Related Experiment Video
Updated: Sep 21, 2025
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
Automated Detection of Brain Metastases on T1-Weighted MRI Using a Convolutional Neural Network: Impact of Volume
Gabriel Chartrand1, Ramón D Emiliani1, Sophie A Pawlowski1
1AFX Medical Inc., Montréal, Canada.
Background:
Detection of brain metastases (BM) and segmentation for treatment planning could be optimized with machine learning methods. Convolutional neural networks (CNNs) are promising, but their trade-offs between sensitivity and precision frequently lead to missing small lesions.
Hypothesis:
Combining volume aware (VA) loss function and sampling strategy could improve BM detection sensitivity.
Study Type:
Retrospective.
Population:
A total of 530 radiation oncology patients (55% women) were split into a training/validation set (433 patients/1460 BM) and an independent test set (97 patients/296 BM).
Field Strength/Sequence:
1.5 T and 3 T, contrast-enhanced three-dimensional (3D) T1-weighted fast gradient echo sequences.
Assessment:
Ground truth masks were based on radiotherapy treatment planning contours reviewed by experts. A U-Net inspired model was trained. Three loss functions (Dice, Dice + boundary, and VA) and two sampling methods (label and VA) were compared. Results were reported with Dice scores, volumetric error, lesion detection sensitivity, and precision. A detected voxel within the ground truth constituted a true positive.
Statistical Tests:
McNemar's exact test to compare detected lesions between models. Pearson's correlation coefficient and Bland-Altman analysis to compare volume agreement between predicted and ground truth volumes. Statistical significance was set at P ≤ 0.05.
Results:
Combining VA loss and VA sampling performed best with an overall sensitivity of 91% and precision of 81%. For BM in the 2.5-6 mm estimated sphere diameter range, VA loss reduced false negatives by 58% and VA sampling reduced it further by 30%. In the same range, the boundary loss achieved the highest precision at 81%, but a low sensitivity (24%) and a 31% Dice loss.
Data Conclusion:
Considering BM size in the loss and sampling function of CNN may increase the detection sensitivity regarding small BM. Our pipeline relying on a single contrast-enhanced T1-weighted MRI sequence could reach a detection sensitivity of 91%, with an average of only 0.66 false positives per scan.
Evidence Level:
3 TECHNICAL EFFICACY: Stage 2.
Insights
Machine learning, specifically convolutional neural networks (CNNs), can improve brain metastasis (BM) detection. Combining volume-aware loss and sampling strategies significantly enhanced sensitivity for detecting small BM on MRI scans.
Area of Science:
- Radiology
- Medical Imaging
- Machine Learning
Background:
- Brain metastases (BM) detection and segmentation are crucial for treatment planning.
- Convolutional neural networks (CNNs) show promise but struggle with detecting small lesions due to sensitivity-precision trade-offs.
Purpose of the Study:
- To optimize BM detection and segmentation using machine learning.
- To improve the sensitivity of CNNs for detecting small brain metastases.
Main Methods:
- A U-Net inspired CNN model was trained on 530 radiation oncology patients' contrast-enhanced T1-weighted MRI scans.
- Compared Dice, Dice+boundary, and volume-aware (VA) loss functions with label and VA sampling strategies.
- Evaluated performance using Dice scores, volumetric error, sensitivity, and precision.
Main Results:
- The combination of VA loss and VA sampling achieved the highest sensitivity (91%) and precision (81%).
- VA loss reduced false negatives by 58% for small BM (2.5-6 mm diameter).
- VA sampling further reduced false negatives by 30% for small BM.
Conclusions:
- Incorporating BM size into CNN loss and sampling functions enhances detection sensitivity for small lesions.
- The proposed pipeline, using a single MRI sequence, achieved 91% detection sensitivity with minimal false positives (0.66 per scan).
More Related Videos
12:50Lesion Explorer: A Video-guided, Standardized Protocol for Accurate and Reliable MRI-derived Volumetrics in Alzheimer's Disease and Normal Elderly
Published on: April 14, 2014
06:44Modeling Brain Metastases Through Intracranial Injection and Magnetic Resonance Imaging
Published on: June 7, 2020