Sequential Models for Endoluminal Image Classification.

Joana Reuss,Hagen Wenzek,Santi Seguí,Guillem Pascual

doi:10.3390/diagnostics12020501

Abstract

Wireless Capsule Endoscopy (WCE) is a procedure to examine the human digestive system for potential mucosal polyps, tumours, or bleedings using an encapsulated camera. This work focuses on polyp detection within WCE videos through Machine Learning. When using Machine Learning in the medical field, scarce and unbalanced datasets often make it hard to receive a satisfying performance. We claim that using Sequential Models in order to take the temporal nature of the data into account improves the performance of previous approaches. Thus, we present a bidirectional Long Short-Term Memory Network (BLSTM), a sequential network that is particularly designed for temporal data. We find the BLSTM Network outperforms non-sequential architectures and other previous models, receiving a final Area under the Curve of . Experiments show that our method of extracting spatial and temporal features yields better performance and could be a possible method to decrease the time needed by physicians to analyse the video material.

Highlights

We used retrospective Wireless Capsule Endoscopy (Medtronic PillCam COLON2) data from 110 patients were conducted on behalf of the NHS Highland Raigmore Hospital in Inverness
Patients with hemorrhoidal bleeding were excluded from this study, as well as all that met standard exclusion criteria for Colon Capsule Endoscopy (CCE), especially wearers of heart rate monitors and patients with diabetes
To further explore the performance of the bidirectional Long Short-Term Memory Network (BLSTM), we look at images and their ground truth and predicted labels to analyse whether the system makes meaningful decisions

Summary

Introduction

The most common method for this purpose is endoscopy In this diagnostic procedure, an endoscope (a long tube with a camera attached to it) gets inserted into the patient’s body to allow physicians to inspect the gastrointestinal (GI) system. An endoscope (a long tube with a camera attached to it) gets inserted into the patient’s body to allow physicians to inspect the gastrointestinal (GI) system This often causes patient discomfort and requires substantial time, medical staff, and infrastructure. While moving through the patient’s GI system, the camera takes up to 400,000 images [4] to capture abnormalities While this alleviates the discomfort patients experience during endoscopy, it comes at the expense of additional post-analysis requirements. Computer-aided detection methods have received increased attention aiming to identify GI abnormalities [5]. These systems usually serve as decision support systems to point the physician at the most conspicuous images that have a high chance of showing a polyp [2]

Methods

Results

Discussion

Conclusion