The 22nd International Conference On Medical Image Computing & Computer Assisted Intervention (MICCAI 2019)
This is about a MICCAI 2019 challenge, Automatic Generation of Cardiovascular Diagnostic Report. The challenge proposal is accepted on March 4, 2019. The registration entry of the challenge has opened on May 20, 2019. The challenge meeting takes place at Oct 13 15:30 , 2019 in Shenzhen, China.
This challenge targets at evaluating cutting-edge techniques and revealing exciting potentials for medical image captioning, and more particularly, the automatic generation of cardiac-cerebral vascular diagnostic report. Despite the extensive endeavors in intelligent diagnosis of cardiovascular diseases, the research on automatic report generation of the diagnosis is left unexploited. Such automatic reports are highly desirable and can be further combined with human-computer interaction to improve the diagnostic efficiency and accuracy. The key challenge lies in how to make use of the heterogeneous medical data as well as the initial/primary detection and classification results. In terms of methodology, the challenge also belongs to a more general topic of vision and language, involving complex understanding of visual contents (i.e. detection of suspected cardiovascular lesions) and linguistic descriptions (i.e. diagnosis and treatment suggestions provided by domain experts). Through this challenge, the organizers target at establishing a novel man-machine collaborative mechanism for diagnosis and treatment of cardiovascular diseases. The resulting standardized reports can further enable in-depth study of man-machine collaborative interaction, design joint analysis model using diagnosis and treatment data and expert diagnosis, as well as realizing the evolution of mechanism from fragmented knowledge to automatic generation of interpretable standardized reports.
Towards providing accurate and intelligent diagnostic of cardiovascular diseases, the challenge provides a certain amount of labeled reports with medical images of cardiovascular diseases, together with a large amount of unlabeled medical images. Both supervised and unsupervised learning methods can be deployed, and the goal is to learn an automatic report generator, which receives new medical images online and outputs the report in a standard format that can be further collaboratively revised between human and machine. To that effect, the automatic diagnostic results of cardiovascular diseases (including suspected lesion areas, disease types, etc.), together with visually similar cases are fused together to automatically generate interpretable standardized reports. Such reports include both images/graphics and texts, where the images include the detection results, and the texts are the diagnosis including both the diseases types and suggested treatments.
Main Process and Important Dates
|2019.05.13||Website Update and notification|
|2019.05.23||Participant Information Verification|
|2019.05.25||Dataset Link Open|
|2019.06.15||Team Registration (Entrance Close)|
|2019.09.10||Notidifacation & Evaluation (1st)|
|2019.09.20||Notidifacation & Evaluation (2ed)|
|2019.09.30||Notidifacation & Evaluation (3rd)|
|2019.10.05||Notidifacation, Communication, and Material Preparason for The Meeting|
Participation and RankingThere are 19 teams participating, including Peking University, University of Chinese Academy of Sciences, The Chinese University of Hong Kong, Nanyang Technological University, etc. We list the top-3 teams as below:
|1||vista||University of Rochester|
|3||Fighting||University of Glasgow|