Fast fully automatic brain detection in fetal MRI using dense rotation invariant image descriptors
read more
Citations
Auto-Context Convolutional Neural Network (Auto-Net) for Brain Extraction in Magnetic Resonance Imaging
Fast Volume Reconstruction From Motion Corrupted Stacks of 2D Slices
A review on automatic fetal and neonatal brain MRI segmentation.
An automated framework for localization, segmentation and super-resolution reconstruction of fetal brain MRI.
References
Shape modeling with front propagation: a level set approach
Decision Forests for Computer Vision and Medical Image Analysis
3D Segmentation in the Clinic: A Grand Challenge
The Medical Imaging Interaction Toolkit
Related Papers (5)
Frequently Asked Questions (13)
Q2. How long does it take to acquire a stack of images?
Each acquisition of a 2D image takes approximately 0.5–1.0s, which makes throughplane movement artifacts very likely until the whole image stack is available.
Q3. What is the typical acquisition of a fetus?
A typical acquisition begins with a localizer scan, which is used to align the scan main axis approximately parallel to the fetus.
Q4. Why did the authors train a classification forest?
The authors train a state-of-the-art Classification Forest ensemble learning method based on decision trees for the SGD image features [11, 12], because of its trade-off between efficiency and classification performance.
Q5. What is the use of 3D template matching?
In [5, 6], 3D template matching is used to detect the eyes, enabling a subsequent 2D/3D graph-cut segmentation to extract the brain.
Q6. Why do the authors use 2D slices parallel to the smallest voxels?
The authors use 2D slices parallel to the two smallest voxel sides for this example because of the large through-plane resolution of their datasets (4mm) and the small volume of the fetus.
Q7. What are the common causes of the border slices?
These border slices are mainly responsible for higher average and maximum surface distance errors and could be left out in practice.
Q8. What are the limitations of the refinement method?
The maximum distance errors show that their refinement method has limitations in regions with large anatomical abnormalities, where an unusually large gradient will stop the level set evolution, and for slices where only a few voxels were detected as brain (border slices).
Q9. How many iterations of the method have been used?
1. An average DC of 0.850% before segmentation refinement shows that their base method is already highly robust and that it shows only a few to no outliers.
Q10. What makes an automatic evaluation of the scan difficult?
Fetal motion and its unpredictable nature (Fig. 2) put high demand on radiologists and make an automatic evaluation of the scan challenging.
Q11. What are the maximum distance errors in the voxel slices?
The level set over-segments in these peripheral slices because of the absence of a clear gradient as well as partial volume effects from the skull.
Q12. What is the way to describe the spherical 3D Gabor?
Skibbe et al. evaluate several different basis functions and show strong evidence that the power spectrum of the expansion with spherical 3D Gabor basisfunctions, represented by a superposition of Bessel functions Bls(k), are highly suitable for both classification accuracy and computation time when used for medical 3D images.
Q13. How many iterations are necessary to reduce the number of background voxels?
The datasets have been cropped to an average size of 200× 200× 70 voxels encompassing the whole fetal body in order to reduce the number of background voxels and therefore the computation time during testing.