Multiscale texture descriptors for automatic small bowel tumors detection in capsule endoscopy

Abstract

Conventional endoscopic exams do not allow the entire visualization of the gastrointestinal (GI) tract. Push enteroscopy (PE) is an effective diagnostic and therapeutic procedure, although it only allows exploration of the proximal small bowel (Pennazio et al., 1995). Simultaneously, convetional colonoscopy is limited at the terminal ileum. Therefore, prior to the wireless capsule endoscopy era, the small intestine was the conventional endoscopy’s last frontier, because it could not be internally visualized directly or in it’s entirely by any method (Herrerías & Mascarenhas-Saraiva, 2007). The small intestine accounts for 75% of the total length and 90% of the surface area of the gastrointestinal tract. In adults it measures about 570 cm at post mortem, which is substantially longer than conventional video endoscopes (100-180 cm) (Swain & Fritscher-Ravens, 2004). Intraoperative enteroscopy is the most complete but also the most invasive means of examining the small bowel (Gay et al., 1998). Given the technical and medical improvements introduced on the assessment of the gastrointestinal (GI) tract, Capsule Endoscopy (CE) is considered as the first major technological innovation in GI diagnostic medicine since the flexible endoscope (Kaffes, 2009). More recently, a new technique, the double-balloon enteroscopy (DBE), has been introduced into clinical practice (Yamamoto & Kita, 2006). DBE has the potential to examine the entire length of the small bowel with biopsy and therapeutic capability. Nevertheless, it is a time consuming procedure that requires specialist training for the operating physician. We should note that DBE and CE are complementary tools and not competitive (Chen et al., 2007). Hence, the diagnostic ease of CE can be complemented with a targeted and often therapeutic DBE (Kaffes, 2009). Therefore, CE can be used as a first line diagnosis method, while DBE can be used as a confirmatory or therapeutic modality for lesions first visualized by CE (Pennazio, 2006). The endoscopic capsule is a pill-like device, with only 11mm x 26 mm, and includes a miniaturized camera, a light source and circuits for the acquisition and wireless transmission of signals (Iddan et al., 2000). As the capsule moves through GI tract, propelled exclusively by peristalsis, it acquires images at a rate of two per second and sends them to a hard disk receiver that is worn in the belt of the patient, in a wireless communication scheme. The acquisition of images is limited by the battery life of the device, usually around eight hours, which imply that in a single CE exam more than 50000 images are acquired. If no complications arise, the capsule should be in the patient’s stool, usually within 24-48 h, and not reused (Pennazio, 2006). Capsule endoscopy has evolved in a few short years to become a first-line, noninvasive diagnostic technique for the small bowel. CE is now being utilized worldwide to assess patients for obscure gastrointestinal bleeding, possible Crohn’s disease, celiac disease and small bowel tumors (Lee & Eisen, 2010). It is now available in over 4500 practice sites around the world (Munoz-Navas, 2009). The time required to a physician to analyze the resulting video is, on average, 40-60 min (Pennazio, 2006). The reading time and interpretation of CE exams is very time consuming given that more than 50,000 images have to be reviewed (Delvaux & Gay, 2006; Mergener et al., 2007), which contributes to the high cost of a CE exam (Westerhof et al., 2009). Thus, a computer assisted diagnosis tool to help the physicians to evaluate CE exams faster and more accurately is an important technical challenge and an excellent economical opportunity.