Impact of visual features on the segmentation of gastroenterology images using normalized cuts

Computer assisted gastric abnormalities detection using hybrid texture descriptors for chromoendoscopy images

BACKGROUND AND OBJECTIVE

The early diagnosis of stomach cancer can be performed by using a proper screening procedure. Chromoendoscopy (CH) is an image-enhanced video endoscopy technique, which is used for inspection of the gastrointestinal-tract by spraying dyes to highlight the gastric mucosal structures. An endoscopy session can end up with generating a large number of video frames. Therefore, inspection of every individual endoscopic-frame is an exhaustive task for the medical experts. In contrast with manual inspection, the automated analysis of gastroenterology images using computer vision based techniques can provide assistance to endoscopist, by finding out abnormal frames from the whole endoscopic sequence.

METHODS

In this paper, we have presented a new feature extraction method named as Gabor-based gray-level co-occurrence matrix (G2LCM) for computer-aided detection of CH abnormal frames. It is a hybrid texture extraction approach which extracts a combination both local and global texture descriptors. Moreover, texture information of a CH image is represented by computing the gray level co-occurrence matrix of Gabor filters responses. Furthermore, the second-order statistics of these co-occurrence matrices are computed to represent images' texture.

RESULTS

The obtained results show the possibility to correctly classifying abnormal from normal frames, with sensitivity, specificity, accuracy, and area under the curve as 91%, 82%, 87% and 0.91 respectively, by using a support vector machine classifier and G2LCM texture features.

CONCLUSION

It is apparent from results that the proposed system can be used for providing aid to the gastroenterologist in the screening of the gastric tract. Ultimately, the time taken by an endoscopic procedure will be sufficiently reduced.

Hue-texture-embedded region-based model for magnifying endoscopy with narrow-band imaging image segmentation based on visual features

BACKGROUND AND OBJECTIVE

Magnification endoscopy with narrow-band imaging (ME-NBI) has become a feasible tool for detecting diseases within the human gastrointestinal tract, and is more applied by physicians to search for pathological abnormalities with gastric cancer such as precancerous lesions, early gastric cancer and advanced cancer. In order to improve the reliability of diseases detection, there is a need for applying or proposing computer-assisted methodologies to efficiently analyze and process ME-NBI images. However, traditional computer vision methodologies, mainly segmentation, do not express well to the specific visual characteristics of NBI scenario.

METHODS

In this paper, two energy functional items based on specific visual characteristics of ME-NBI images have been integrated in the framework of Chan-Vese model to construct the Hue-texture-embedded model. On the one hand, a global hue energy functional was proposed representing a global color information extracted in H channel (HSI color space). On the other hand, a texture energy was put forward presenting local microvascular textures extracted by the PIF of adaptive threshold in S channel.

RESULTS

The results of our model have been compared with Chan-Vese model and manual annotations marked by physicians using F-measure and false positive rate. The value of average F-measure and FPR was 0.61 and 0.16 achieved through the Hue-texture-embedded region-based model. And the C-V model achieved the average F-measure and FPR value of 0.52 and 0.32, respectively. Experiments showed that the Hue-texture-embedded region-based outperforms Chan-Vese model in terms of efficiency, universality and lesion detection.

CONCLUSIONS

Better segmentation results are acquired by the Hue-texture-embedded region-based model compared with the traditional region-based active contour in these five cases: chronic gastritis, intestinal metaplasia and atrophy, low grade neoplasia, high grade neoplasia and early gastric cancer. In the future, we are planning to expand the universality of our proposed methodology to segment other lesions such as intramucosal cancer etc. As long as these issues are solved, we can proceed with the classification of clinically relevant diseases in ME-NBI images to implement a fully automatic computer-assisted diagnosis system.

Pairwise Latent Semantic Association for Similarity Computation in Medical Imaging

Retrieving medical images that present similar diseases is an active research area for diagnostics and therapy. However, it can be problematic given the visual variations between anatomical structures. In this paper, we propose a new feature extraction method for similarity computation in medical imaging. Instead of the low-level visual appearance, we design a CCA-PairLDA feature representation method to capture the similarity between images with high-level semantics. First, we extract the PairLDA topics to represent an image as a mixture of latent semantic topics in an image pair context. Second, we generate a CCA-correlation model to represent the semantic association between an image pair for similarity computation. While PairLDA adjusts the latent topics for all image pairs, CCA-correlation helps to associate an individual image pair. In this way, the semantic descriptions of an image pair are closely correlated, and naturally correspond to similarity computation between images. We evaluated our method on two public medical imaging datasets for image retrieval and showed improved performance.

Magnifying endoscopy of gastric epithelial dysplasia based on the morphologic characteristics

AIM: To investigate the difference in magnifying endoscopic findings of gastric epithelial dysplasias (GEDs) according to the morphologic characteristics.

METHODS: This study included 46 GED lesions in 45 patients who underwent magnifying endoscopy using narrow band imaging (ME-NBI) before endoscopic resection. During ME-NBI, the microvascular and microsurface (MS) patterns and the presence of light blue crest (LBC) and white opaque substance were investigated. GEDs were categorized as adenomatous, foveolar, and hybrid types, and their mucin phenotype was evaluated.

RESULTS: Of the 46 lesions, 27 (59%) were categorized as adenomatous, 15 (32%) as hybrid, and the remaining 4 (9%) as foveolar. All adenomatous GEDs showed the round pit and/or tubular MS patterns, all foveolar GEDs showed the papillary pattern, and hybrid GEDs showed mixed patterns (P < 0.001). LBC was more frequently observed in adenomatous GEDs than in hybrid or foveolar GEDs (52%, 33%, 0%, respectively), although this difference was not significant (P = 0.127). The papillary MS pattern was associated with MUC5AC and MUC6 expression, and the round pit and/or tubular MS patterns were associated with CD10 expression.

CONCLUSION: The MS pattern in ME-NBI findings is useful for predicting the morphologic category and mucin phenotype of GEDs, and ME-NBI findings may guide decisions regarding GED treatment.