Author

Overview of our proposed pre-training framework.
Input CT images are randomly cropped into sub-volumes and
augmented with random inner cutout and rotation, then fed to the
Swin UNETR encoder as input. We use masked volume inpainting,
contrastive learning and rotation prediction as proxy tasks for
learning contextual representations of input images

Self-Supervised Pre-Training of Swin Transformers for 3D Medical Image Analysis

Jul. 25, 2022—Tang, Yucheng, Dong Yang, Wenqi Li, Holger R. Roth, Bennett Landman, Daguang Xu, Vishwesh Nath, and Ali Hatamizadeh. “Self-supervised pre-training of swin transformers for 3d medical image analysis.” In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 20730-20740. 2022. Full text: Abstract Vision Transformers (ViT)s have shown great performance in self-supervised...

Representative images are predicted to associate with comorbidities and ICD-10 codes
(phenotype components) identified in each risk category. The red outlines show the pancreas
tissue can be different under phenotyping contexts. (1) is from a nominally healthy pancreas
group with potential lung infections; (2) is from type I diabetes and other chronic kidney disease
patients with atrophic pancreas; (3) is from other metabolic syndromes including type II diabetes;
(4) is from patients with weight loss and pancreatitis.

Pancreas CT Segmentation by Predictive Phenotyping

Dec. 14, 2021—Y. Tang, R.Gao, H.H.Lee, Q.Yang, X.Yu,Y.Zhou, S.Bao, Y.Huo, J.Spraggins, J.Virostko, Z.Xu, B.A. Landman. “Pancreas CTSegmentation by Predictive Phenotyping”. International Conference on MedicalImage Computing and Computer Assisted Intervention(MICCAI), 2021 Full Text: https://link.springer.com/chapter/10.1007/978-3-030-87193-2_3 Abstract Pancreas CT segmentation offers promise at understanding the structural manifestation of metabolic conditions. To date, the medical primary record of conditions that impact...

Method framework. Given a CT scan with at high resolution of ~0.8 x ~0.8 x ~2 mm, a low-res section (left panel) is trained with multi-channel segmentation. The low-res part contains down-sampling and normalization in order to preserve the complete spatial information. After the coarse segmentations are acquired from low- res UNet, we interpolate the mask to match the image’s original resolution. Next, random patch sampling (mid panel) is employed to collect patches, and patches are concatenated with corresponding coarse segmentation masks. Finally, we trained a patch-based high-res (right panel) segmentation model, the high-dimensional probability maps are acquired from integration of all patches on field of views. Majority vote is used to merge estimates into a final segmentation.

High-resolution 3D abdominal segmentation with random patch network fusion

Dec. 14, 2021—Y. Tang,R.Gao,S.Han, Y.Chen, D.Gao, V.Nath, C.Bermudez, M.R. Savona, R.G. Abramson, S.Bao,I.Lyu, Y.Huo and B.A. Landman,“High-resolution 3D Abdominal Segmentation with Random PatchNetworkFusion”,Medical Image Analysis, 2021. Full Text: https://www.sciencedirect.com/science/article/pii/S1361841520302589 Abstract Deep learning for three dimensional (3D) abdominal organ segmentation on high-resolution computed to- mography (CT) is a challenging topic, in part due to the limited memory provide...

A representative subject was evaluated with URN (left) and BUSN (right). Green scatters are inliers of influence to the regression, yellow
scatters are outliers of no influence to the distributed data. Darker blue line indicates the normal linear regression on scatters points, lighter blue
line is the RANSAC regressor result according to inliers. Left panel presents the single URN regression with amounts of outliers result in failure of
linearity nature in chest and pelvis regions. Right panel shows the testing result of BUSN method, the distributed scores follows good linearity in
chest, abdomen and pelvis regions in CT scan. In summary, BUSN takes advantage of self-supervised network, which presents better continuity in
regression result among neighbor slices and shows scatter plots without number of outliers.

Body Part Regression With Self-Supervision

Jan. 14, 2021—Y.Tang, R.Gao, S.Han, Y.Chen, D.Gao, V.Nath, C.Bermudez, M.R. Savona, R.G. Abramson, S.Bao,I.Lyu, Y.Huo and B.A. Landman,”Body Part Regression with Self-supervision”,IEEETransactions onMedicalImaging,2021 Full Text: https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9350603 Abstract Body part regression is a promising new technique that enables content navigation through selfsupervised learning. Using this technique, the global quantitative spatial location for each axial view slice is obtained...

Prediction of Type II Diabetes Onset with Computed Tomography and Electronic Medical Records

Dec. 7, 2020—Yucheng Tang, Riqiang Gao, Ho Hin Lee, Quinn Stanton Wells, Ashley Spann, James Gregory Terry, Jeff Carr, Yuankai Huo, Shunxing Bao and Bennett A. Landman, “Prediction of Type II Diabetes Onset with Computed Tomography and Electronic Medical Records”, MICCAI CLIP, 2020. Full Text: Abstract Type II diabetes mellitus (T2DM) is a significant public health concern...

Learning from dispersed manual annotations with an optimized data weighting policy

Dec. 7, 2020—Yucheng Tang, Riqiang Gao, Yunqiang Chen, Dashan Gao, Michael R. Savona, Richard G. Abramson, Shunxing Bao, Yuankai Huo and Bennett A. Landman, “Learning from Dispersed Manual Annotations with an Optimized Data Weighting Policy”, Journal of Medical Imaging, 2020. Full Text: Abstract https://pubmed.ncbi.nlm.nih.gov/32775501/ Purpose: Deep learning methods have become essential tools for quantitative interpretation of medical...

Figure 1. Renal structures segmentation of clinically acquired CT image. Yellow box shows a representative right kidney, green box shows the left kidney. The zoom-in patches show the CT kidney field-of-view (FOV) and labels. The label patches display the renal cortex (blue), the medulla (red) and the pelvicalyceal system (green). Those small structures make automatic renal segmentation challenging.

Renal Cortex, Medulla and Pelvicaliceal System Segmentation on Arterial Phase CT Images with Random Patch-based Networks

Nov. 21, 2020—Yucheng Tang, Riqiang Gao, Ho Hin Lee, Brent V. Savoie, Shunxing Bao, Yuankai Huo, Jeffrey Spraggins and Bennett A, Landman, Renal Cortex, Medulla, Pelvis Segmentation on Arterial Phase CT Images with Random Patch-based Networks, SPIE 2021 Medical Imaging Full Text Abstract Renal segmentation on contrast-enhanced computed tomography (CT) provides distinct spatial context and morphology. Current studies for renal segmentations...

Validation and Optimization of Multi-Organ Segmentation on Clinical Imaging Archives

Feb. 12, 2020—Olivia Tang, Yuchen Xu, Yucheng Tang, Ho Hin Lee, Yunqiang Chen, Dashan Gao, Shizhong Han, Riqiang Gao, Michael R. Savona, Richard G. Abramson, Yuankai Huo, Bennett A. Landman, “Validation and Optimization of Multi-Organ Segmentation on Clinical Imaging Archives”, SPIE IP:MI 2020. Houston, TX. https://arxiv.org/abs/2002.04102 Abstract Segmentation of abdominal computed tomography (CT) provides spatial context, morphological...

Examples of segmentations generated by the baseline algorithm. A. An inlier, where the algorithm correctly predicted larger organs, like the liver (on the left side of image in purple) and spleen (on the right side of image in red- pink) and suggested mostly accurate areas of smaller organs. B. An outlier (global failure), where the liver (purple) and spleen (red-pink) were largely correct but major inconsistencies are visible with other organs.

Outlier Guided Optimization of Abdominal Segmentation

Feb. 12, 2020—Yuchen Xu*, Olivia Tang*, Yucheng Tang, Ho Hin Lee, Yunqiang Chen, Dashan Gao, Shizhong Han, Riqiang Gao, Michael R. Savona, Richard G. Abramson, Yuankai Huo, Bennett A. Landman, “Outlier Guided Optimization of Abdomen Segmentation”, SPIE IP:MI 2020. Houston, TX https://arxiv.org/abs/2002.04098 Abstract Abdominal multi-organ segmentation of computed tomography (CT) images has been the subject of extensive...