Use of a convolutional neural network for classifying microvessels of superficial esophageal squamous cell carcinomas

Abstract

Background and Aim:The morphological diagnosis of microvessels on the surface of superficial esophageal squamous cell carcinomas using magnifying endoscopy with narrow-band imaging is widely used in clinical practice. Nevertheless, inconsistency, even among experts, remains a problem. We constructed a convolutional neural network-based computer-aided diagnosis system to classify the microvessels of superficial esophageal squamous cell carcinomas and evaluated its diagnostic performance. Methods: In this retrospective study, a cropped magnifying endoscopy with narrow-band images from superficial esophageal squamous cell carcinoma lesions was used as the dataset. All images were assessed by three experts, and classified into three classes, Type B1, B2, and B3, based on the Japan Esophagus Society classification. The dataset was divided into training and validation datasets. A convolutional neural network model (ResNeXt-101) was trained and tuned with the training dataset. To evaluate diagnostic accuracy, the validation dataset was assessed by the computer-aided diagnosis system and eight endoscopists. Results: In total, 1777 and 747 cropped images (total, 393 lesions) were included in the training and validation datasets, respectively. The diagnosis system took 20.3 s to evaluate the 747 images in the validation dataset. The microvessel classification accuracy of the computer-aided diagnosis system was 84.2%, which was higher than the average of the eight endoscopists (77.8%, P < 0.001). The area under the receiver operating characteristic curves for diagnosing Type B1, B2, and B3 vessels were 0.969, 0.948, and 0.973, respectively. Conclusions:The computer-aided diagnosis system showed remarkable performance in the classification of microvessels on superficial esophageal squamous cell carcinomas.

Systematic review with meta‐analysis: artificial intelligence in the diagnosis of oesophageal diseases

“…From these we identified 67 separate articles that appeared to be relevant to the study question. In total, 42 studies 17‐58 reported on the performance of AI in the diagnosis of various ODs and were included in the qualitative synthesis (Supplementary Table S1). Among the included studies, 19 17‐35 reported complete data for extraction and were included in the meta‐analysis: 9 on BN, 17‐25 5 on OSCC, 26‐30 2 on abnormal IPCLs 31,32 and 3 on GERD 33‐35 (Figure 2).…”

“…From these we identified 67 separate articles that appeared to be relevant to the study question. In total, 42 studies 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 reported on the performance of AI in the diagnosis of various ODs and were included in the qualitative synthesis (Supplementary Table S1 ). Among the included studies, 19 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 ,…”

“…Among the 42 studies included in the qualitative synthesis, 23 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 could not be included in the meta‐analysis for various reasons. They also included investigations on eosinophilic oesophagitis (n = 1), reflux monitoring (n = 1), optical endoscopic diagnosis of GERD (n = 1), motility assessment (n = 1), diagnosis of cytomegalovirus and herpes simplex virus oesophagitis (n = 1) and varices (n = 1).…”

Applications of Artificial Intelligence for the Diagnosis of Gastrointestinal Diseases

“…Dedicated models analyzing WLI and NBI images have shown high disease-specific diagnostic accuracy not only for BE and EAC [ 4 ] but also for esophageal squamous cell carcinoma (ESCC) [ 5 ]. Dedicated algorithms have also been implemented to analyze enhanced endoscopy imaging, allowing to evaluate disease-specific mucosal and vascular patterns [ 6 ], the presence of submucosal invasion [ 7 ], the depth of invasion [ 8 ], and microendoscopy use for both ESCC [ 9 ] and BE [ 10 ]. Moreover, a recent meta-analysis has shown an overall high accuracy in the detection of early EAC, with a significantly better performance compared to endoscopists in terms of the pooled sensitivity (0.94 vs. 0.82, p = 0.01).…”

A systematic review on application of deep learning in digestive system image processing

“…However, NBI is beneficial to enhance histological diagnostic grading accuracy [ 6 ]. Moreover, NBI can enhance the ability to differentiate squamous cell carcinoma microvessels [ 20 ]. A multi-center study shows that magnifying endoscopy narrow-band imaging (ME-NBI) reached senior endoscopic physicians’ predictive performance in early gastric cancer.…”