Research Directions

Intelligent Auxiliary Diagnostic Platform for Cerebral Hemorrhage
UpdateTime:2024-03-18 01:50

The research team has established the China Intracerebral Hemorrhage Imaging Database (CICHID), focusing primarily on hypertensive cerebral hemorrhage and encompassing various bleeding types and multimodal data. The database has collected complete and continuous medical records during the hospitalization of 10,000 patients from over 30 tertiary hospitals, including nearly 50,000 head scan data, as well as medical records, auxiliary examination reports, etc., making it one of the largest databases in the field of radiomics research in China. The institute's team has also established China's first "Hypertensive Cerebral Hemorrhage Surgical Treatment Alliance" focused on hypertensive cerebral hemorrhage, currently comprising 116 hospitals. A preliminary intelligent assistance model for the diagnosis and treatment of hypertensive cerebral hemorrhage has been completed and is ready for further clinical validation within the alliance hospitals. Based on the above data, related work in the field of cerebral hemorrhage is being carried out: 


A series of research works has been conducted in the application of artificial intelligence in the field of cerebral parenchymal hemorrhage, including assisting in precise assessment, software platform development, and neurosurgical navigation based on mobile computing platforms: 


In terms of assisting precise assessment and diagnosis, based on more than 2,000 head CT scans, an automatic hemorrhage segmentation algorithm has been developed using deep learning networks to achieve regional quantitative diagnosis of cerebral hemorrhage. The consistency with manual segmentation gold standard reached 0.98, far superior to the traditional Kothari formula. Creatively, regional cerebral hemorrhage volume has been used to predict patient prognosis, finding that the AUC value for predicting death reached 0.71, a significant improvement over the traditional method's 0.62. In predicting hematoma expansion, deep learning networks and latent space algorithms have been used to construct an image-to-image (end-to-end) prediction model. It can not only predict whether the hematoma will expand but also the volume of expansion and even the morphology of the hematoma after expansion. The current test set DICE value reached 0.749.


The institute's team has already developed an automated clinical assistance software platform for cerebral hemorrhage. Based on the aforementioned research results, an intelligent cloud-based automatic cerebral hemorrhage assessment system has been built. It can conveniently provide doctors with data such as regional bleeding volume, midline shift, and cerebrospinal fluid/brain parenchyma ratio, offering a basis for precise clinical treatment.

2893dbceb5da4f37932953adc1f02eff.jpg


官方公众号
Access statistics:You are the visitor
Copyright Reserved © 2022 national health data institute,shenzhen | 粤ICP备16049670号 technical support:yichuang Technology Co., Ltd