AI-Aided Analysis of Capsule Endoscopy Images

Project Reference :


Institution :

National University of Singapore (NUS)

Principal Investigator :

Professor Guo Yongxin

Technology Readiness :

7 (System prototype demonstration in operational environment)

Technology Categories :

AI - Deep Learning

Background/Problem Statement

With the increasing global prevalence of gastrointestinal disorders, the rise in the geriatric population, and the preference for minimally invasive techniques by patients for diagnosis, the demand for capsule endoscopy is expected to grow to $1.2 billion by 2026.

But the process of detecting lesions or abnormalities from the images taken by the capsule endoscope is very tedious, time-consuming and error-prone. It takes about two hours for a doctor to read an image due to which the missed diagnosis rate could be high.



An intelligent AI platform comprising three deep learning networks:

  • A lesion classification network that can be used to classify vascular lesion images, inflammatory images, and normal images with more than 95% accuracy
  • A segmentation network that can be used to clarify the location and area of the lesions with an IoU of by more than 85%
  • A super-resolution network that enlarges the resolution to twice its original resolution, resulting in clearer images


  • Supports doctors to quickly and accurately read numerous images taken by capsule endoscopes
  • Improves the image quality taken by capsule endoscopes to provide doctors with a better user experience

Potential Application(s)

Hospitals and clinics that perform capsule endoscopy can use this solution to reduce the time taken for medical image analysis and improve accuracy of diagnosis.

We welcome interest from the industry for collaboration/ co-development / customisation of the technology into a new product or service. If you have any enquiries or are keen to collaborate, please contact us.