Related Experiment Video
Updated: Jul 16, 2025
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
272
Evaluation of a Deep Learning-based Algorithm for Post-Radiotherapy Prostate Cancer Local Recurrence Detection Using
Enis C Yilmaz1, Stephanie A Harmon1, Mason J Belue1
1Molecular Imaging Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD, United States.
European Journal of Radiology
|September 17, 2023
Summary
A biparametric MRI (bpMRI)-based artificial intelligence (AI) model shows promise in detecting local prostate cancer recurrence after radiotherapy. While performing below radiologists, the AI model identified most recurrent lesions, especially after external beam radiation treatment.
Area of Science:
- Radiology
- Artificial Intelligence
- Oncology
Background:
- Prostate cancer (PCa) recurrence after radiotherapy poses a diagnostic challenge.
- Biparametric MRI (bpMRI) is increasingly used for PCa assessment.
- Artificial intelligence (AI) models are being developed to enhance diagnostic accuracy.
Purpose of the Study:
- To evaluate the efficacy of a bpMRI-based AI model in detecting local prostate cancer recurrence in patients previously treated with radiotherapy.
- To compare the AI model's performance against a prospective radiologist assessment.
Main Methods:
- Retrospective analysis of 62 post-radiotherapy patients who underwent bpMRI and subsequent biopsy.
- A bpMRI-based AI model, initially developed for treatment-naïve patients, was applied to detect recurrent PCa.
- Performance metrics (sensitivity) of the AI model were compared to radiologist interpretation using the Wald test.
- Subgroup analyses were performed based on treatment modality (EBRT vs. brachytherapy) and prostate volume.
Main Results:
- The AI model detected 40 of 56 recurrent PCa foci in 35 patients.
- Radiologist interpretation demonstrated higher sensitivity than the AI model on both patient-level (91.3% vs. 76.1%) and lesion-level (87.5% vs. 71.4%).
- The AI model showed improved performance in the external beam radiation treatment (EBRT) group and in patients with larger prostate volumes (>34 ml).
Conclusions:
- The bpMRI-based AI model successfully detected a majority of locally recurrent prostate cancer after radiotherapy.
- The AI model's performance was lower than expert radiologist interpretation.
- External validation and further testing are recommended before clinical implementation of the AI model for post-radiotherapy PCa recurrence detection.
