Artificial Intelligence-Assisted Sac Diameter Assessment for Complex Endovascular Aortic Repair

No prior work had resolved the efficiency challenges associated with manual aortic diameter measurements in complex endovascular cases. Clinicians currently rely on time-intensive manual assessments for monitoring patients after surgical interventions. This gap motivated the development of automated tools to assist in routine morphology evaluation. Prior research has shown that deep learning methods can potentially enhance diagnostic accuracy in various medical imaging fields. However, the performance of these tools in the context of complex aortic anatomies remains under investigation. That uncertainty drove the need for rigorous verification of new software against established clinical standards. Existing manual workflows often require significant expertise and time from vascular specialists. This study addresses these limitations by testing a specific automated platform in a real-world clinical environment.

The aim of this study was to evaluate the accuracy of an automated deep learning-based method for assessing aortic aneurysm morphology. Researchers sought to determine if this software could serve as a viable aide for clinicians managing complex aortic aneurysms. The study specifically focused on patients who underwent fenestrated endovascular repair. This investigation addressed the need for more efficient methods to process the high volume of preoperative and postoperative imaging data. The authors hypothesized that automated tools could reduce the time-intensive nature of manual measurements. They aimed to verify the performance of the software against standard clinician-led techniques. By comparing automated results with manual multiplanar reconstruction and curved planar reformatting, the team assessed the reliability of the technology. This work was motivated by the desire to improve routine clinical workflows in specialized aortic centers.

The review approach involved a retrospective analysis of imaging data from fifty patients treated with fenestrated endovascular repair. Investigators extracted preoperative and postoperative computed tomography angiography scans from a single institutional picture archiving and communication system. All images underwent automated morphology assessment using the specified deep learning software. Clinicians verified the identification of the aortic outer wall by reviewing AI-generated overlays for every patient. Two experts performed manual measurements using multiplanar reconstruction and curved planar reformatting techniques. The team compared these manual results against the automated outputs for studies where the software correctly identified the vessel wall. Researchers identified outliers by scrolling through axial segmentation cuts of the entire aorta. This systematic evaluation ensured a robust comparison between automated and human-led diagnostic processes.

The strongest finding indicates that the software accurately identified the aortic outer wall in 89% of the reviewed scans. Automated diameter measurements were comparable to clinician results, with median absolute differences of 2.4 mm for preoperative and 1.6 mm for postoperative images. No significant differences appeared between manual multiplanar reconstruction and curved planar reformatting measurements, which ranged from 0.5 mm to 0.9 mm. The software maintained accuracy despite the presence of complex aortic anatomies in the patient cohort. Furthermore, the tool performed effectively on postoperative scans even when metal artifacts from stent grafts and embolization materials were present. Clinicians easily identified imprecise automated overlays by inspecting the axial segmentation cuts. These results confirm that the automated method provides a high level of precision for routine morphology assessment. The data support the integration of this technology into standard vascular imaging workflows.

The authors propose that their automated software offers a reliable alternative to traditional manual assessment methods for aortic aneurysms. Their findings suggest that the technology maintains high accuracy even when complex anatomical features are present. The researchers indicate that the tool performs effectively despite the presence of metal artifacts from surgical implants. This synthesis implies that integrating such systems could significantly improve the efficiency of routine clinical practice. The data demonstrate that human oversight remains a necessary component for identifying occasional outliers in automated segmentation. The authors conclude that the software provides a viable solution for managing the high volume of imaging data in vascular centers. This work highlights the potential for artificial intelligence to support clinicians without compromising measurement precision. The study confirms that automated morphology assessment is a feasible approach for patients undergoing complex endovascular aortic repair.