In October 2022,Cell Reports Medicine (IF:16.988) published the research results of Professor ZHENG Chuansheng’s team of the Radiology Department of Union Hospital, titled “Self-supervised learning enables 3D digital subtraction angiography reconstruction from ultra-sparse 2D projection views: A multicenter study”. The research proposed a self-supervised 3D-DSA Reconstruction (SSDR) method to ensure the reconstruction quality of 3D-DSA imaging while significantly reduce the radiation dosage received by the human body.
Subarachnoid hemorrhage is a severe subtype of stroke associated with permanent brain damage or death. Early diagnosis and treatment are the most effective means for this disease. Approximately 85% of cases are caused by intracranial aneurysms (IAs). Two-dimensional (2D) and 3D images recorded using the digital subtraction angiography (DSA) technique has long been used as the gold standard for the diagnosis of IAs, for it can be analyzed to directly visualize the bleeding points. The most classic 3D-DSA reconstruction method involves the use of the Feldkamp-Davis-Kress (FDK) algorithm. However, this standard 3D-DSA scanning and reconstruction method requires the patient to be continuously exposed to radiation. To date, effective methods that can help significantly reduce the radiation dosage (ensuring the quality of the reconstructed images) have not been reported.
Herein, the team proposed a SSDR method to reconstruct high-quality 3D-DSA images using only eight 2D projections (from different views).In 3D reconstruction, the multi-view 2D images are directly input to the trained model to reconstruct the corresponding 3D image in real time. The clinical application of this method can be expected to greatly reduce the radiation dosage received by human body during 3D-DSA imaging,and thus has great translational value.
Source: https://www.whuh.com/info/1021/12829.htm
Edited by: International Exchange Office
