RESUMEN
BACKGROUND: Blue laser imaging (BLI) can provide useful information on colorectal laterally spreading tumors (LSTs) by visualizing the surface and vessel patterns in detail. The present research aimed to evaluate the diagnostic performance of BLI-combined JNET (Japan NBI Expert Team) classification for identifying LSTs. METHODS: This retrospective, multicenter study included 172 LSTs consisted of 6 hyperplastic polyps/sessile serrated polyps, 94 low-grade dysplasias (LGD), 60 high-grade dysplasias (HGD), 6 superficial submucosal invasive (m-SMs) carcinomas, and 4 deep submucosal invasive carcinomas. The relationship between the JNET classification and the histologic findings of these lesions were then analyzed. RESULTS: For all LSTs, non-experts and experts had a 79.7% and 90.7% accuracy for Type 2A (P = 0.004), a sensitivity of 94.7% and 96.8% (P = 0.718), and a specificity of 61.5% and 83.3% (P = 0.002) for prediction of LGD, respectively. The results also demonstrated 80.8% and 91.3% accuracy for Type 2B (P = 0.005), a sensitivity of 65.2% and 83.3% (P = 0.017), and a specificity of 90.6% and 96.2% (P = 0.097) for predicting HGD or m-SMs. For LST-granular (LST-G) lesions, Type 2A in experts had higher specificity (65.6% vs. 83.6%, P = 0.022) and accuracy (81.8% vs. 91.2%, P = 0.022). Type 2B in experts only had higher accuracy (82.5% vs. 92.0%, P = 0.019). However, no significant differences were noted for any comparisons between non-experts and experts for LST-non-granular (LST-NG) lesions. CONCLUSIONS: BLI combined with JNET classification was an effective method for the precise prediction of pathological diagnosis in patients with LSTs. Diagnostic performance of JNET classification by experts was better than that by non-experts for all examined LST or LST-G lesions when delineating between Type 2A and 2B, but there was no difference for the identification of LST-NG lesions by these two groups.
