Medical-image Analysis and Statistical Interpretation (MASI) Lab

Qualitative results of the DN segmentation. Here, we show the output of the automatic segmentation using unimodal U-Net, a multisequence U-Net, and single atlas registration of the SUIT template in the same subject. The manual label is shown in green while the predicted label is shown in red. Overlap between the two structures is shown in yellow. The Dice coefficient is 0.95 for the unimodal T1 model, 0.95 for the unimodal T2 model, 0.97 for the unimodal FA model, 0.93 for the multisequence model, 0.41 for the SUIT atlas, and 0.43 for the multi-atlas segmentation. Note: the multiatlas segmentation results were registered to the SUIT template for visualization.

Using deep learning for a diffusion-based segmentation of the dentate nucleus and its benefits over atlas-based methods

Dec. 6, 2019—Noguera, C. B., Bao, S., Petersen, K. J., Lopez, A. M., Reid, J., Plassard, A. J., … & Landman, B. A. (2019). Using deep learning for a diffusion-based segmentation of the dentate nucleus and its benefits over atlas-based methods. Journal of Medical Imaging, 6(4), 044007. Full Text: https://www.ncbi.nlm.nih.gov/pubmed/31824980 Abstract The dentate nucleus (DN) is a...

Example of Internal-transfer Weighting (ITW) of multi-task learning. The LC, AA, CB, COPD, E represent the tasks of lung cancer, adult asthma, chronic bronchitis, chronic obstructive pulmonary disease and emphysema, respectively. The red number is the loss weight for the task. Note that we apply PFLP on top of the displayed weights.

Internal-transfer Weighting of Multi-task Learning for Lung Cancer Detection

Dec. 5, 2019—Yiyuan Yang, Riqiang Gao*, Yucheng Tang, Sanja L. Antic, Steve Deppen, Yuankai Huo, Kim L. Sandler, Pierre P. Massion, Bennett A. Landman, “Internal-transfer Weighting of Multi-task Learning for Lung Cancer Detection,” SPIE MI:IP 2020. Houston, TX. [full text] Abstract Deep learning has achieved many successes in medical imaging, including lung nodule segmentation and lung cancer prediction on...

Vision for joint consideration of diagnosis and cancer-free progression time (CFPT) in cancer and non-cancer patients. The blue dashed lines represent the defined CFPT. The red number (%) in small rectangle is the risk score computed by the PLCOm2012 [3] model. Current scores suggest screening at every point since they indicate high risk (>= 2%). If CFPT were known, the green rectangles could have been avoided.

Deep Multi-task Prediction of Lung Cancer and Cancer-free Progression from Censored Heterogenous Clinical Imaging

Dec. 5, 2019—Riqiang Gao, Lingfeng Li, Yucheng Tang, Sanja L. Antic, Alexis B. Paulson, Yuankai Huo, Kim L. Sandler, Pierre P. Massion, Bennett A. Landman, “Deep Multi- task Prediction of Lung Cancer and Cancer-free Progression from Censored Heterogenous Clinical Imaging,” SPIE MI:IP 2020. Houston, TX. [full text] Abstract Annual low dose computed tomography (CT) lung screening is currently advised for individuals...

Improving Human Cortical Sulcal Curve Labeling in Large Scale Cross-Sectional MRI using Deep Neural Networks

Nov. 11, 2019—Prasanna Parvathaneni, Vishwesh Nath, Maureen McHugo, Yuankai Huo, Susan Resnick, Neil Woodward, Bennett Landman, Ilwoo Lyu. “Improving Human Cortical Sulcal Curve Labeling in Large Scale Cross-Sectional MRI using Deep Neural Networks”. Journal of Neuroscience Methods, 324, 108311, 2019. [Full text] Abstract BACKGROUND: Human cortical primary sulci are relatively stable landmarks and commonly observed across the population....

Hierarchical Spherical Deformation for Cortical Surface Registration

Nov. 11, 2019—Ilwoo Lyu, Hakmook Kang, Neil D. Woodward, Martin A. Styner and Bennett A. Landman. “Hierarchical Spherical Deformation for Cortical Surface Registration”. Medical Image Analysis, 57, pp. 72-88, 2019. [Full text][Code][Docker] Abstract We present hierarchical spherical deformation for a group-wise shape correspondence to address template selection bias and to minimize registration distortion. In this work, we...

The framework of DLSTM (three “steps” in the example). x_t is the input data at time point t, and d_t is the time distance from the time point t to the latest time point. “F” represents the learnable DLSTM component (convolutional version in this paper). H_t and C_t are the hidden state and cell state, respectively. The input data, x_t, could be 1D, 2D, or 3D.

Distanced LSTM: Time-Distanced Gates in Long Short-Term Memory Models for Lung Cancer Detection

Oct. 13, 2019—Gao, R., Huo, Y., Bao, S., Tang, Y., Antic, S.L., Epstein, E.S., Balar, A.B., Deppen, S., Paulson, A.B., Sandler, K.L. and Massion, P.P., 2019, October. Distanced LSTM: Time-Distanced Gates in Long Short-Term Memory Models for Lung Cancer Detection. In International Workshop on Machine Learning in Medical Imaging (pp. 310-318). Springer, Cham. [Full Text] Abstract The field of...

Transfer learning pipeline for out of sample whole brain segmentation

Generalizing Deep Whole Brain Segmentation for Pediatric and Post-Contrast MRI with Augmented Transfer Learning

Aug. 13, 2019—Bermudez, C., Blaber, J., Remedios, S.W., Reynolds, J.E., Lebel, C., McHugo, M., Heckers, S., Huo, Y., Landman, B.A. Generalizing Deep Whole Brain Segmentation for Pediatric and Post-Constrast MRI with Augmented Transfer Learning. SPIE Medical Imaging: Image Processing 2020. Houston, TX. Full Text: NIHMSID Abstract Generalizability is an important problem in deep neural networks, especially in...

Anatomical context improves deep learning on the brain age estimation task

Jul. 12, 2019—Bermudez, C., Plassard, A. J., Chaganti, S., Huo, Y., Aboud, K. E., Cutting, L. E., … & Landman, B. A. (2019). Anatomical context improves deep learning on the brain age estimation task. Magnetic Resonance Imaging. Full Text: https://www.ncbi.nlm.nih.gov/pubmed/31247249 Abstract Deep learning has shown remarkable improvements in the analysis of medical images without the need for...

An overview of the proposed method. Three geometric features (iH, SD, H) are used for training the spherical U-Net to predict 49 cortical parcellation labels. For each geometric property, intermediate deformation fields draw a total of 11+11 respective samples after boundary and geometric alignment for data augmentation. The cortical parcellation is then performed using the original geometric features of testing subjects.

Cortical Surface Parcellation using Spherical Convolutional Neural Networks

Jun. 6, 2019—*Prasanna Parvathaneni, *Shunxing Bao, Vishwesh Nath, David H. Zald, Neil D. Woodward, Daniel O. Claassen, Carissa Cascio, Yuankai Huo, Bennett A. Landman, Ilwoo Lyu, “Cortical Surface Parcellation using Spherical Convolutional Neural Networks”, MICCAI 2019, LNCS11766, 501-509. Abstract We present cortical surface parcellation using spherical deep convolutional neural networks. Traditional multi-atlas cortical surface parcellation requires inter-subject surface registration using geometric features with high processing...

Improved gray matter surface based spatial statistics in neuroimaging studies

May. 21, 2019—Prasanna Parvathaneni; Ilwoo Lyu; Yuankai Huo; Baxter P. Rogers; Kurt G. Schilling; Vishwesh Nath; Justin A Blaber; Allison E Hainline; Adam W Anderson; Neil D. Woodward; Bennett A Landman. “Improved gray matter surface based spatial statistics in neuroimaging studies.” Magnetic Resonance Imaging, 61, 285-295, 2019. Full text Abstract Neuroimaging often involves acquiring high-resolution anatomical images along with...

« Previous Page