LAUSR

Predicting the invasion depth of superficial esophageal neoplasms (SENs) is crucial for an indication toward endoscopic treatment. We aimed to develop imaging biomarkers to predict the invasion depth of SENs using the intrapapillary capillary loop (IPCL) patterns. Texture and fractal analyses were performed on IPCL demarcations. Logistic regression models were built to distinguish narrow‐band images (NBIs) of T1b from T1a and high‐grade dysplasia (HGD). Classification performances of a base model composed of the age, sex, height, and body mass index of the patient, and a morphometric model comprising the texture and fractal metrics along with the base model parameters were compared. NBIs of 39 patients including T staging of 6 HGD, 22 T1a, and 11 T1b were analyzed. In the logistic regression by a multivariable generalized linear model classifying histological depths, only texture entropy was significantly and positively associated with the invasion depth. Lower texture entropy of IPCL demarcations was significantly associated with diagnoses of T1b (odds ratio: 0.144, 95% confidence interval: 0.019–0.687) in comparison to the stratum with higher texture entropy. The performance of the morphometric model represented by areas under the receiver operating curves (0.795) was better than that of the base model (0.562). We suggest morphometric analyses, particularly texture entropy, have the potential as image biomarkers to aid in the prediction of SEN invasion depth using IPCL patterns.