Related Experiment Video
Updated: Jul 5, 2025
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
Deep Learning-Based Detection and Classification of Bone Lesions on Staging Computed Tomography in Prostate Cancer: A
Mason J Belue1, Stephanie A Harmon1, Dong Yang2
1Molecular Imaging Branch, National Cancer Institute, National Institutes of Health, 10 Center Dr., MSC 1182, Building 10, Room B3B85, Bethesda, Maryland, USA (M.J.B., S.A.H., N.S.L., E.C.Y., T.E.P., B.S., L.L., E.M., P.L.C., B.T.).
This study developed an artificial intelligence (AI) model to detect and classify bone lesions in prostate cancer (PCa) staging CT scans. The AI demonstrated performance comparable to radiologists in identifying metastatic versus benign lesions.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Oncology
- Prostate Cancer Diagnostics
Background:
- Detecting metastatic bone lesions in prostate cancer (PCa) staging CT scans is critical but time-consuming for experts.
- Current methods often require additional imaging like PET/CT, increasing costs and patient burden.
- Automated solutions are needed to improve efficiency and accuracy in lesion characterization.
Purpose of the Study:
- To develop and evaluate an ensemble of two deep learning AI models for bone lesion detection, segmentation, and classification on staging CTs in PCa patients.
- To compare the performance of the developed AI models against radiologists in distinguishing benign from metastatic bone lesions.
Main Methods:
- Developed two AI models: 3DAISeg for lesion segmentation and 3DAIClass for classification, using 297 staging CT scans from PCa patients.
- Validated metastatic lesions through follow-up scans, bone biopsy, or PET/CT.
- Assessed AI performance using Dice similarity coefficient, F1-score, and accuracy, and conducted a multi-reader study with junior and senior radiologists.
Main Results:
- The AI models achieved 100% patient-level detection of metastatic lesions in unseen CT scans.
- The classification AI (3DAIClass) demonstrated F1-scores of 94.8% with radiologist contours and 92.4% with AI contours.
- 3DAIClass showed comparable positive predictive value (PPV) and negative predictive value (NPV) to junior and senior radiologists.
Conclusions:
- The developed AI model for lesion detection and classification performs on par with radiologists in identifying benign and metastatic bone lesions on PCa staging CTs.
- This AI ensemble offers a promising tool to aid in the efficient and accurate assessment of bone metastases in prostate cancer care.
