AUTHOR=Duan Yunyun , Shan Wei , Liu Liying , Wang Qun , Wu Zhenzhou , Liu Pan , Ji Jiahao , Liu Yaou , He Kunlun , Wang Yongjun 

TITLE=Primary Categorizing and Masking Cerebral Small Vessel Disease Based on “Deep Learning System”

JOURNAL=Frontiers in Neuroinformatics

VOLUME=Volume 14 - 2020

YEAR=2020

URL=https://www.frontiersin.org/journals/neuroinformatics/articles/10.3389/fninf.2020.00017

DOI=10.3389/fninf.2020.00017

ISSN=1662-5196

ABSTRACT=Objective: 
To supply the attending doctor's diagnosis of the persisting of cerebral small vessel disease and speed up their work effectively, we developed a “deep learning system (DLS)” for cerebral small vessel disease diagnosis. The reliability and the disease area segmentation accuracy, of the proposed DLS, was also investigated.
Methods: 
A deep learning model based on the convolutional neural network was designed and trained on 1,010 DWI-b1000 images for patients diagnosed with infarction, 359 T2* images for patients diagnosed with microbleed, as well as 824 T1-weighted and T2-FLAIR images for patients diagnosed with Lacune and WMH. Accuracy, recall, and f1-score were calculated to evaluate the proposed deep learning model. Finally, we also compared the DLS prediction capability with that of 6 doctors with 3 to 18 years’ clinical experience (8  6 years).  
Results:
The results support that an appropriately-trained deep learning system can achieve a high-level accuracy, more than 0.8 in the training section over all these four classifications, validation accuracy is 0.6 in lacune, 0.75 in Wmh, 0.8 in infarction, and 0.78 in cerebral microbleeds. It is comparable to attending doctor with a few years of experience, regardless of whether the emphasis is placed on the segmentation or detection of lesions with less time-spending compared with manual analysis, about 4.4 seconds/case, which is dramatically less than doctors about 634 seconds/case. 
Conclusion:
The results of our comparison lend support to the case that an appropriately-trained deep learning system can be trusted to the same extent as one would trust an attending doctor with a few years of experience, regardless of whether the emphasis is placed on the segmentation or detection of lesions.