Coronary artery lumen segmentation by deep learning neural networks
Medical images segmentation manually is always a challenging problem, especially for segmentation of tiny lumen in coronary artery. Detection and delineation are important but tedious steps to grade stenosis. Auto segmentation using deep learning has recently been successful applied to many applicat...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/74763 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Medical images segmentation manually is always a challenging problem, especially for segmentation of tiny lumen in coronary artery. Detection and delineation are important but tedious steps to grade stenosis. Auto segmentation using deep learning has recently been successful applied to many applications such as medical images. This report studies coronary artery stenosis detection and segmentation using 3D U-net convolutional neural networks, tests its performance and compares the testing results with multiple datasets on two settings. The CTCA images are adapted into small patches and directly into the learning machine. Our experiment involves additional effort to select and test different sets of data, so as to reduce problem of overfitting. Compared with data using traditional normalization, the performance has shown an improvement on dataset while we fixed a specific number rather than using mean and standard deviation. The performance of testing was evaluated by loss function with ground truth masks and prediction images. The results of dice coefficients also showed the proposed approach is more successful than some other methods. This work [22] has been accepted for presentation in IEEE Engineering in Medicine and Biology Society (EMBC) 2018 conference. |
|---|