PTU-127 Poor quality of capsule endoscopy images has negative effect on diagnosis of small bowel malignancy

Introduction Capsule endoscopy (CE) is the prime mode of investigation for small bowel (SB) pathology. It relies heavily on image quality, which is commonly affected by poor preparation. Currently, there is no widely-accepted method for quantifying visualisation quality. We studied the contribution of various image parameters to visualisation quality and their effect on certainty of diagnosis of small bowel lesions. Methods Five clear CE images of common SB pathology were processed for 3 parameters to simulate increasingly poor SB preparation: opacity (colour-matched to luminal content; 10–90%, 10% increments), blurriness (radius 1–10px; 1px increments), contrast (-50–50%; 10% increments). 9 expert readers evaluated whether images were adequate for diagnostic purposes. The points where perception of image quality changed significantly were determined for each parameter. Three further sets of SBCE images (vascular, inflammatory and malignant lesions) were processed for 4 points/parameter. 20 experienced-expert CE readers from reviewed the resulting images. Results In vascular and inflammatory lesions, diagnostic certainty was least affected by increasing image opacity, requiring opacities >90% before most readers considered images inadequate for diagnosis. The greatest negative effects of image opacity were seen in malignancy where significantly fewer readers found images adequate at >50% opacity. Similar results obtained with increasing blur radius, simulating motion blur and poor focus. The proportions of readers finding vascular and inflammatory images adequate for diagnosis did not drop significantly at wider blur radii, while the proportion who found images of malignancies diagnostically adequate dropped at blur radius 6px. Decreasing contrast had greater negative effect than raised contrast, most obvious in malignant lesions. Conclusions Poor visualisation quality in all parameters had the greatest effect on malignant lesions. Software to increase contrast and sharpen images can improve visualisation quality; smart frame rate adaptation could also improve the number of high-quality frames obtained. Furthermore, our results suggest that thoroughness in SB cleansing is most important when there is suspicion of SB malignancy, to improve diagnostic certainty of images obtained.