THE CADIMPACT STUDY: COMPUTER AIDED DETECTION AS SECOND READER IN A PROSPECTIVE MULTICENTER SETTING

PURPOSE The main study objective was to prospectively compare per-polyp sensitivity of non-assisted CT Colonography (CTC) with that of CTC using Computer-Aided Detection (CAD) as second reader, for lesions ≥ 6 mm. METHOD AND MATERIALS Fecal tagging regimen was adopted in all cases. All CTC studies were performed using a low dose protocol. Exams were read by an experienced radiologist unassisted (test A) and then with the aid of a commercial CAD system (test B) (CAD-COLON, im3D SpA, Turin, Italy). The reference standard was unblinded conventional colonoscopy (CC) and histology. Per-polyp sensitivity was computed for polyps of different size categories (≥6mm, ≥10mm and 6-9 mm). Sensitivity of unassisted reading was compared to that of CAD aided reading. Differences between tests were assessed by McNemar test. False positive (FP) findings and mean reading time were also recorded. RESULTS A total of 401 individuals were enrolled into the study in 3 different centers. Twenty-six were excluded for: study drop-out (n=16), CC not performed (n=5) and protocol violation (n=5). The remaining 374 individuals (200 males, mean age 58 yrs; range 34-87 yrs) included 98 patients with a total of 120 benign polyps (63 between 6-9mm; 57 ≥10mm). Sensitivity for polyps ≥6mm was 74.2% (89/120) and 80% (96/120) for non-assisted and CAD assisted readings respectively (P=0.0078). Sensitivity differences were more pronounced in the 6-9 mm class (59% for test A vs 68% for test B; P=0.02) while a non-significant increase was observed for lesions ≥10mm (93% for test A vs. 94% for test B; P>0.05). Overall there were 14 FPs ≥6 mm: 6 with test A, 1 with test B and 7 with both. Mean time of the unassisted read was 5’ 48’’ (SD 2’40’’); mean additional time for CAD-prompt evaluation was 1’ 50’’ (SD 1’32’’). CONCLUSION Using CAD as a second reader increases per-polyp sensitivity with respect to non-assisted reading. Differences between the 2 tests are more pronounced for intermediate size polyps. FP rate is not affected by the use of CAD. Reading time is shorter than what reported in landmark clinical trials. CLINICAL RELEVANCE/APPLICATION CAD used in second-read mode significantly improves performances of CTC