Small-bowel Capsule Endoscopy Images Research Articles

Segmentation and region quantification of bubbles in small bowel capsule endoscopy images using wavelet transform

ObjectiveA large number of captured frames by the wireless capsule endoscopy have been contaminated with different amounts of bubbles. Bubbles can degrade the visualization quality of the small intestine mucosa. The aim of this study is to develop an objective method for evaluating the amount of bubbles in WCE images. MethodsFrames with varying levels of bubble occlusion were selected from the Kvasir capsule endoscopy dataset. The round shape bubbles have an edge in their boundaries. Edges in the spatial domain correspond to high-frequency bands in the frequency domain. Two automated edge detection approaches have been developed in a rule-based manner and evaluated to assess the amount of bubbles. The first approach involved high pass filtering using fast Fourier transform (FFT), while the second approach has been based on wavelet image decomposition and reconstruction by omitting approximation coefficients subband. ResultsBoth Fourier and wavelet transforms obtained approximately the same dice similarity score (DSC), and precision metrics, which were equal to 0.87, and 0.91, respectively. Based on the specificity measure, the FFT outperformed the Hough and wavelet transforms. However, the wavelet transform obtained a higher dice similarity score (DSC) (0.93), accuracy (0.95), and sensitivity metric (0.97) and was the fastest, with an execution time of 0.01 s per frame, making it suitable for real-time applications. ConclusionThe proposed technique provides an easy-to-implement method for quality reporting or objective comparison tools of different bowel preparation paradigms in real-time applications due to its fast execution time. The obtained results from two different datasets proved that the presented method has good generalization.

Poor Quality of Small Bowel Capsule Endoscopy Images Has a Significant Negative Effect in the Diagnosis of Small Bowel Malignancy.

Background and AimsCapsule endoscopy (CE) is a visual modality; hence, diagnosis relies on image quality. We studied the contribution of image parameters to visualization quality and their effect on diagnostic certainty of small bowel (SB) lesions.MethodsFive clear CE images of common SB pathology – two vascular lesions, two inflammatory, one polyp – were processed for three image parameters to simulate poor SB conditions: opacity (color-matched to luminal content; 10–90%, 10% increments); blurriness (radius 1–10 pixels; one pixel increments); and contrast (−50-50%; 10% increments). Nine expert readers evaluated whether images were adequate for diagnosis. Points where perception of image quality changed significantly were determined for each parameter. Three further sets of SBCE images (vascular, inflammatory, and neoplastic lesions; nine images/set) were processed for four points/parameters. Twenty experienced/expert CE readers reviewed these images.ResultsThe negative effects of opacity in diagnostic certainty were mostly evident in images of neoplasia; images of vascular and inflammatory lesions were less affected. Similar results were observed with increasing blur radius, simulating movement, and poor focus. The proportions of readers finding vascular and inflammatory images adequate for diagnosis did not drop significantly at wider blur radii, while images of neoplasia were quickly deemed inadequate. Low contrast had a greater negative effect than high, most consistently in neoplastic lesions.ConclusionPoor visualization quality in all parameters affected mostly neoplastic lesions. Software to increase contrast and sharpen images can improve visualization quality; smart frame rate adaptation could improve the number of high-quality frames obtained. Thoroughness in SB cleansing is most important when there is a suspicion of neoplasia.

Gastroenterologist-Level Identification of Small-Bowel Diseases and Normal Variants by Capsule Endoscopy Using a Deep-Learning Model

Capsule endoscopy has revolutionized investigation of the small bowel. However, this technique produces a video that is 8-10 hours long, so analysis is time consuming for gastroenterologists. Deep convolutional neural networks (CNNs) can recognize specific images among a large variety. We aimed to develop a CNN-based algorithm to assist in the evaluation of small bowel capsule endoscopy (SB-CE) images. We collected 113,426,569 images from 6970 patients who had SB-CE at 77 medical centers from July 2016 through July 2018. A CNN-based auxiliary reading model was trained to differentiate abnormal from normal images using 158,235 SB-CE images from 1970 patients. Images were categorized as normal, inflammation, ulcer, polyps, lymphangiectasia, bleeding, vascular disease, protruding lesion, lymphatic follicular hyperplasia, diverticulum, parasite, and other. The model was further validated in 5000 patients (no patient was overlap with the 1970 patients in the training set); the same patients were evaluated by conventional analysis and CNN-based auxiliary analysis by 20 gastroenterologists. If there was agreement in image categorization between the conventional analysis and CNN model, no further evaluation was performed. If there was disagreement between the conventional analysis and CNN model, the gastroenterologists re-evaluated the image to confirm or reject the CNN categorization. In the SB-CE images from the validation set, 4206 abnormalities in 3280 patients were identified afterfinal consensus evaluation. The CNN-based auxiliary model identified abnormalities with 99.88% sensitivity in theper-patient analysis (95% CI, 99.67-99.96) and 99.90% sensitivity in the per-lesion analysis (95% CI, 99.74-99.97). Conventional reading by the gastroenterologists identified abnormalities with 74.57% sensitivity (95% CI, 73.05-76.03) in the per-patient analysis and 76.89% in the per-lesion analysis (95% CI, 75.58-78.15). The mean reading time per patient was 96.6 ± 22.53 minutes by conventional reading and5.9 ± 2.23 minutes by CNN-based auxiliary reading (P<.001). We validated the ability of a CNN-based algorithm to identify abnormalities in SB-CE images. The CNN-based auxiliary model identified abnormalities with higher levels of sensitivity and significantly shorter reading times than conventional analysis by gastroenterologists. This algorithm provides an important tool to help gastroenterologists analyze SB-CE images more efficiently and more accurately.

Improving diagnostic yield in obscure gastrointestinal bleeding--how virtual chromoendoscopy may be the answer.

A significant proportion of patients presenting with obscure gastrointestinal bleeding (OGIB) have negative small bowel capsule endoscopy (SBCE) examinations, and yet remain at risk of rebleeding. We aimed to evaluate whether a second-look review of SBCE images using flexible spectral color enhancement (FICE) may improve the detection of potentially bleeding lesions. This was a retrospective, single-center study including consecutive patients with OGIB subjected to SBCE, whose standard white light examination was nondiagnostic. Each SBCE was reviewed using FICE 1. New findings were labeled as either P1 or P2 lesions according to bleeding potential. Patients were followed up to assess the incidence of rebleeding. A total of 42 consecutive patients were included. Sixteen patients (38%) experienced rebleeding after a mean follow-up of 26 months. Review of SBCE images using FICE 1 enabled the identification of previously unrecognized P2 lesions, mainly angioectasias, in nine patients (21%) and P1 lesions, mainly erosions, in 26 patients (62%). Among patients who experienced rebleeding, 13/16 (81%) were diagnosed with P1 lesions with FICE 1 (P=0.043), whereas 3/16 (19%) had confirmed nondiagnostic SBCE and only 1/16 (6%) had newly diagnosed P2 (plus P1) lesions. An alternative source of bleeding outside the small bowel was found in only 3/16 (19%) patients with rebleeding during the follow-up. In a significant proportion of patients with OGIB, FICE 1 may detect potentially bleeding lesions previously missed under conventional white light SBCE. Review of nondiagnostic SBCE with FICE 1 may be a valuable strategy to obviate the need for further investigations in patients with OGIB, particularly for those who experience rebleeding.

A critical analysis of the effect of view mode and frame rate on reading time and lesion detection during capsule endoscopy.

Factors influencing reading time and detection of lesions include the view mode (VM) and frame rate (FR) applied during reading of small bowel capsule endoscopy images. The aims of this study were to examine the impact of VM and FR on reading time and lesion detection using a standardized, single-type lesion model. A selected video clip containing a known number of positive images (n = 60) of small bowel angioectasias was read using nine different combinations of VM and FR (VM1, VM2, and VM4 × FR10, FR15, and FR25) in randomized order by six capsule endoscopists. Readers were asked to count all positive images of angioectasias (maximum number of positive images, MPIs) seen during reading. The main outcome measurements were effect of VM and FR on reading time and lesion detection. Mean MPIs for all VM2 and VM4 were 36 (60 %) and 38 (64 %). They were significantly higher than VM1 of 24 (40 %) (P = 0.011, 0.008). A statistical difference was found when the total MPIs at FR10 were compared to FR15 (P = 0.008) and to FR25 (P < 0.001). Both VM and FR significantly influence lesion detection during capsule endoscopy reading.

PTU-021 3D Reconstruction in Capsule Endoscopy; a Feasibility Study

IntroductionThree-dimensional (3D) imaging in small-bowel capsule endoscopy (SBCE) is not currently feasible due to hardware limitations. However, there has been increasing use of software algorithms to enable 3D reconstruction in...

Evaluation of 4 three-dimensional representation algorithms in capsule endoscopy images

To evaluate the three-dimensional (3-D) representation performance of 4 publicly available Shape-from-Shading (SfS) algorithms in small-bowel capsule endoscopy (SBCE). SfS techniques recover the shape of objects using the gradual variation of shading. There are 4 publicly available SfS algorithms. To the best of our knowledge, no comparative study with images obtained during clinical SBCE has been performed to date. Three experienced reviewers were asked to evaluate 54 two-dimensional (2-D) images (categories: protrusion/inflammation/vascular) transformed to 3-D by the aforementioned SfS 3-D algorithms. The best algorithm was selected and inter-rater agreement was calculated. Four publicly available SfS algorithms were compared. Tsai's SfS algorithm outperformed the rest (selected as best performing in 45/54 SBCE images), followed by Ciuti's algorithm (best performing in 7/54 images) and Torreão's (in 1/54 images). In 26/54 images; Tsai's algorithm was unanimously selected as the best performing 3-D representation SfS software. Tsai's 3-D algorithm superiority was independent of lesion category (protrusion/inflammatory/vascular; P = 0.678) and/or CE system used to obtain the 2-D images (MiroCam/PillCam; P = 0.558). Lastly, the inter-observer agreement was good (kappa = 0.55). 3-D representation software offers a plausible alternative for 3-D representation of conventional capsule endoscopy images (until optics technology matures enough to allow hardware enabled-"real" 3-D reconstruction of the gastrointestinal tract).

Nurses Can Interpret Small-Bowel Capsule Endoscopy Accurately

Capsule endoscopy images are not operator-dependent; therefore, they can be read and used by nonphysician personnel. In two previous studies, researchers have shown a high correlation between endoscopists’ and nurses’ interpretations of small-bowel capsule endoscopy images. In this study, a highly experienced …