Ensemble of Deep Convolutional Neural Networks for Classification of Early Barrett’s Neoplasia Using Volumetric Laser Endomicroscopy

Abstract

Barrett’s esopaghagus (BE) is a known precursor of esophageal adenocarcinoma (EAC). Patients with BE undergo regular surveillance to early detect stages of EAC. Volumetric laser endomicroscopy (VLE) is a novel technology incorporating a second-generation form of optical coherence tomography and is capable of imaging the inner tissue layers of the esophagus over a 6 cm length scan. However, interpretation of full VLE scans is still a challenge for human observers. In this work, we train an ensemble of deep convolutional neural networks to detect neoplasia in 45 BE patients, using a dataset of images acquired with VLE in a multi-center study. We achieve an area under the receiver operating characteristic curve (AUC) of 0.96 on the unseen test dataset and we compare our results with previous work done with VLE analysis, where only AUC of 0.90 was achieved via cross-validation on 18 BE patients. Our method for detecting neoplasia in BE patients facilitates future advances on patient treatment and provides clinicians with new assisting solutions to process and better understand VLE data.

Accuracy of artificial intelligence‐assisted detection of esophageal cancer and neoplasms on endoscopic images: A systematic review and meta‐analysis

“…Fifty-two potentially relevant articles were then retrieved for further review, of which 36 were excluded due to insufficient data (n = 18), not analyzed based on endoscopic image (n = 4), or animal studies (n = 14). Finally, a total of 16 full-text manuscripts [12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27] were enrolled for the meta-analysis. The flowchart of study enrollment is shown in Figure 1.…”

“…The results of QUADAS-2 showed that the risk for patient selection was unclear in six studies, 14,20,22,[25][26][27] as shown in Figure 2, and the methodological quality was generally high.…”

“…Among the 16 studies endoscopic images or video clips were collected retrospectively as the test dataset in 13 studies [12][13][14][15][16][17][18][19]21,[23][24][25]27 and prospectively in three studies. 20,22,26 Only two studies 20,27 used AI on real patients with non-dysplastic BE and confirmed Barrett's neoplasia. Most studies were conducted in Asian populations, [12][13][14][15][16][17][18][23][24][25] while in the other six studies [19][20][21][22]26,27 endoscopic images were obtained from Western populations.…”

“…20,22,26 Only two studies 20,27 used AI on real patients with non-dysplastic BE and confirmed Barrett's neoplasia. Most studies were conducted in Asian populations, [12][13][14][15][16][17][18][23][24][25] while in the other six studies [19][20][21][22]26,27 endoscopic images were obtained from Western populations.…”

The Role and Impact of Deep Learning Methods in Computer-Aided Diagnosis Using Gastrointestinal Endoscopy

“…A major advantage was that the model was quick enough to realize real-time detection and diagnosis. Fonollà et al [ 38 ] also discovered that a brief but deep model was the most promising strategy to identify a volumetric laser endomicroscopy (VLE) RoI between nondysplastic BE (NDBE) and high-grade dysplasia (HGD) because of the unbalanced property of the dataset. Thus, they chose to use a combination of horizontal flip, motion blur, and optical grid distortion of three DCNNs, each of them based on the VGG16 network, in order to find neoplasia using a valid VLE image dataset.…”