Artificial Intelligence-Based Diagnosis of Gastric Mesenchymal Tumors Using Digital Endosonography Image Analysis.

Background/Objectives: Artificial intelligence (AI)-assisted endoscopic ultrasonography (EUS) diagnostic tools have shown excellent performance in diagnosing gastric mesenchymal tumors. This study aimed to assess whether incorporating clinical and endoscopic factors into AI-assisted EUS classification models based on digital image analysis could improve the diagnostic performance of AI-assisted EUS diagnostic tools. Methods: We retrospectively analyzed the data of 464 patients who underwent both EUS and surgical resection of gastric mesenchymal tumors, including 294 gastrointestinal stromal tumors (GISTs), 52 leiomyomas, and 41 schwannomas. AI-assisted classification models for GISTs and non-GIST tumors were developed utilizing clinical and endoscopic factors and digital EUS image analysis. Results: Regarding the baseline EUS classification models, the area under the receiver operating characteristic (AUC) values of the logistic regression, decision tree, random forest, K-nearest neighbor (KNN), and support vector machine (SVM) models were 0.805, 0.673, 0.781, 0.740, and 0.791, respectively. Using the new classification models incorporating clinical and endoscopic factors into the baseline classification models, the AUC values of the logistic regression, decision tree, random forest, KNN, and SVM models increased to 0.853, 0.715, 0.896, 0.825, and 0.794, respectively. In particular, the random forest and KNN models exhibited significant improvement in performance in Delong's test (both p < 0.001). Conclusion: The diagnostic performance of the AI-assisted EUS classification models improved when clinical and endoscopic factors were incorporated. Our results provided direction for developing new AI-assisted EUS models for gastric mesenchymal tumors.

Intelligent Automatic Segmentation of Wrist Ganglion Cysts Using DBSCAN and Fuzzy C-Means.

Ganglion cysts are common soft tissue masses of the hand and wrist, and small size cysts are often hypoechoic. Thus, identifying them from ultrasonography is not an easy problem. In this paper, we propose an automatic segmentation method using two artificial intelligence algorithms in sequence. A density based unsupervised learning algorithm called DBSCAN is performed as a front-end and its result determines the number of clusters used in the Fuzzy C-Means (FCM) clustering algorithm for quantification of ganglion cyst object. In an experiment using 120 images, the proposed method shows a higher extraction rate (89.2%) and lower false positive rate compared with FCM when the ground truth is set as the human expert's decision. Such human-like behavior is more apparent when the size of the ganglion cyst is small that the quality of ultrasonography is often not very high. With this fully automatic segmentation method, the operator subjectivity that is highly dependent on the experience of the ultrasound examiner can be mitigated with high reliability.

Application of A Convolutional Neural Network in The Diagnosis of Gastric Mesenchymal Tumors on Endoscopic Ultrasonography Images.

BACKGROUND AND AIMS: Endoscopic ultrasonography (EUS) is a useful diagnostic modality for evaluating gastric mesenchymal tumors; however, differentiating gastrointestinal stromal tumors (GISTs) from benign mesenchymal tumors such as leiomyomas and schwannomas remains challenging. For this reason, we developed a convolutional neural network computer-aided diagnosis (CNN-CAD) system that can analyze gastric mesenchymal tumors on EUS images. METHODS: A total of 905 EUS images of gastric mesenchymal tumors (pathologically confirmed GIST, leiomyoma, and schwannoma) were used as a training dataset. Validation was performed using 212 EUS images of gastric mesenchymal tumors. This test dataset was interpreted by three experienced and three junior endoscopists. RESULTS: The sensitivity, specificity, and accuracy of the CNN-CAD system for differentiating GISTs from non-GIST tumors were 83.0%, 75.5%, and 79.2%, respectively. Its diagnostic specificity and accuracy were significantly higher than those of two experienced and one junior endoscopists. In the further sequential analysis to differentiate leiomyoma from schwannoma in non-GIST tumors, the final diagnostic accuracy of the CNN-CAD system was 72.5%, which was significantly higher than that of two experienced and one junior endoscopists. CONCLUSIONS: Our CNN-CAD system showed high accuracy in diagnosing gastric mesenchymal tumors on EUS images. It may complement the current clinical practices in the EUS diagnosis of gastric mesenchymal tumors.

Automatic Characterizations of Lumbar Multifidus Muscle and Intramuscular Fat with Fuzzy C-means based Quantization from Ultrasound Images.

BACKGROUND: Low Back Pain (LBP) is a common disorder involving the muscles and bones and about half of the people experience LBP at some point of their lives. Since the social economic cost and the recurrence rate over the lifetime is very high, the treatment/rehabilitation of chronic LBP is important to physiotherapists, both for clinical and research purposes. Trunk muscles such as the lumbar multifidi is important in spinal functions and intramuscular fat is also important in understanding pain control and rehabilitations. However, the analysis of such muscles and related fat require many human interventions and thus suffers from the operator subjectivity especially when the ultrasonography is used due to its cost-effectiveness and no radioactive risk. AIMS: In this paper, we propose a fully automatic computer vision based software to compute the thickness of the lumbar multifidi muscles and to analyze intramuscular fat distribution in that area. METHODS: The proposed system applies various image processing algorithms to enhance the intensity contrast of the image and measure the thickness of the target muscle. Intermuscular fat analysis is done by Fuzzy C-Means (FCM) clustering based quantization. RESULTS: In experiment using 50 DICOM format ultrasound images from 50 subjects, the proposed system shows very promising result in computing the thickness of lumbar multifidi. CONCLUSION: The proposed system have minimal discrepancy(less than 0.2 cm) from human expert for 72% (36 out of 50 cases) of the given data. Also, FCM based intramuscular fat analysis looks better than conventional histogram analysis.

Digital image analysis-based scoring system for endoscopic ultrasonography is useful in predicting gastrointestinal stromal tumors.

BACKGROUND: When gastric mesenchymal tumors (GMTs) measuring 2-5 cm in size are found, whether to undergo further treatment or not is controversial. Endoscopic ultrasonography (EUS) is useful for the evaluation of malignant potential of GMTs, but has limitations, such as subjective interpretation of EUS images. Therefore, we aimed to develop a scoring system based on the digital image analysis of EUS images to predict gastrointestinal stromal tumors (GISTs). METHODS: We included 103 patients with histopathologically proven GIST, leiomyoma or schwannoma on surgically resected specimen who underwent EUS examination between January 2007 and June 2018. After standardization of the EUS images, brightness values, including the mean (Tmean), indicative of echogenicity, and the standard deviation (TSD), indicative of heterogeneity, in the tumors were analyzed. RESULTS: Age, Tmean, and TSD were significantly higher in GISTs than in non-GISTs. The sensitivity and specificity were almost optimized for differentiating GISTs from non-GISTs when the critical values of age, Tmean, and TSD were 57.5 years, 67.0, and 25.6, respectively. A GIST-predicting scoring system was created by assigning 3 points for Tmean ≥ 67, 2 points for age ≥ 58 years, and 1 point for TSD ≥ 26. When GMTs with 3 points or more were diagnosed as GISTs, the sensitivity, specificity, and accuracy of the scoring system were 86.5%, 75.9%, and 83.5%, respectively. CONCLUSIONS: The scoring system based on the information of digital image analysis is useful in predicting GISTs in case of GMTs that are 2-5 cm in size.

Semi-dynamic Control of FCM Initialization for Automatic Extraction of Inflamed Appendix from Ultrasonography.

BACKGROUND: Current naked-eye examination of the ultrasound images for inflamed appendix has limitations due to its intrinsic operator subjectivity problem. OBJECTIVE: In this paper, we propose a fully automatic intelligent method for extracting inflamed appendix from ultrasound images. Accurate and automatic extraction of inflamed appendix from ultrasonography is a major decision making resource of the diagnosis and management of suspected appendicitis. METHODS: The proposed method uses Fuzzy C-means learning algorithm in pixel clustering with semi-dynamic control of initializing the number of clusters based on the intensity contrast dispersion of the input image. Thirty percent of the prepared ultrasonography samples are classified into four different groups based on their intensity contrast distribution and then different number of clusters are assigned to the images in accordance with such groups in Fuzzy C-means learning process. RESULTS: In the experiment, the proposed system successfully extracts the target without human intervention in 82 of 85 cases (96.47% accuracy). The proposed method also shows that it can cover the false negative cases occurred previously that used self-organizing map as the learning engine. CONCLUSION: Such high level reliable correct extraction of inflamed appendix encourages to use the automatic extraction software in the diagnosis procedure of suspected acute appendicitis.

Automatic Extraction of Appendix from Ultrasonography with Self-Organizing Map and Shape-Brightness Pattern Learning.

Accurate diagnosis of acute appendicitis is a difficult problem in practice especially when the patient is too young or women in pregnancy. In this paper, we propose a fully automatic appendix extractor from ultrasonography by applying a series of image processing algorithms and an unsupervised neural learning algorithm, self-organizing map. From the suggestions of clinical practitioners, we define four shape patterns of appendix and self-organizing map learns those patterns in pixel clustering phase. In the experiment designed to test the performance for those four frequently found shape patterns, our method is successful in 3 types (1 failure out of 45 cases) but leaves a question for one shape pattern (80% correct).

Computer Vision Based Automatic Extraction and Thickness Measurement of Deep Cervical Flexor from Ultrasonic Images.

Deep Cervical Flexor (DCF) muscles are important in monitoring and controlling neck pain. While ultrasonographic analysis is useful in this area, it has intrinsic subjectivity problem. In this paper, we propose automatic DCF extractor/analyzer software based on computer vision. One of the major difficulties in developing such an automatic analyzer is to detect important organs and their boundaries under very low brightness contrast environment. Our fuzzy sigma binarization process is one of the answers for that problem. Another difficulty is to compensate information loss that happened during such image processing procedures. Many morphologically motivated image processing algorithms are applied for that purpose. The proposed method is verified as successful in extracting DCFs and measuring thicknesses in experiment using two hundred 800 × 600 DICOM ultrasonography images with 98.5% extraction rate. Also, the thickness of DCFs automatically measured by this software has small difference (less than 0.3 cm) for 89.8% of extracted DCFs.

An Effective Color Quantization Method Using Octree-Based Self-Organizing Maps.

Color quantization is an essential technique in color image processing, which has been continuously researched. It is often used, in particular, as preprocessing for many applications. Self-Organizing Map (SOM) color quantization is one of the most effective methods. However, it is inefficient for obtaining accurate results when it performs quantization with too few colors. In this paper, we present a more effective color quantization algorithm that reduces the number of colors to a small number by using octree quantization. This generates more natural results with less difference from the original image. The proposed method is evaluated by comparing it with well-known quantization methods. The experimental results show that the proposed method is more effective than other methods when using a small number of colors to quantize the colors. Also, it takes only 71.73% of the processing time of the conventional SOM method.

Quantification of Hepatorenal Index for Computer-Aided Fatty Liver Classification with Self-Organizing Map and Fuzzy Stretching from Ultrasonography.

Accurate measures of liver fat content are essential for investigating hepatic steatosis. For a noninvasive inexpensive ultrasonographic analysis, it is necessary to validate the quantitative assessment of liver fat content so that fully automated reliable computer-aided software can assist medical practitioners without any operator subjectivity. In this study, we attempt to quantify the hepatorenal index difference between the liver and the kidney with respect to the multiple severity status of hepatic steatosis. In order to do this, a series of carefully designed image processing techniques, including fuzzy stretching and edge tracking, are applied to extract regions of interest. Then, an unsupervised neural learning algorithm, the self-organizing map, is designed to establish characteristic clusters from the image, and the distribution of the hepatorenal index values with respect to the different levels of the fatty liver status is experimentally verified to estimate the differences in the distribution of the hepatorenal index. Such findings will be useful in building reliable computer-aided diagnostic software if combined with a good set of other characteristic feature sets and powerful machine learning classifiers in the future.

Developing an Intelligent Automatic Appendix Extraction Method from Ultrasonography Based on Fuzzy ART and Image Processing.

Ultrasound examination (US) does a key role in the diagnosis and management of the patients with clinically suspected appendicitis which is the most common abdominal surgical emergency. Among the various sonographic findings of appendicitis, outer diameter of the appendix is most important. Therefore, clear delineation of the appendix on US images is essential. In this paper, we propose a new intelligent method to extract appendix automatically from abdominal sonographic images as a basic building block of developing such an intelligent tool for medical practitioners. Knowing that the appendix is located at the lower organ area below the bottom fascia line, we conduct a series of image processing techniques to find the fascia line correctly. And then we apply fuzzy ART learning algorithm to the organ area in order to extract appendix accurately. The experiment verifies that the proposed method is highly accurate (successful in 38 out of 40 cases) in extracting appendix.

Digital image analysis of endoscopic ultrasonography is helpful in diagnosing gastric mesenchymal tumors.

BACKGROUND: Endoscopic ultrasonography (EUS) is a valuable imaging tool for evaluating subepithelial lesions in the stomach. However, there are few studies on differentiation between gastrointestinal stromal tumors (GISTs) and benign mesenchymal tumors, such as leiomyoma or schwannoma, with the use of EUS. In addition, there are limitations in the analysis of the characteristic features of such tumors due to poor interobserver agreement as a result of subjective interpretation of EUS images. Therefore, the aim of this study was to evaluate the role of digital image analysis in distinguishing the features of GISTs from those of benign mesenchymal tumors on EUS. METHODS: We enrolled 65 patients with histopathologically proven gastric GIST, leiomyoma or schwannoma on surgically resected specimens who underwent EUS examination at our endoscopic unit from January 2007 to September 2010. After standardization of the EUS images, brightness values including the mean (Tmean), indicative of echogenicity, and the standard deviation (TSD), indicative of heterogeneity, in the tumors were analyzed. RESULTS: The Tmean and TSD were significantly higher in GIST than in leiomyoma and schwannoma (p < 0.001). However, there was no significant difference in the Tmean or TSD between benign and malignant GISTs. The sensitivity and specificity were almost optimized for differentiating GIST from leiomyoma or schwannoma when the critical values of Tmean and TSD were 65 and 75, respectively. The presence of at least 1 of these 2 findings in a given tumor resulted in a sensitivity of 94%, specificity of 80%, positive predictive value of 94%, negative predictive value of 80%, and accuracy of 90.8% for predicting GIST. CONCLUSIONS: Digital image analysis provides objective information on EUS images; thus, it can be useful in diagnosing gastric mesenchymal tumors.

Image analysis of endosocopic ultrasonography in submucosal tumor using fuzzy inference.

Endoscopists usually make a diagnosis in the submucosal tumor depending on the subjective evaluation about general images obtained by endoscopic ultrasonography. In this paper, we propose a method to extract areas of gastrointestinal stromal tumor (GIST) and lipoma automatically from the ultrasonic image to assist those specialists. We also propose an algorithm to differentiate GIST from non-GIST by fuzzy inference from such images after applying ROC curve with mean and standard deviation of brightness information. In experiments using real images that medical specialists use, we verify that our method is sufficiently helpful for such specialists for efficient classification of submucosal tumors.

Area extraction of the liver and hepatocellular carcinoma in CT scans.

In Korea, hepatocellular carcinoma is the third frequent cause of cancer death, occupying 17.2% among the whole deaths from cancer, and the rate of death from hepatocellular carcinoma comes to about 21 out of 100,000. This paper proposes an automatic method for the extraction of areas being suspicious as hepatocellular carcinoma from computed tomography (CT) scans and evaluates the availability as an auxiliary tool for the diagnosis of hepatocellular carcinoma. For detecting tumors in the internal of the liver from a CT scan, first, an area of the liver is extracted from about 45-50 CT slices obtained by scanning in 2.5-mm intervals starting from the lower part of the chest. In the extraction of an area of the liver, after the unconcerned areas outside of the bony thorax are removed, areas of the internal organs are segmented by using information on the intensity distribution of each organ, and an area of the liver is extracted among the segmented areas by using information on the position and morphology of the liver. Because hepatocellular carcinoma is a hypervascular tumor, the area corresponding to hepatocellular carcinoma appears more brightly than the surroundings in a CT scan, and also takes a spherical shape if the tumor shows expansile growth pattern. By using these features, areas being brighter than the surroundings and globe-shaped are segmented as candidate areas for hepatocellular carcinoma in the area of the liver, and then, areas appearing at the same position in successive CT slices among the candidates are discriminated as hepatocellular carcinoma. For the performance evaluation of the proposed method, experimental results obtained by applying the proposed method to CT scans were compared with the diagnoses by radiologists. The evaluation results showed that all areas of the liver and hypervascular tumors were extracted exactly and the proposed method has a high availability as an auxiliary diagnosis tool for the discrimination of liver tumors.

Analysis of endoscopic electronic image of intramucosal gastric carcinoma using a software program for calculating hemoglobin index.

Hemoglobin is the predominent pigment in the gastrointestinal mucosa, and the development of electronic endoscopy has made it possible to quantitatively measure the mucosal hemoglobin volume, by using a hemoglobin index (IHb). The aims of this study were to make a software program to calculate the IHb and then to investigate whether the mucosal IHb determined from the electronic endoscopic data is a useful marker for evaluating the color of intramucosal gastric carcinoma with regard to its value for discriminating between the histologic types. We made a software program for calculating the IHb in the endoscopic images. By using this program, the mean values of the IHb for the carcinoma (IHb-C) and those of the IHb for the surrounding non-cancerous mucosa (IHb-N) were calculated in 75 intestinal-type and 34 diffuse-type intramucosal gastric carcinomas. We then analyzed the ratio of the IHb-C to the IHb-N (C/N ratio). The C/N ratio in the intestinal-type carcinoma group was higher than that in the diffuse-type carcinoma group (p<0.001). In the diffuse-type carcinoma group, the C/N ratio in the body was lower than that in the antrum (p=0.022). The accuracy rate, sensitivity, specificity, and the positive and negative predictive values for the differential diagnosis of the diffuse-type carcinoma from the intestinal-type carcinoma were 94.5%, 94.1%, 94.7%, 88.9% and 97.3%, respectively. IHb is useful for making quantitative measurement of the endoscopic color in the intramucosal gastric carcinoma, and the C/N ratio by using the IHb would be helpful for distinguishing the diffuse-type carcinoma from the intestinal-type carcinoma.