Joshua R Astley1,2, Alberto M Biancardi1, Helen Marshall1, Laurie J Smith1, Guilhem J Collier1, Paul J Hughes1, Michael Walker1, Matthew Q Hatton2, Jim M Wild1, and Bilal A Tahir1,2
1POLARIS, University of Sheffield, Sheffield, United Kingdom, 2Oncology and Metabolism, University of Sheffield, Sheffield, United Kingdom
1POLARIS, University of Sheffield, Sheffield, United Kingdom, 2Oncology and Metabolism, University of Sheffield, Sheffield, United Kingdom
We present a generalisable deep learning model for automated lung segmentation of proton MRI acquired at different resolutions and inflation levels from healthy subjects and patients with respiratory diseases. The model generated accurate segmentations, outperforming a previous method.
Figure 3. Example coronal slices of DL and SFCM segmentations for three cases with different image resolutions and diseases compared to the expert segmentations. DSC and Avg HD values are given for each case.
Figure 4. a) Comparison of segmentation performance of DL and SFCM for all scans in the testing set and for each acquisition protocol. Means are given; the best result for each metric is in bold. b) Comparison of DL performance for each of the three acquisition protocols using DSC (left) and Average boundary Hausdorff distance (right). Significances of differences between acquisitions were assessed using a Mann–Whitney U test.