Our team is funded to perform multidisciplinary research focusing on developing deep learning technology to develop the novel diagnosis methods and implants for spine disorders.

Teng Zhang, Ashish Diwan, Nan Meng, Morgan Cheng, Jason Cheung

We have developed an AI-powered mobile app with an auto-detection and spine pathology prediction AI engine. Patients can have real-time feedback on their spine conditions by using this app. It significantly facilitates spine surgeons with fast clinical management and out of hospital consultation. Mobile applications (AlignProCARE) are available at App Store and Google Play, free for downloading.

Wukong: Light-based Disease Detection

Teng Zhang, Nan Meng, Morgan Cheng, Jason Cheung, Kenneth Wong

Conventional body appearance analysis lack of quantitative appearance measurements and the conventional X-ray diagnosis for spine deformities involves in radiation exposure. We use depth sensing and AI technologies for accurate body geometry detections and spine curvature generations at our Body and Motion Analysis Laboratory, which is radiation free, portable, and fast with consistent results. 


Spine Quantitative AI Analysis 

Xihe Kuang, Teng Zhang, Jason Cheung

We have developed an AI-driven pipeline for the quantitative analysis of the spine. We have established an unsupervised deep learning framework called MRI-SegFlow for the segmentation of multiple spinal tissues in spinal MRI, which achieved comparable performance with the state-of-art supervised methods and without relying on any manually masked ground truth. Based on the segmentation result, the quantitative assessment of multiple spinal tissues was conducted. Our method has great significance in biomechanical simulation, 3D printing, and tissue engineering.

progression prediction_edited.jpg

Diseases Classifications and Progress Predictions

Xihe Kuang, Kenneth Chu, Teng Zhang, Jason Cheung

Our team has established a large dataset with follow-up spine MRIs and clinical labels. And based on the dataset, several deep learning based methods have been developed for automated disease detection and pathology progression prediction, which can significantly improve clinical efficiency and help radiologists to produce more consistent and objective diagnoses.


Orthopaedic Biomechanics and Implant Design

Yongqiang Jin, Teng Zhang, Jason Cheung, Tak Man Wong, KY Sze

We have demonstrated that trabecular microarchitecture (TraMicroArcht) can vary the apparent modulus (E), yield strength of trabecular bone accurately. Micro-finite element analysis can quantify the effects of TraMicroArcht on mechanical properties. MicroCT images can be converted directly into digital models via our in-house developed new system and simulate virtual loading effects. Our new system can also have significant application in fast implant design and optimization.


Medical Image Registration

Moxin Zhao, Nan Meng, Teng Zhang, Jason Cheung

Our self-developed medical image registration pipeline (entitled CA3D-MIR) overcomes the drawbacks of traditional methods, for example, sensitivity to intensity variations and time consumption. It creatively utilizes the shape model of the object scanned with different modalities. To the best of our knowledge, we are the pioneer to develop a fast and accurate algorithm for micro-CT and CT image registration. Experimental results show that CA3D-MIR outperforms the traditional registration methods in both speed and accuracy.


Clinical Database

Jason Cheung, Moxin Zhao, Noel Liang, Teng Zhang

We has established a large dataset with comprehensive medical data, including image (MRI, CT, Xray), clinical information, and follow-up, and the relevant data registry system and UI interface are under developing. It will support multi-domain big data analysis tasks in spine clinic, and the development of AI based automated diagnosis methods.