Congratulations to Assistant Professor Liu Jun and Research Assistant Wu Qian for winning the 2nd place (amongst a total of 339 international groups) in the MICCAI 2023 CL-Detection Grand Challenge. The research group applied their Human Pose Estimation technique to handle the medical landmark detection problem and got the first place on two metrics and the second place in one metric.

The 26th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) 2023 was held on 8 October 2023, in Vancouver, Canada. This conference is a premier international event in the field of medical image intelligent computing, renowned for showcasing the latest research and hosting international challenges related to various medical image-assisted diagnostic scenarios. The team members, Wu Qian and Liu Jun from the SUTD-VLG group, secured the second place (2/339) in the Cephalometric Landmark Detection (CL-Detection) in Lateral X-ray Images (MICCAI CL-Detection 2023) challenge.

The localisation of cephalometric landmarks in lateral X-ray images is of significant importance in the fields of orthodontics, maxillofacial surgery and craniofacial treatment, as it aids in diagnosing dental and skeletal abnormalities in patients, devising treatment plans, assessing facial reconstruction outcomes, and predicting the growth trends of bones and teeth in adolescents. The CL-Detection Challenge aimed to design precise and reliable algorithms for the automatic positioning of head X-ray measurement points, providing reference tools for clinical practitioners.

SUTD-VLG creatively revisited the challenges and differences in medical landmark detection from the perspective of human pose estimation to propose an innovative solution based on efficient heatmap super-resolution. This solution reduced the computational load during upsampling while mitigating the challenges of heatmap localisation, leading to more precise detection. Their approach outshone competitors from prestigious institutions and organisations like Cornell University, the German Cancer Research Center (DKFZ), and Zhejiang University. Specifically, SUTD-VLG claimed the top position in the two-stage rankings for the Successful Detection Rate under 2mm (SDR2mm) metric and achieved first and third place in the Mean Radial Error (MRE).

References:
https://conferences.miccai.org/2023/en/challenges.asp
https://cl-detection2023.grand-challenge.org/