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

Real-time gastric intestinal metaplasia diagnosis tailored for bias and noisy-labeled data with multiple endoscopic imaging

This work presents real-time segmentation viz. gastric intestinal metaplasia (GIM). Recently, GIM segmentation of endoscopic images has been carried out to differentiate GIM from a healthy stomach. However, real-time detection is difficult to achieve. Conditions are challenging, and include multiple color modes (white light endoscopy and narrow-band imaging), other abnormal lesions (erosion and ulcer), noisy labels etc. Herein, our model is based on BiSeNet and can overcome the many issues regarding GIM. Application of auxiliary head and additional loss are seen to improve performance as well as enhance multiple color modes accurately. Further, multiple pre-processing techniques are utilized for leveraging detection performance: namely, location-wise negative sampling, jigsaw augmentation, and label smoothing. Finally, the decision threshold can be adjusted separately for each color mode. Work undertaken at King Chulalongkorn Memorial Hospital examined 940 histologically proven GIM images and 1239 non-GIM images, obtained over 173 frames per second (FPS). In terms of accuracy, our model is seen to outperform all baselines. Our results demonstrate sensitivity, specificity, positive predictive, negative predictive, accuracy, and mean intersection over union (IoU), achieving GIM segmentation values of 91%, 96%, 91%, 91%, 96%, and 55%, respectively.

Development and validation of a feature extraction-based logical anthropomorphic diagnostic system for early gastric cancer: A case-control study

BACKGROUND

Prompt diagnosis of early gastric cancer (EGC) is crucial for improving patient survival. However, most previous computer-aided-diagnosis (CAD) systems did not concretize or explain diagnostic theories. We aimed to develop a logical anthropomorphic artificial intelligence (AI) diagnostic system named ENDOANGEL-LA (logical anthropomorphic) for EGCs under magnifying image enhanced endoscopy (M-IEE).

METHODS

We retrospectively collected data for 692 patients and 1897 images from Renmin Hospital of Wuhan University, Wuhan, China between Nov 15, 2016 and May 7, 2019. The images were randomly assigned to the training set and test set by patient with a ratio of about 4:1. ENDOANGEL-LA was developed based on feature extraction combining quantitative analysis, deep learning (DL), and machine learning (ML). 11 diagnostic feature indexes were integrated into seven ML models, and an optimal model was selected. The performance of ENDOANGEL-LA was evaluated and compared with endoscopists and sole DL models. The satisfaction of endoscopists on ENDOANGEL-LA and sole DL model was also compared.

FINDINGS

Random forest showed the best performance, and demarcation line and microstructures density were the most important feature indexes. The accuracy of ENDOANGEL-LA in images (88.76%) was significantly higher than that of sole DL model (82.77%, p = 0.034) and the novices (71.63%, p<0.001), and comparable to that of the experts (88.95%). The accuracy of ENDOANGEL-LA in videos (87.00%) was significantly higher than that of the sole DL model (68.00%, p<0.001), and comparable to that of the endoscopists (89.00%). The accuracy (87.45%, p<0.001) of novices with the assistance of ENDOANGEL-LA was significantly improved. The satisfaction of endoscopists on ENDOANGEL-LA was significantly higher than that of sole DL model.

INTERPRETATION

We established a logical anthropomorphic system (ENDOANGEL-LA) that can diagnose EGC under M-IEE with diagnostic theory concretization, high accuracy, and good explainability. It has the potential to increase interactivity between endoscopists and CADs, and improve trust and acceptability of endoscopists for CADs.

FUNDING

This work was partly supported by a grant from the Hubei Province Major Science and Technology Innovation Project (2018-916-000-008) and the Fundamental Research Funds for the Central Universities (2042021kf0084).

An In Vivo Proposal of Cell Computing Inspired by Membrane Computing

Intractable problems are challenging and not uncommon in Computer Science. The computing generation we are living in forces us to look for an alternative way of computing, as current computers are facing limitations when dealing with complex problems and bigger input data. Physics and Biology offer great alternatives to solve these problems that traditional computers cannot. Models like Quantum Computing and cell computing are emerging as possible solutions to the current problems the conventional computers are facing. This proposal describes an in vivo framework inspired by membrane computing and based on alternative computational frameworks that have been proven to be theoretically correct such as chemical reaction series. The abilities of a cell as a computational unit make this proposal a starting point in the creation of feasible potential frameworks to enhance the performance of applications in different disciplines such as Biology, BioMedicine, Computer networks, and Social Sciences, by accelerating drastically the way information is processed by conventional architectures and possibly achieving results that presently are not possible due to the limitations of the current computing paradigm. This paper introduces an in vivo solution that uses the principles of membrane computing and it can produce non-deterministic outputs.

Application of computer image segmentation algorithm based on visual features

In order to further satisfy the needs of visual perception, a computer image segmentation algorithm based on visual characteristics is proposed. The computer image geometric features are linearly segmented to obtain multiple sub-panels and various geometric features of each sub-panel are extracted. The geometric features are input as low-level features into the deep neural network model to learn to generate high-level features. Finally, based on the high-level features, the clustering center is obtained by Gaussian mixture model, and the final segmentation result is obtained by using graph cut. The results of experiments on Princeton standard data set and COSEG dataset show that the rand index RI value is the best value for each 3D model, which shows that the proposed method is better than the traditional segmentation method. It has good consistent segmentation results. The research showed that using a variety of geometric features compared to a single geometric feature, the obtained features have a more comprehensive geometric meaning, which can effectively make image segmentation meet the visual characteristics requirements.

Objective image analysis of non-magnifying image-enhanced endoscopy for diagnosis of small depressed early gastric cancers

Background  Gastric cancers (GC) after H. pylori eradication are difficult to diagnose even by magnifying narrow-band imaging (NBI) or blue laser imaging (BLI) endoscopy. Little is known with regard to non-magnifying (NM)-NBI/BLI for early GC so we examined the efficacy of NM-NBI/BLI for early GC diagnosis. Methods  We retrospectively analyzed the images of 29 small (≤ 1 cm) intramucosal GC that had been treated with endoscopic submucosal dissection and 137 benign depressed lesions (BDLs). The brightness and shape of the GCs and BDLs by NM-NBI/BLI were assessed with ImageJ software. Results  The NBI/BLI-index, which indicates the brightness of NBI/BLI for visualization, was significantly higher in GC than BDLs in both the H. pylori -infected ( P  = 0.009) and -eradicated group ( P  < 0.0001), indicating that GC exhibited brighter colors than the normal surrounding mucosa. The C-index, which refers to the circularity of the lesion, was also significantly higher in GC than BDLs in both H. pylori -infected ( P  = 0.006) and -eradicated cases ( P  = 0.004). Based on receiver-operating characteristic curve analysis, cutoff values for the NBI/BLI- and C-indices for GC were 1.04 and 0.58 in the H. pylori -infected cases, and 0.98 and 0.64 in the H. pylori -eradicated cases. With the reference value of the NBI/BLI-index set at ≥ 0.69 with the C-index at ≥ 0.21 in the H. pylori -infected and the NBI/BLI-index at ≥ 0.80 with the C-index at ≥ 0.32 in the H. pylori -eradicated cases, both the sensitivity and negative predictive value for early GC were 100 %. A high NBI/BLI-index tended to be associated with a wide length of the intervening part histologically in the H. pylori -eradicated cases ( P  = 0.09). Conclusions  The small depressed-type early GC had brighter color and rounder shape compared to BDLs in both H. pylori -infected and -eradicated cases. The NBI/BLI- and C-indices calculated by the image analysis may facilitate identification of small depressed-type GC.

Computer-aided diagnosis for identifying and delineating early gastric cancers in magnifying narrow-band imaging

BACKGROUND AND AIMS

Magnifying narrow-band imaging (M-NBI) is important in the diagnosis of early gastric cancers (EGCs) but requires expertise to master. We developed a computer-aided diagnosis (CADx) system to assist endoscopists in identifying and delineating EGCs.

METHODS

We retrospectively collected and randomly selected 66 EGC M-NBI images and 60 non-cancer M-NBI images into a training set and 61 EGC M-NBI images and 20 non-cancer M-NBI images into a test set. After preprocessing and partition, we determined 8 gray-level co-occurrence matrix (GLCM) features for each partitioned 40 × 40 pixel block and calculated a coefficient of variation of 8 GLCM feature vectors. We then trained a support vector machine (SVM_Lv1) based on variation vectors from the training set and examined in the test set. Furthermore, we collected 2 determined P and Q GLCM feature vectors from cancerous image blocks containing irregular microvessels from the training set, and we trained another SVM (SVM_Lv2) to delineate cancerous blocks, which were compared with expert-delineated areas for area concordance.

RESULTS

The diagnostic performance revealed accuracy of 96.3%, precision (positive predictive value [PPV]) of 98.3%, recall (sensitivity) of 96.7%, and specificity of 95%, at a rate of 0.41 ± 0.01 seconds per image. The performance of area concordance, on a block basis, demonstrated accuracy of 73.8% ± 10.9%, precision (PPV) of 75.3% ± 20.9%, recall (sensitivity) of 65.5% ± 19.9%, and specificity of 80.8% ± 17.1%, at a rate of 0.49 ± 0.04 seconds per image.

CONCLUSIONS

This pilot study demonstrates that our CADx system has great potential in real-time diagnosis and delineation of EGCs in M-NBI images.