Key result
Semiautomatic algorithms match or exceed TotalSegmentator accuracy for descending aorta segmentation.
Why the study?
Automatic machine learning segmentation methods can fail on poor-quality scans or corner cases, require time-consuming manual labeling, and have not yet solved practical tasks like diaphragm detection.
Semiautomatic model-based algorithms can provide comparable or superior accuracy to deep learning methods for specific CT segmentation tasks like the descending aorta, serving as a robust fallback or data-labeling tool.
Answers are grounded in this paper and cite their sources.
Semiautomatic algorithms may serve as fallback for aorta CT segmentation; leaves open broader validation and clinical integration.
Segmentation of various organs and body parts based on CT scans is an essential step in the computer-aided design of the surgeries. Although, with the emergence of the machine learning-based algorithms many segmentation tasks can now be solved automatically, there is room for semiautomatic algorithms. On the one hand, in certain circumstances – such as poor quality of the scans, presence of artifacts, and in some corner cases – even machine learning based methods fail to perform the segmentation accurately. In these cases, a robust semiautomatic algorithm may be used as a fallback mechanism. On the other hand, training datasets which typically cover a wide spectrum of cases and range of qualities for a better generalization ability of the trained model, require however time consuming manual labeling. Furthermore, a slightly different application requirement may require the re-labeling of the data. Semiautomatic algorithms can be a viable solution for speeding up the manual labeling of the training datasets. Finally, there are practical tasks that have not yet been solved by automatic segmentation methods – such as diaphragm detection – that healthcare professionals can benefit from. In this paper we introduce semiautomatic algorithms for aorta and heart segmentation and diaphragm detection. By relying on medical expert knowledge and integrating classical image processing algorithms such as Hough-transform and dynamic programming, the proposed algorithms can be reliably applied in practical applications. We evaluated the proposed methods on the SegThor dataset and compared their results with a deep neural network-based method called TotalSegmentator. Results indicate that in the case of the descending aorta segmentation the accuracy of the proposed method is comparable or superior. In the case of heart segmentation the proposed method solves a slightly different, practical task.
No takes yet. Share an insight, caveat, or question.
Révy et al. (2026) studied CT scan segmentation. Semiautomatic algorithms for aorta and heart segmentation and diaphragm detection vs. TotalSegmentator (deep neural network-based method) was evaluated on Segmentation accuracy. Semiautomatic algorithms for descending aorta segmentation achieved accuracy comparable or superior to the deep neural network-based TotalSegmentator on the SegThor dataset.
Synapse has enriched 10 related papers. Consider them for comparative context: