The Japan NBI Expert Team (JNET) classification enables optical prediction of colorectal polyp histology. Evidence from Western settings remains limited. Lesion size may also impact diagnostic accuracy. We evaluated the diagnostic performance of JNET in a Untied Kingdom (UK) cohort and explore whether lesion size influences accuracy, using histopathology as the gold standard.

A retrospective observational study was performed across three UK endoscopy units, Northwick Park (St Marks), Central Middlesex and Ealing Hospitals London during the period of January 2020–August 2024. Endoscopy reports were screened for the keyword “JNET.” Procedures were included if a colorectal polyp had an assigned JNET classification. Exclusion criteria were absence of histology and inflammatory bowel disease.

For each lesion, data collected included location, size, JNET, morphology (Paris or Laterally spreading tumour (LST)) and histopathology. Accuracy data was calculated for each JNET subtype and compared across lesion size. We used the Cochran–Armitage test to assess trends across size categories.

A total of 697 lesions were analysed: 270 in the right colon, 169 in the left colon, and 258 in the rectosigmoid/rectum. Histology comprised 25 hyperplastic, 20 sessile serrated, 330 low-grade dysplasia, 234 high-grade dysplasia and 88 T1 cancers (22 T1a, and 66 T1b). Morphology included: 6% Ip, 11% Isp, 27% Is, 9% IIa, 11% IIc, 29% LST-G, and 7% LST-NG. Lesion size distribution was 149 ≤10 mm, 160 = 11–20 mm, 217 = 21–40 mm, and 171 > 40 mm.

The overall accuracy was 95%, 70%, 67% and 90% for JNET 1, 2A, 2B and 3 respectively. Accuracy significantly improved with increasing lesion size for JNET 1 (85% to 99%, p < 0.001) but conversely declined significantly for JNET 2A (74% to 60%, p = 0.004), 2B (89% to 50%, p < 0.001) and JNET 3 (97% to 85%, p = 0.001). 

The JNET classification demonstrated reasonable, but slightly lower overall diagnostic yield compared with Asian cohorts. Lesion size significantly influenced JNET performance. Larger lesions improved accuracy for JNET 1 but reduced it for JNET 2A, 2B, and 3. These findings emphasise that lesion size is an important consideration when using JNET in practice. Standardised training may optimise diagnostic reliability and support implementation of optical diagnosis pathways in colorectal polyp assessment in the UK.