A Detection Method for Circumferential Alignment of Diminutive Lesions Using Wavelet Transform Modulus Maxima and Higher-Order Local Autocorrelation

The number of patients with Crohn’s disease is increasing annually. Since the cause of Crohn’s disease remains unclear, early detection and appropriate treatment are crucial. Diagnostic criteria for Crohn’s disease include features such as erosions, ulcers, and the circumferential alignment of diminutive lesions. Medical professionals employ capsule endoscopy for diagnosis. While existing research focuses on erosions and ulcers, studies on the circumferential alignment of diminutive lesions are limited. Therefore, this paper presents a classification of images showing the circumferential alignment of diminutive lesions and normal images obtained through capsule endoscopy. We propose a classification method that utilizes the Wavelet Transform Modulus Maxima (WTMM) to extract contours of circumferential alignment, obtains features using Higher-order Local Autocorrelation (HLAC), and classifies them using a Support Vector Machine (SVM). Our method accurately classified circumferential alignment of diminutive lesions and normal images with an accuracy of 98.2%.