Original title: Feature-Based Patterns Associated With Japan NBI Expert Team Type 1 Classification of Serrated Colorectal Lesions: A Post Hoc Decision Tree Analysis
Taku Sakamoto, Hiroyuki Takamaru, Daizen Hirata, Iwatate Mineo +4 · September 2026
Doctors use a special camera view to classify growths in the colon called serrated lesions. One category they look for is Japan NBI Expert Team Type 1. However, different doctors often disagree on which visual features they see when making this choice. Researchers wanted to see how doctors choose this label. They reviewed web test data where specialists and less experienced doctors examined images of growths. The team used decision trees to map which visual clues drove their choices between Type 1 and non-Type 1. They found that experts primarily based their first split on signs that argued against Type 1. In contrast, non-experts first split on seeing regular dark or white spots. The experts showed more stable visual patterns and less randomness. The non-expert patterns changed when unassessable answers were removed. These findings show pattern links rather than a proven diagnostic rule.
Imagine you and a friend are sorting mushrooms into 'safe' and 'poisonous' piles. You both have a list of features like cap color, gill spacing, and stem ring. But you each start with a different feature as the most important check. You first look at whether the mushroom has a cup-like volva; your friend first checks if the cap has white spots. Neither method is wrong per se, but you end up disagreeing on some mushrooms. In this paper, expert endoscopists and non-experts likewise prioritized different visual features when deciding whether a colon lesion is Type 1, and the experts' priority was more consistent across different cases.
Doctors who look inside the colon often see suspicious growths called polyps. Their team's distinction between two polyp types—serrated lesions—affects whether a patient needs more surveillance or treatment. This study shows that expert and non-expert endoscopists rely on different visual clues to make that call, and experts are more consistent. That gap matters because it means training and experience directly influence diagnostic accuracy, which can change patient outcomes.
This is a small, exploratory post hoc analysis of 13 serrated lesions and reader responses. The decision trees show associations between features and classification, not the actual cognitive steps endoscopists took. The study does not address how to distinguish sessile serrated lesions from hyperplastic polyps, nor does it validate a diagnostic algorithm. Results may change if unassessable responses are excluded, especially for non-experts.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Feature-Based Patterns Associated With Japan NBI Expert Team Type 1 Classification of Serrated Colorectal Lesions: A Post Hoc Decision Tree Analysis — Taku Sakamoto, Hiroyuki Takamaru, Daizen Hirata et al. — DEN open — doi:10.1002/deo2.70428
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper.