Identification of Barrett's esophagus in endoscopic images using deep learning

Abstract

Background Development of a deep learning method to identify Barrett's esophagus (BE) scopes in endoscopic images. Methods 443 endoscopic images from 187 patients of BE were included in this study. The gastroesophageal junction (GEJ) and squamous-columnar junction (SCJ) of BE were manually annotated in endoscopic images by experts. Fully convolutional neural networks (FCN) were developed to automatically identify the BE scopes in endoscopic images. The networks were trained and evaluated in two separate image sets. The performance of segmentation was evaluated by intersection over union (IOU). Results The deep learning method was proved to be satisfying in the automated identification of BE in endoscopic images. The values of the IOU were 0.56 (GEJ) and 0.82 (SCJ), respectively. Conclusions Deep learning algorithm is promising with accuracies of concordance with manual human assessment in segmentation of the BE scope in endoscopic images. This automated recognition method helps clinicians to locate and recognize the scopes of BE in endoscopic examinations.

Automatic esophagus Z‐line delineation in endoscopic images using a new boundary linking method

“…For annotation evaluation, only having the overlap of the segmented neoplastic images with the ground truth is considered in [14]. The averages of intersection over union for GEJ and SCJ are 0.56 and 0.86, respectively, and these values for the Dice coefficient are 0.71 and 0.90 in [21]. The average of the reported dice coefficient and Jaccard index of the EAC annotation compared to the five experts in [28] are 0.448 and 0.307, respectively.…”

“…Consequently, this method can be used when there are remarkable changes in the texture or colour of the tissue. The gastroesophageal junction (GEJ) and squamous-columnar junction (SCJ) of BE are segmented by the proposed multi-layer fully convolutional networks [21]. For the annotation of GEJ and SCJ, two separate, fully connected networks are used.…”

“…[14] ✓ Segmentation is performed on the neoplastic high resolution images ✓ Only the overlap with the ground truth is used for the annotation evaluation [16] ✓ Annotation is done for the dysplastic areas of high resolution frames ✓ The dice coefficient of segmentation with only one expert is reported [18] ✓ The Lesion-based method, the method works when changes in tissue colour or texture have occurred [21] ✓ A large dataset is needed ✓ It is useful for high quality images [28] ✓ EAC is segmented, so the method works when tissue colour or texture changes have occurred.…”

“…The classification category methods aim to classify between the images with normal and abnormal regions and detect the type of the disorder [11][12][13][14][15][16]. These works have employed either deep learning tools [11][12][13][14][15][16][17][18][19][20][21][22][23] or conventional image processing techniques [22][23][24][25][26][27]. ROI detection is achieved by automatically determining the region with abnormal tissue by introducing the bounding box [18,[24][25][26][27].…”

The investigation of construction and clinical application of image recognition technology assisted bronchoscopy diagnostic model of lung cancer

“…IRT has achieved important results in assisting the diagnosis of breast cancer, skin cancer, diabetic retinopathy and other diseases ( 9 – 11 ). Its use in assisting medical endoscopy in recognition of lesions is currently mainly applied in digestive endoscopy, such as the identification of Barrett’s esophagus, early gastric cancer, differentiation of colon polyps and etc ( 12 – 14 ). It has been verified to have higher efficiency, sensitivity and specificity than ordinary endoscopists.…”