|
AUTHOR(S): Ioannis Theocharakis, Eleftherios Kontopodis, Dimitris Arampatzis, Emmanouil Athanasiadis, Ιlias Theodorakopoulos, Dimitris Glotsos, Pantelis Asvestas, Anastasios Raptis, Christos Manopoulos, Konstantinos Moulakakis, John Kakisis, Ioannis Kalatzis, Spiros Kostopoulos
|
TITLE A Comparative Study of Machine Learning Systems in Abdominal Aortic Segmentation |
 PDF |
ABSTRACT Abdominal aortic segmentation is a critical procedure for the diagnosis and monitoring of pathologies such as aneurysms and strictures. Accurate and efficient imaging of this anatomical region requires the use of advanced computational imaging technologies. In this work, we investigate the application of modern machine learning (ML) tools for automatic segmentation of the abdominal aorta from computed tomography (CT) data without retraining of available ML networks. We compared two ML image segmentation systems against annotations performed by skilled personnel. Nvidia's Medical Open Network for Artificial Intelligence (MONAI) and TotalSegmentator are two state-of-the-art tools that in recent years have become almost universally prevalent in the field of Medical Image Analysis as solutions that combine deep learning algorithms and ML to improve accuracy and efficiency in medical image segmentation. We used 19 CT datasets acquired in the framework of the SAFE-AORTA action, along with annotations of the lumen and intraluminal thrombus. We used the Dice Similarity coefficient (DSC) and the average Hausdorff Distance (HD) to quantitatively assess the performance of the two systems in segmenting the aorta region (lumen and intraluminal thrombus). Preliminary results indicate minimal differences between the two ML tools and their adaptation to expert segmentation remains within acceptable limits (DSC~0.83, HD~3 mm with 95%HD~10 mm). In a next phase, we will explore the possibility of improving the models through dataset enrichment and retraining, with the aim of increasing the accuracy of abdominal aortic segmentation. |
KEYWORDS X-ray CT; aortic aneurysm; segmentation; machine learning; neural networks |
|
Cite this paper Ioannis Theocharakis, Eleftherios Kontopodis, Dimitris Arampatzis, Emmanouil Athanasiadis, Ιlias Theodorakopoulos, Dimitris Glotsos, Pantelis Asvestas, Anastasios Raptis, Christos Manopoulos, Konstantinos Moulakakis, John Kakisis, Ioannis Kalatzis, Spiros Kostopoulos. (2025) A Comparative Study of Machine Learning Systems in Abdominal Aortic Segmentation. International Journal of Biology and Biomedicine, 10, 14-21 |
|
