowledged the potential for bias with annotations from a reference standard based on cognitive fusion biopsy. They also noted challenges with stratification analysis due to wide variety in the experiences of the reviewing radiologists.

References

1. Sun Z, Wang K, Kong Z, et al. A multicenter study of artificial intelligence-aided software for detecting visible clinically significant prostate cancer on mpMRI. Insights Imaging. 2023 Apr 30;14(1):72. doi: 10.1186/s13244-023-01421-w.

2. Zhu L, Gao G, Liu Y, et al. Feasibility of integrating computer-aided diagnosis with structured reports of prostate multiparametric MRI. Clin Imaging. 2020;60(1):123-130.

Related Videos
Prostate cancer cells | Image Credit: © Dr_Microbe - stock.adobe.com
Eric Li, MD, answers a question during a Zoom video interview
Prostate cancer cells dividing | Image Credit: © PRB ARTS - stock.adobe.com
© 2024 MJH Life Sciences

All rights reserved.