Assessing the Accuracy of an Artificial Intelligence-Based Segmentation Algorithm for the Thoracic Aorta in Computed...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiovascular Imaging

Background:

• Accurate measurement of the thoracic aorta (TA) is crucial for diagnosing aortic pathology.
• Manual segmentation and diameter measurement of the TA on CT scans can be time-consuming and subject to inter-observer variability.

Purpose of the Study:

• To evaluate the accuracy and reliability of a novel artificial intelligence (AI)-based algorithm for automated TA segmentation and diameter measurement using CT.
• To compare the AI algorithm's measurements against those of experienced radiologists.

Main Methods:

• Retrospective analysis of 122 dual-source CT scans, with 93 including contrast.
• Automated segmentation and diameter measurement of the TA by the AI prototype according to American Heart Association guidelines.
• Reference standard established by two independent, blinded radiologists; inter-reader agreement assessed using intra-class correlation (ICC).

Main Results:

• The AI algorithm achieved 99.2% assessability for TA parameters, with minor failures in 9 patients.
• No significant difference was found between AI and radiologist measurements (p > 0.05), establishing measurement equivalence.
• Excellent inter-reader agreement was observed between the AI and radiologists (ICC ≥ 0.961) and between radiologists (ICC ≥ 0.879).

Conclusions:

• The AI-based algorithm demonstrates high accuracy and reliability in measuring TA diameters, comparable to expert radiologists.
• The algorithm's performance is independent of contrast use or underlying pathology, highlighting its robustness.
• This AI tool shows substantial potential for efficient and accurate clinical assessment of aortic diseases.