The morphology of glands from histopathological images has been used routinely by pathologists to assess the degree of malignancy of adenocarcinomas. Accurate segmentation of glands is often a crucial step to obtain reliable morphological statistics. To tackle this challenging problem, we proposed a deep contour-aware network (DCAN) by exploring multi-level feature representations with fully convolutional neural networks. Our method can not only output accurate segmentation probability maps, but also depict the clear contours simultaneously, which further boosts the object segmentation performance. Our method (CUMedVision team) won the winner out of 13 teams in the 2015 MICCAI Gland Segmentation Challenge.

[related link]: http://www2.warwick.ac.uk/fac/sci/dcs/research/combi/research/bic/glascontest/results/

[Representative Articles]:
•H Chen, X Qi, L Yu, PA Heng. DCAN: deep contour-aware networks for accurate gland segmentation. Conference on Computer Vision and Pattern Recognition.
•H Chen, Q Dou, X Wang, J Qin, PA Heng. Mitosis detection in breast cancer histology images via deep cascaded networks. Thirtieth AAAI Conference on Artificial Intelligence.

The localization and segmentation of intervertebral disc (IVD) is a prior step for quantification diagnosis. The automatic computerized methods are highly demanded to alleviate the workload as well as improve the efficiency and robustness.  We proposed a 3D convolutional network by harnessing the deep and hierarchical feature representations for segmentation related tasks, where each voxel was regarded as a classifier. We evaluated our method on the 3D T2 MRI data of MICCAI 2015 Challenge on Automatic Intervertebral Disc Localization and Segmentation. Our results achieved the first place on the automatic IVD localization challenge.

[related link]: http://ijoint.istb.unibe.ch/challenge/index.html

[Representative Articles]:
•Q Dou, H Chen, L Yu, L Zhao, J Qin, D Wang, VCT Mok, L Shi, PA Heng. Automatic detection of cerebral microbleeds from MR images via 3D convolutional neural networks. IEEE Transactions on Mdical Imaging 35(5), 1182-1195.

In this project, we aim to develop a planning, training and intraoperative magnetic resonance (MR) images and real-time TRUS images for targeted biopsy guidance so as to improve the positioning accuracy. The research and development challenges include 3D US-MR image regiastration for compensating the large and inhomogeneous prostate deformation due to the factors of TRUS probe zones, interactive navigation and visualization, and real-time intraoperative guidance requirement. We proposed a non-rigid registration method based on the prior knowledge of the patient-specific biomechnical deformation, and further implemented our registration method using GPU-based accelaration approach to ensure the real-time advantage of our system for intraoperative guidance.

[Representative Articles]:
•Wang Y, Cheng JZ, Ni D, et al. Towards Personalized Statistical Deformable Model and Hybrid Point Matching for Robust MR-TRUS Registration. Medical imaging, IEEE Transactions on Medical Imaging.

In this project, we proposed a virtual reality system for training vascular and interventional radiology procedures. Our system provides a cost-effective approach to providing standardized clinical education, training and accelerated learning for various percutaneous interventions on the treatment of tumors, and blood vessel diseases etc. The proposed system will benefit trainees with articulated learning experiences allowing practice with no harm to patients.

[Representative Articles]:
•J Guo, S Li, YP Chui, et al. Mesh quality oriented 3D geometric vascular modeling based on parallel transport frame. Computers in biology and medicine. 43(7), 2013, 879-888.
•S Li, J Guo, Q Wang, et al. A A catheterization-training simulator based on a fast multigrid solver. Computer graphics and applications, IEEE 32(6), 2012, 56-70.
•S Li, J Qin, J Guo, et al. A novel FEM-based numerical solver for interactive catheter simulation in virtual catheterization. International journal of biomedical imaging, 2011.

We have built a prototype virtual reality ultrasound guided organ biopsy system to facilitate the training of radiologists and physicians in ultrasound guided interventional procedures. The research issues addressed include a 3D anatomical model reconstruction, data fusion of multiple modalities data, realistic visualization and interactive navigation, and multi-sensory feedbacks. The proposed system can provide trainees with a structured learning experience, permitting practice with no danger to patients.

[Representative Articles]:
•WY Chan, PA Heng. Visualization of needle access pathway and a five-DoF evaluation. IEEE Journal of biomedical and health informatics, 2014.
•WY Chan, J Qin, YP Chui, et al. A serious game for learning ultrasound-guided needle placement skills. IEEE Transactions on information technology in biomedicine, 16(6), 2012, 1032-1042.
•D Ni, WY Chan, J Qin, et al. An ultrasound-guided organ biopsy simulation with 6DOF haptic feedback. 11th international conference on medical imaging computing and computer assisted intervention (MICCAI 2008), 551-559.
•D Ni, Q Yu, X Yang, et al. Volumetric ultrasound panorama based on 3D SIFT. 11th international conference on medical imaging computing and computer assisted intervention (MICCAI 2008), 52-60.

Training the multitude of novice medical officers and interns to acquire the skill of endoscopic surgery and / or obstetric ultrasound examination (diagnostic) is a major task in medical education. Particularly, hand-on experiences on different cases may be difficult to be arranged. Virtual reality (VR) based simulation systems provide a very elegant solution to the problem, because we can provide virtual models of different anatomic structures to simulate different procedures in realism within the virtual environment.

[Representative Articles]:
•PA Heng, CY Cheng, TT Wong, et al. Virtual reality technique: application to anatomical visualization and orthopaedics training. Clinical orthopaedics and related research, 442, 2006, 13-20.
•W Wu, PA Heng. An improved scheme of interactive finite element model for 3D soft tissue cutting and deformation. The visual computer, 21(8), 2005, 707-716.
•PA Heng, CY Cheng, TT Wong, et al. A virtual reality training system for knee arthroscopic surgery. IEEE Transactions on information technology in biomedicine, 8(2), 2004.

The Chinese Visible Human Project research team from The Third Military Medical University has successfully collected the first Chinese Visible Human data set in October 2002. Our research centre has been invited to establish long-term research collaboration in developing advanced visualization and virtual reality technologies for the Chinese Visible Human Project. We have processed and compressed the visible human data so that it becomes possible to be displayed interactively on a PC with 1:1 accuracy as to the original data. As a result, the delicated anatomic structure, which is originally difficult to be revealed on an ordinary PC, could be displayed well now. With this advance, we can observe real-time volume data information from any orientation on a low-end platform.

[Representative Articles]:
•Q Meng, YP Chui, J Qin, et al. CvhSlicer: an interactive cross-sectional anatomy navigation system based on high-resolution Chinese visible human data.Medicine meets virtual reality 18: NextMed 163, 2011, 354.
•PA Heng, SX Zhang, YM Xie, et al. Photorealistic virtual anatomy based on Chinese visible human data. Clinical anatomy, 19(3), 2006, 232-239.
•SX Zhang, PA Heng, ZJ Liu, et al. The Chinese Visible Human (CVH) datasets incorporate technical and imaging advances on earlier digital humans. Journal of anatomy, 204(3), 2004, 165-173.
•SX Zhang, PA Heng, ZJ Liu, et al. Creation of the Chinese visible human data set. The anatomical record part B: the new anatomist 275(1), 190-195.

This project aims at developing an intelligent virtual environment for Chinese acupuncture learning and training using state-of-the-art virtual reality technology. It is the first step towards developing a comprehensive virtual human model for studying Chinese medicine.

[Representative Articles]:
•CFD Lam, KS Leung, PA Heng, et al. Chinese Acupuncture Expert System (CAES) - A useful tool to practice and learn medical acupuncture. Journal of medical systems, 36(3), 2012, 1883-1890.
•PA Heng, SX Zhang, Y Xie, et al. Deploying Chinese visible human data on anatomical exploration: from western medicine to Chinese acupuncture. Complex medical engineering, 2007, 351-360.
•SX Zhang, PA Heng, ZJ Liu, et al. Virtual acupuncture human based on Chinese visible human dataset. Journal of anatomy, 204(3), 2004, 165-173.

In this project we develop an intelligent virtual environment with the ability of providing knowledge-based image segmentation, multi-modal cardiac image fusion, image data mining, and dynamic cardiac feature visualization with multi-sensory feedback.

[Representative Articles]:
•H Wu, PA Heng, TT Wong. Cardiac motion recovery using an incompressible B-solid model. Medical engineering and physics. 35(7), 2013, 958-968.
•L Wang, J Qin, TT Wong, et al. Application of L1-norm regularization to epicardial potential reconstruction based on gradient projection. Physics in medicine and biology, 56(19), 2011, 6291-6310.
•G Luo, PA Heng. LV shape and motion: B-spline based deformable model and sequential motion decomposition. IEEE Transactions on information technology in biomedicine, 9(3), 2005, 430-446.
•M Solaiyappan, T Poston, PA Heng, et al. Interactive visualization for speedy non-invasive cardiac assessment. IEEE Computer, 29(1), 1996, 55-62.