Distinguishing patients with celiac disease by quantitative analysis of videocapsule endoscopy images

Abstract

Background Although videocapsule endoscopy images are helpful in the evaluation of celiac disease, their interpretation is subjective. Quantitative disease markers could assist in determining the extent of villous atrophy and response to treatment. Method Capsule endoscopy images were acquired from celiac patients with small bowel pathology ( N = 11) and from control patients ( N = 10). Image resolution was 576 × 576 pixels in dimension, 256 grayscale levels, and had a 2 s −1 frame rate. Pixel brightness and image texture were measured over 10 × 10 pixel subimages and then averaged for 56 × 56 subimages per frame. Measurements were obtained at five locations from proximal to distal small intestine in each patient. At each location, measurements were calculated using 200 consecutive image frames (100 s). Mean frame-to-frame pixel brightness, image texture, and periodicity in brightness, an estimate of wall motion or intestinal motility, were computed and used for classification with a nonlinear discriminant function. Results From pooled data, celiac images had greater texture than did images from control patients ( p < 0.001) and exhibited more frame-to-frame brightness variation as well ( p = 0.032). The dominant period of brightness was longer in celiacs ( p = 0.001), possibly indicating decreased motility. Using the markers for three-dimensional nonlinear classification of celiacs versus controls, sensitivity was 92.7% and specificity was 93.5%. The relationship between dominant period and small intestinal transit time was approximately linear for both celiacs and controls ( r 2 = 0.42 and r 2 =0 .55, respectively). Conclusions Videocapsule images can be quantified to detect villous atrophy throughout the small intestine, and to distinguish individuals with celiac disease from individuals lacking mucosal atrophy.