Artificial Intelligence-based Large Vessel Occlusion Diagnosis Algorithm Verificatio

Not Applicable

Completed

• Conditions: Diseases of the circulatory system

• Registration Number: KCT0009457
• Lead Sponsor: JLK

Brief Summary

This clinical trial was conducted to evaluate the clinical effectiveness of JBS-LVO, an artificial intelligence-based CT vascular imaging large vessel occlusion diagnostic medical device. The primary efficacy endpoint of the clinical trial is the sensitivity and specificity calculated by comparing the large vessel occlusion interpretation results of the software (JBS-LVO) with the reference standard, and the secondary efficacy variables are the location of large vessel occlusion (carotid artery [ICA]), proximal middle cerebral artery [M1], and distal middle cerebral artery [M2]) to calculate the sensitivity and specificity for predicting large vessel occlusion of JBS-LVO. This study retrospectively extracted the patient's brain CT vascular images that had already been collected prospectively and applied them to a clinical trial. Since no side effects or adverse reactions occurred in the patient, it was evaluated that there was no safety problem with the test performance. Therefore, no safety evaluation was conducted. Patients with large vessel occlusion arrive relatively faster, are older, often have atrial fibrillation, and have higher severity than non-occurring cases. The CT vascular image information used for analysis by medical devices for clinical trials does not directly reflect the relevant characteristics, so it does not affect the validity of this test. CT vascular images of a total of 603 people (205 from Daejeon Eulji Hospital and 398 from Chonnam National University Hospital) were collected according to the test subject selection and exclusion criteria, and a total of 8 people were dropped according to the dropout criteria, and finally, 595 brain CT scans were performed. Among 595 patients, 275 (46.2%) were diagnosed as having large vessel occlusion based on expert consensus interpretation. As a result of the clinical trial, no significant differences were found in the performance analyzed by demographic characteristics (age, gender) of positive and negative subjects, presence or absence of atrial fibrillation, time from symptom onset to imaging, whether reperfusion was performed, and stroke severity and both groups showed equally high performance. Additionally, the sensitivity and specificity for diagnosing large vessel occlusion of JBS-LVO, which considered the consensus result of three cerebrovascular disease specialists as the correct answer, were 0.858 (95% CI: 0.811-0.897) and 0.969 (95% CI: 0.943-0.985), respectively. It exceeded the primary efficacy evaluation criteria: sensitivity for diagnosing large vessel occlusion of 0.7489 and specificity for diagnosing large vessel occlusion of 0.9155. In the secondary effectiveness evaluation, high diagnostic sensitivity and specificity were achieved with sensitivity of 0.862 and specificity of 0.960 for diagnosing intracranial carotid artery occlusion, sensitivity of 0.954 and specificity of 0.960 for diagnosing proximal middle cerebral artery occlusion, and sensitivity of 0.880 and specificity of 0.960 for diagnosing distal middle cerebral artery occlusion.

Detailed Description

Not available

Study Timeline

• Study Completion: 2024-01-31
• Last Update Posted: 27 May 2024

Recruitment & Eligibility

• Status: Completed
• Sex: All
• Target Recruitment: 603

Inclusion Criteria

1) Age: 18 years or older
2) Period: January 1, 2017 to December 31, 2022
3) Data on patients who underwent brain CTA due to suspected LVO among patients who visited the emergency rooms of Chonnam National University Hospital and Daejeon Eulji University Hospital.
4) Data of patients with corrupt CTA
5) Guest data whose ownership is recognized for two reasons
6) Data from patients with CTA plus section thickness of 0.5 to 2.0 mm

Exclusion Criteria

1) Patients with symptomatic LVO and asymptomatic LVO
2) Patients who took a brain CTA image, but an error occurred in the DICOM storage format and the image could not be read, making analysis impossible
3) A patient who had a brain CTA image taken, but analysis was not possible because the area within the brain could not be diagnosed due to strong metal artifacts.
4) Patients with insufficient analysis due to missing slices in brain CTA images
5) Those who have had a brain CTA image taken, but the subject's age, gender, date of shooting, and CT image sequence information for identification are unclear.
6) Patients who do not have enough vascular information to determine LVO because the contrast agent is insufficient or the contrast agent does not sufficiently fill the blood vessels.
7) Patients used for artificial intelligence model learning and internal verification

Study & Design

• Study Type: Observational Study
• Study Design: Not specified