Related Experiment Video
Updated: Aug 21, 2026
In Vivo Tracking of Edema Development and Microvascular Pathology in a Model of Experimental Cerebral Malaria Using Magnetic Resonance Imaging
Published on: June 8, 2017
A geometric-to-neural cascade for cerebral microbleed detection in susceptibility-weighted MRI
Sam Bogdanov1, Gaurav Rudravaram2, Adam M Saunders2
1Medical Scientist Training Program, Vanderbilt University, Nashville, TN 37240, USA.
Abstract:
Cerebral microbleeds (CMBs) are established imaging biomarkers of cerebral small vessel disease and are a defining feature of cerebral amyloid angiopathy (CAA), yet their automated detection in susceptibility-weighted imaging (SWI) remains challenging due to high false-positive rates from vessel cross-sections, iron and calcium deposits, and other hypointense mimics. We present a fully automated, three-stage cascade pipeline that combines subject-adaptive unsupervised candidate generation with two successive lightweight 3D ResNet classifiers, trained with only human-in-the-loop quality-assurance (QA) labels (yes/no per candidate) rather than dense voxel-wise segmentation masks. The candidate generation stage is performed by fitting a Gaussian Mixture Model (GMM) to each subject's SWI intensity histogram to define an adaptive low-intensity threshold, followed by anatomical masking to exclude physiologically irrelevant regions (image edges, ventricles/CSF/choroid plexus, and cerebellum), and filters candidates by size and sphericity. The model was trained and evaluated across a nested 3×5-fold cross-validation on N = 30 subjects from a publicly available labeled microbleed dataset and a CAA cohort (11,424 CMB candidate lesions) with data augmentation during training. Stage A classifies all geometric candidates as CMB or non-CMB and Stage B refines the predicted positives to suppress false positives (cascade AUC = 0.9587, sensitivity = 0.712, specificity = 0.975, PPV = 0.676, F1 = 0.693). The cascade reduces Stage A false positives by 76.8% (888/1,157 false positives eliminated) while retaining competitive sensitivity. Inference was performed on 141 SWI scans, detecting a mean 40.3 CMBs per scan and being preferred for use in 85% of high CMB cases, as evaluated by a blinded neurologist. The inference pipeline outputs binary CMB segmentation NIfTI images and radiologist-ready QA visualizations.
Insights
This study introduces an automated pipeline for detecting cerebral microbleeds (CMBs) in brain imaging, significantly reducing false positives and improving accuracy for diagnosing small vessel disease and cerebral amyloid angiopathy.
Area of Science:
- Neuroimaging
- Medical Image Analysis
- Artificial Intelligence in Medicine
Background:
- Cerebral microbleeds (CMBs) are key imaging biomarkers for cerebral small vessel disease and cerebral amyloid angiopathy (CAA).
- Automated CMB detection in susceptibility-weighted imaging (SWI) is hindered by mimics like vessel cross-sections, leading to high false-positive rates.
- Current methods often require complex, time-consuming voxel-wise segmentation, limiting clinical applicability.
Purpose of the Study:
- To develop and validate a fully automated, three-stage cascade pipeline for accurate CMB detection in SWI.
- To minimize false positives by employing a novel approach using quality-assurance labels instead of dense segmentation masks.
- To provide a computationally efficient and clinically relevant tool for CMB analysis.
Main Methods:
- A three-stage cascade pipeline combining unsupervised candidate generation (Gaussian Mixture Model, anatomical masking, size/sphericity filtering) and two 3D ResNet classifiers.
- Training utilized human-in-the-loop quality-assurance labels (yes/no per candidate) on a combined dataset of 30 subjects (11,424 candidates).
- Nested 3x5-fold cross-validation and data augmentation were employed during model training and evaluation.
Main Results:
- The cascade pipeline achieved a high AUC of 0.9587, with sensitivity 0.712, specificity 0.975, PPV 0.676, and F1 score 0.693.
- The system significantly reduced false positives by 76.8% compared to initial candidates while maintaining competitive sensitivity.
- Inference on 141 scans detected a mean of 40.3 CMBs per scan, with a neurologist preferring the output in 85% of high CMB cases.
Conclusions:
- The developed automated pipeline offers a robust and efficient method for CMB detection in SWI.
- This approach effectively mitigates false positives, enhancing diagnostic accuracy for small vessel disease and CAA.
- The pipeline's output, including NIfTI segmentations and visualizations, is suitable for clinical radiologist review.
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
05:23Construction and Application of Cerebral Functional Region-Based Cerebral Blood Flow Atlas Using Magnetic Resonance Imaging-Arterial Spin Labeling
Published on: May 31, 2024