# Intelligent DRL-Based Adaptive Region of Interest for Delay-sensitive Telemedicine Applications

**Authors:** Abdulrahman Soliman, Amr Mohamed, Elias Yaacoub, Nikhil V. Navkar, Aiman Erbad  
**arXiv:** [2310.05099](https://arxiv.org/abs/2310.05099)  
**Date:** 2023-10-08  
**Summary page:** https://www.summarizepaper.com/en/paper/2310.05099v1/

## Key points

- Telemedicine has gained increased interest and potential, especially post-COVID-19.
- Remote experiences allow for complex surgeries and knowledge sharing without the need for travel.
- Video streaming delays are a significant challenge in telemedicine applications despite internet speed advancements.
- Techniques like image compression and ROI adjustments are used to reduce data size and transmission requirements.
- A recent study introduced a DRL model that dynamically adjusts ROI size and non-ROI quality based on estimated throughput.
- The DRL model reduced delay by 13% while maintaining overall quality within an acceptable range.
- Previous works focused on reducing delay or compressing frames, but visible latency hindered real-time transmission in practical telemedical settings.
- An approach using low-complexity image compression with region-based coding and IWT successfully reconstructed original frames with optimal quality for efficient picture transmission in telemedicine applications.

## Summary

The field of telemedicine has seen a surge in interest and potential, particularly in the wake of the COVID-19 pandemic. With remote experiences becoming more prevalent, individuals can now undergo complex surgeries or share knowledge with local surgeons without the need to travel abroad. Despite advancements in internet speeds, video streaming delays remain a significant challenge for telemedicine applications. To address this issue, techniques such as image compression and region of interest (ROI) adjustments are utilized to reduce data size and transmission requirements. In a recent study titled "Intelligent DRL-Based Adaptive Region of Interest for Delay-sensitive Telemedicine Applications," researchers proposed a Deep Reinforcement Learning (DRL) model that dynamically adjusts the ROI size and non-ROI quality based on estimated throughput. By comparing delay and structural similarity index measure (SSIM), the researchers found that the DRL model was able to reduce delay by 13% while maintaining overall quality within an acceptable range. This significant reduction in latency represents a valuable enhancement for telemedicine applications. Previous works have focused on reducing delay or compressing frames, but visible latency has hindered real-time transmission in practical telemedical settings. One notable approach introduced a low-complexity image compression method for Digital Imaging and Communications in Medicine (DICOM). By leveraging region-based coding combined with Integer Wavelet Transform (IWT), this technique successfully reconstructed original frames with optimal quality, facilitating efficient picture transmission within bandwidth constraints specific to telemedicine applications. Overall, this research underscores the importance of intelligent adaptive strategies like DRL in optimizing telemedicine processes by minimizing delays and ensuring high-quality transmission. These findings offer promising advancements for enhancing the efficiency and effectiveness of remote medical procedures and consultations.