Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Learning for Improved Polyp Detection from Synthetic Narrow-Band Imaging

Mathias Ramm Haugland, Hemin Ali Qadir, Ilangko Balasingham

TL;DR

This paper tackles the challenge of improving automatic polyp detection when NBI hardware is unavailable by translating white-light colonoscopy images into synthetic NBI (SNBI) using a CycleGAN. A dual-network pipeline pairs unpaired WLI↔SNBI translation with EfficientDet-D0-based detection, trained under semi-paired conditions to align modalities. Results show that real NBI yields better detection than WLI, and SNBI enhances WLI-based detection, with still-image tests confirming gains across polyp types; SNBI performance closely approaches that of real NBI. The work demonstrates a practical pre-processing approach to boost CRC screening workflows without additional equipment, with future work aimed at extending to NICE-based polyp classification and clinical impact assessments.

Abstract

To cope with the growing prevalence of colorectal cancer (CRC), screening programs for polyp detection and removal have proven their usefulness. Colonoscopy is considered the best-performing procedure for CRC screening. To ease the examination, deep learning based methods for automatic polyp detection have been developed for conventional white-light imaging (WLI). Compared with WLI, narrow-band imaging (NBI) can improve polyp classification during colonoscopy but requires special equipment. We propose a CycleGAN-based framework to convert images captured with regular WLI to synthetic NBI (SNBI) as a pre-processing method for improving object detection on WLI when NBI is unavailable. This paper first shows that better results for polyp detection can be achieved on NBI compared to a relatively similar dataset of WLI. Secondly, experimental results demonstrate that our proposed modality translation can achieve improved polyp detection on SNBI images generated from WLI compared to the original WLI. This is because our WLI-to-SNBI translation model can enhance the observation of polyp surface patterns in the generated SNBI images.

Deep Learning for Improved Polyp Detection from Synthetic Narrow-Band Imaging

TL;DR

This paper tackles the challenge of improving automatic polyp detection when NBI hardware is unavailable by translating white-light colonoscopy images into synthetic NBI (SNBI) using a CycleGAN. A dual-network pipeline pairs unpaired WLI↔SNBI translation with EfficientDet-D0-based detection, trained under semi-paired conditions to align modalities. Results show that real NBI yields better detection than WLI, and SNBI enhances WLI-based detection, with still-image tests confirming gains across polyp types; SNBI performance closely approaches that of real NBI. The work demonstrates a practical pre-processing approach to boost CRC screening workflows without additional equipment, with future work aimed at extending to NICE-based polyp classification and clinical impact assessments.

Abstract

To cope with the growing prevalence of colorectal cancer (CRC), screening programs for polyp detection and removal have proven their usefulness. Colonoscopy is considered the best-performing procedure for CRC screening. To ease the examination, deep learning based methods for automatic polyp detection have been developed for conventional white-light imaging (WLI). Compared with WLI, narrow-band imaging (NBI) can improve polyp classification during colonoscopy but requires special equipment. We propose a CycleGAN-based framework to convert images captured with regular WLI to synthetic NBI (SNBI) as a pre-processing method for improving object detection on WLI when NBI is unavailable. This paper first shows that better results for polyp detection can be achieved on NBI compared to a relatively similar dataset of WLI. Secondly, experimental results demonstrate that our proposed modality translation can achieve improved polyp detection on SNBI images generated from WLI compared to the original WLI. This is because our WLI-to-SNBI translation model can enhance the observation of polyp surface patterns in the generated SNBI images.
Paper Structure (9 sections, 3 equations, 3 figures, 2 tables)

This paper contains 9 sections, 3 equations, 3 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. MATERIALS AND METHODS
  3. Method
  4. Datasets
  5. Evaluation metrics
  6. Results and Discussion
  7. Results on videos
  8. Results on still images
  9. CONCLUSION AND FUTURE WORK

Figures (3)

  • Figure 1: The proposed framework. CycleGAN for modality translation (left), followed by EfficientDet-D0 for polyp detection (right).
  • Figure 2: The same polyp (from the PICCOLO set) is shown in (a) original WLI, (b) original NBI, and (c) SNBI generated from (a) by our CycleGAN.
  • Figure 3: An output detection shows that the detection model misses a hyperplastic polyp in WLI but detects it in SNBI.