Neuroradiological projects
3D classification of pseudoprogression and true progression in glioblastoma
For people with glioblastoma who are undergoing chemotherapy and radiation therapy, a critical and unresolved challenge is telling pseudoprogression from true progression. A misdiagnosis can lead to incorrect treatments, misinterpreted trial results and increased anxiety in patients.
This project aims to develop a model to tell glioblastoma's true progression from pseudoprogression using MRI.
Automated MRI segmentation for vestibular schwannomas
This project addresses the challenges of manual tumor tracking by benchmarking state-of-the-art architectures, specifically focusing on self-optimizing convolution neural networks and transformer-based models. By training and evaluating these models across large-scale, heterogeneous MRI datasets, the research ensures robust generalization and clinical reliability for real-world applications.
The resulting automated volumetry provides highly accurate and reproducible monitoring of tumor growth. This is essential for optimizing serial imaging and treatment planning for patients with vestibular schwannomas.
Two-stage workflow for vestibular schwannoma segmentation
The workflow first detects the region of interest from the full MRI volume. Then it performs dedicated vestibular schwannoma segmentation within the localized region.
Classification of brain lesions
This project uses deep learning with MRI to group tumefactive brain lesions into these classifications:
- Gliomas.
- Metastases.
- Lymphomas.
- Tumefactive multiple sclerosis.
Head CT scans
This fully automated anatomical segmentation of head CT scans was obtained with the lab's deep learning model.
Deep brain structure segmentation
The project developed a U-Mamba-based model for automated brain substructure segmentation using T1-weighted MRI data and data from the Alzheimer's Disease Neuroimaging Initiative. The model outperformed state-of-the-art methods, providing robust tools for analyzing brain structure-disease relationships and supporting future clinical applications in neuroimaging and disease mechanism research.
Deep learning-based segmentation of 122 brain MRI subregions
This image illustrates the deep learning workflow for brain MRI segmentation across 122 subregions, achieving a median Dice similarity coefficient of 0.9112.
Deep learning to determine gene status in glioblastoma multiforme
Dr. Erickson's team is developing a deep learning model to predict methylation status in the MGMT gene in people with glioblastoma multiforme. This model uses different brain MRI sequences.
Deep learning tool for the early diagnosis of progressive supranuclear palsy
This project investigates structural and volumetric changes in the brainstem and midbrain in patients with progressive supranuclear palsy versus patients without the disease and patients with Parkinson's disease. The lab is creating two segmentation models and one classifier for this project.
Fully automatic deep learning framework for subarachnoid hemorrhage detection on noncontrast head CT scans
This project develops a deep learning model to automatically segment and quantify hemorrhage volume in patients with subarachnoid hemorrhage using CT scans. Based on 152 manually annotated Mayo Clinic cases, the customized nnU-Net V2 model achieved accurate subarachnoid hemorrhage segmentation, showed strong agreement with physician annotations and enabled rapid inference in approximately 6.7 seconds per case.
Automated deep learning segmentation and quantification of subarachnoid hemorrhage
This deep learning model enables automatic subarachnoid hemorrhage segmentation and volume measurement, closely matching ground truth and tracking posttreatment reduction.