Related Experiment Video
Updated: May 27, 2025
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
ProstAtlasDiff: Prostate cancer detection on MRI using Diffusion Probabilistic Models guided by population spatial
Cynthia Xinran Li1, Indrani Bhattacharya2, Sulaiman Vesal3
1Institute of Computational and Mathematical Engineering, Stanford University, Stanford, CA 94305, USA.
Abstract:
Magnetic Resonance Imaging (MRI) is increasingly being used to detect prostate cancer, yet its interpretation can be challenging due to subtle differences between benign and cancerous tissue. Recently, Denoising Diffusion Probabilistic Models (DDPMs) have shown great utility for medical image segmentation, modeling the process as noise removal in standard Gaussian distributions. In this study, we further enhance DDPMs by introducing the knowledge that the occurrence of cancer varies across the prostate (e.g., ∼70% of prostate cancers occur in the peripheral zone). We quantify such heterogeneity with a registration pipeline to calculate voxel-level cancer distribution mean and variances. Our proposed approach, ProstAtlasDiff, relies on DDPMs that use the cancer atlas to model noise removal and segment cancer on MRI. We trained and evaluated the performance of ProstAtlasDiff in detecting clinically significant cancer in a multi-institution multi-scanner dataset, and compared it with alternative models. In a lesion-level evaluation, ProstAtlasDiff achieved statistically significantly higher accuracy (0.91 vs. 0.85, p<0.001), specificity (0.91 vs. 0.84, p<0.001), positive predictive value (PPV, 0.50 vs. 0.35, p<0.001), compared to alternative models. ProstAtlasDiff also offers more accurate cancer outlines, achieving a higher Dice Coefficient (0.33 vs. 0.31, p<0.01). Furthermore, we evaluated ProstAtlasDiff in an independent cohort of 91 patients who underwent radical prostatectomy to compare its performance to that of radiologists, relative to whole-mount histopathology ground truth. ProstAtlasDiff detected 16% (15 lesions out of 93) more clinically significant cancers compared to radiologists (sensitivity: 0.90 vs. 0.75, p<0.01), and was comparable in terms of ROC-AUC, PR-AUC, PPV, accuracy, and Dice coefficient (p≥0.05). Furthermore, we evaluated ProstAtlasDiff in a second independent cohort of 537 subjects and observed that ProsAtlasDiff outperformed alternative approaches. These results suggest that ProstAltasDiff has the potential to assist in localizing cancer for biopsy guidance and treatment planning.
Insights
ProstAtlasDiff, a novel approach using Denoising Diffusion Probabilistic Models (DDPMs) and prostate cancer atlases, significantly improves the detection and segmentation of clinically significant prostate cancer on MRI scans.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Magnetic Resonance Imaging (MRI) is crucial for prostate cancer detection, but distinguishing benign from cancerous tissue is challenging.
- Denoising Diffusion Probabilistic Models (DDPMs) have shown promise in medical image segmentation.
- Existing methods lack the ability to incorporate known patterns of prostate cancer distribution.
Purpose of the Study:
- To develop and evaluate ProstAtlasDiff, an enhanced DDPM approach for prostate cancer segmentation on MRI.
- To integrate prostate cancer heterogeneity knowledge into DDPMs for improved detection accuracy.
- To assess ProstAtlasDiff's performance against alternative models and radiologists.
Main Methods:
- Developed ProstAtlasDiff, a DDPM-based model incorporating a prostate cancer atlas to guide noise removal.
- Quantified cancer distribution heterogeneity using a registration pipeline for voxel-level mean and variance calculations.
- Trained and validated ProstAtlasDiff on multi-institution, multi-scanner datasets and compared it with existing models and expert radiologists.
Main Results:
- ProstAtlasDiff demonstrated statistically significant improvements in accuracy (0.91 vs. 0.85), specificity (0.91 vs. 0.84), and PPV (0.50 vs. 0.35) compared to alternative models.
- Achieved higher Dice Coefficient (0.33 vs. 0.31) for more accurate cancer outlines.
- In comparison to radiologists, ProstAtlasDiff detected 16% more clinically significant cancers (sensitivity: 0.90 vs. 0.75) and showed comparable performance in other metrics.
Conclusions:
- ProstAtlasDiff significantly enhances prostate cancer detection and segmentation accuracy on MRI by leveraging cancer distribution knowledge.
- The model shows potential to outperform current methods and assist radiologists in biopsy guidance and treatment planning.
- Further validation in larger cohorts confirms ProstAtlasDiff's superiority over alternative approaches.
