Reducing Data Transmission Efficiency in Wireless Capsule Endoscopy through DL-CEndo Framework: Reconstructing Lossy Low-Resolution Luma Images and Improving Summarization
Abderrahmane Salmi,Wei Zhang,Feng Jiang
DOI: https://doi.org/10.1007/s11036-024-02334-8
2024-05-23
Mobile Networks and Applications
Abstract:Wireless Capsule Endoscopy (WCE) is the preferred technology for the diagnosis and evaluation of small bowel disorders. The concluding WCE video comprises as many as 150,000 images. Within this extensive collection of images, the majority are classified as normal , with only a singular or a few frames identified as pathological. The transmission of such a substantial volume of data from the WCE is constrained by the device's limited battery capacity. The primary objective of this study is to minimize the computational load by shifting the maximum of computations at the decoder side while reducing the information transmitted from the encoder. This paper presents three primary contributions. First, a new Capsule Endoscopy framework leveraging Deep Learning (DL-CEndo) is proposed for the WCE compression system. This framework aims to significantly decrease the transmission of unnecessary data by sending Low-Resolution (LR) luma images, thereby conserving energy in the WCE device. Second, a new colorization model named EndoColorGAN diffusion-based is implemented to reconstruct the colors of unnecessary WCE images. Lastly, the paper introduces a demo for fast and smart summarization, easing the review process for physicians by focusing on essential images within the WCE video. The DL-CEndo framework achieved a high compression ratio of 93.20% with a reconstruction NiQE value of 4.63 and ensures fast diagnostics for physicians. Our demo is available at https://capsule-colorization-framework.vercel.app/.
computer science, information systems,telecommunications, hardware & architecture