Keywords:
Artificial Intelligence, eHealth, Vascular, Ultrasound, Diagnostic procedure, Embolism / Thrombosis
Authors:
J. Oppenheimer, R. Mandegaran, B. Kainz, M. P. Heinrich, F. Noor, S. Mischkewitz, A. Ruttloff, P. Klein-Weigel
DOI:
10.26044/ecr2022/C-10357
Results
40 patients were included in the study, of which 13 patients received scanning of both legs. 47 complete 2-point DVT-scans from 37 patients, including 43 complete 3-point DVT-scans from 33 patients were collected and uploaded to the platform. 4 patients showed positive results for DVT in the gold standard, on-site diagnostic scan (7.1%) of which full scans were acquired for 3 patients (see figure 8).
For 3-point compression scans, both raters showed a 100% sensitivity for DVT-diagnosis, with specificities of 62.50% and 65.00% respectively. Negative predictive value (NPV) was 100% for both reviewers and positive predictive value (PPV) 16.67% and 17.65%. 93.02% of scans were of diagnostic quality (ACEP ≥ 3) for both raters, average scoring was 3.86 (±0.25) and 3.84 (±0.12).
Subgroup analysis for 2-point compressions showed an increase of specificity to 63.64% and 68.18% with Sensitivity and NPV remaining at 100%, showing no DVT missed when excluding the thigh compression. PPV 15.79% and 17.65%. 100% of ACEP image quality scores were of diagnostic quality for both reviewers. Average scores increased to 4.40 (±0.12) and 3.87 (±0.06). Full ACEP scoring results are shown below in figure 9 - table 2. Confusion matrices are shown in figures 10 - 13.