Applying Unet for extraction of vascular metrics from T1-weighted and T2-weighted MRI
Orooji, F.; Butler, R.
Show abstract
We apply deep learning to the problem of segmenting the arterial system from T1w and T2w images. We use the freely available 7-Tesla forrest dataset from OpenNeuro, (which contains TOF, T1w, and T2w) and use supervised learning with T1w or T2w as input, and TOF segmentation as ground truth, to train a Unet architecture capable of segmenting arteries and quantifying arterial diameters from T1w or T2w images alone. We demonstrate arterial segmentations from both T1w and T2w images, and show that T2w images have sufficient vessel contrast to estimate arterial diameters comparable to those estimated from TOF. We then apply our Unet to T2w images from a separate dataset (IXI) and show our model generalizes to images acquired at different field strength. We consider this work proof-of-concept that arterial segmentations can be derived from MRI sequences with poor contrast between arteries and surrounding tissue (T1w and T2w), due to the ability of deep convolutional networks to extract complex features based on local image intensity. Future work will focus on improving the generalizability of the network to non-forrest datasets, with the eventual goal of leveraging the entire pre-existing corpus of neuroimaging data for study of human cerebrovasculature.
Matching journals
The top 10 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
- Neural Network-derived perfusion maps: a Model-free approach to computed tomography perfusion in patients with acute ischemic stroke 96%
- Exploring white matter dynamics and morphology through interactive numerical phantoms: The White Matter Generator 94%
- 3D segmentation of perivascular spaces on T1-weighted 3 Tesla MR images with a convolutional autoencoder and a U-shaped neural network 94%
Similar papers in this journal
- Improved segmentation of the intracranial and ventricular volumes in populations with cerebrovascular lesions and atrophy using 3D CNNs 95%
- MRIO: The Magnetic Resonance Imaging Acquisition and Analysis Ontology 94%
- Regularized Bagged Canonical Component Analysis for Multiclass Learning in Brain Imaging 94%
Similar papers in this journal
- Automated 3D reconstruction of the fetal thorax in the standard atlas space from motion-corrupted MRI stacks for 21-36 weeks GA range 95%
- STAMP: Simultaneous Training and Model Pruning for Low Data Regimes in Medical Image Segmentation 95%
- MISPEL: A supervised deep learning harmonizationmethod for multi-scanner neuroimaging data 95%
Similar papers in this journal
- Radius-Optimized Efficient Template Matching for Lesion Detection from Brain Images 96%
- Segmentation of Pancreatic Ductal Adenocarcinoma (PDAC) and surrounding vessels in CT images using deep convolutional neural networks and Texture Descriptors 95%
- Probabilistic Brain MR Image Transformation Using Generative Models 95%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.