Related Experiment Video
Updated: Dec 18, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.2K
Primary Categorizing and Masking Cerebral Small Vessel Disease Based on "Deep Learning System".
Yunyun Duan1,2, Wei Shan2,3,4, Liying Liu2
1Department of Radiology, Beijing Tiantan Hospital, Capital Medical University, Beijing, China.
Frontiers in Neuroinformatics
|June 12, 2020
Summary
A new deep learning system (DLS) aids in diagnosing cerebral small vessel disease, matching doctor performance in lesion detection and segmentation. This AI tool significantly speeds up diagnosis, offering a reliable alternative for attending physicians.
Area of Science:
- Neurology
- Artificial Intelligence
- Medical Imaging
Background:
- Cerebral small vessel disease (CSVD) diagnosis requires efficient and accurate tools for attending physicians.
- Deep learning systems (DLS) show promise in medical image analysis.
Purpose of the Study:
- To develop and evaluate a DLS for predicting CSVD.
- To assess the reliability and segmentation accuracy of the DLS.
- To compare DLS performance against human experts.
Main Methods:
- A convolutional neural network-based DLS was trained on diverse neuroimaging data (DWI, T2*, T1-weighted, T2-FLAIR) from over 1000 patients.
- The DLS was evaluated using accuracy, recall, and F1-score for segmenting subcortical infarction, cerebral microbleed, lacune, and white matter hyperintensities (WMH).
- DLS performance was compared to that of six neuroradiologists with varying clinical experience.
Main Results:
- The DLS achieved high Dice accuracy across four classifications (0.598 overall training, 0.496-0.728 validation).
- DLS performance was comparable to doctors with several years of experience in lesion segmentation and detection.
- The DLS processed cases significantly faster (4.4 s/case) than manual analysis by doctors (634 s/case).
Conclusions:
- An appropriately trained DLS can be reliably used for CSVD diagnosis.
- The DLS offers comparable diagnostic performance to experienced attending physicians.
- The DLS dramatically reduces diagnosis time, enhancing workflow efficiency for physicians.
