Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Self-Supervised Ultrasound Screen Detection

Alberto Gomez, Jorge Oliveira, Ramon Casero, Agis Chartsias

TL;DR

The paper tackles the bottleneck of accessing ultrasound data by proposing a self-supervised pipeline that detects US screen content in photographs, corrects perspective, and reconstructs usable frames for downstream analysis. It introduces synthetic, self-annotated data generation and a multi-task CNN that localizes screen corners and predicts screen presence, coupled with a homography-based rectification and basic post-processing. The approach is evaluated on synthetic and real data, showing strong screen-detection performance and reasonable image reconstruction, though real-world performance declines due to reflections and labeling ambiguities. Overall, the method enables rapid testing and prototyping of US-image analysis without hardware changes, potentially accelerating development of real-time imaging workflows.

Abstract

Ultrasound (US) machines display images on a built-in monitor, but routine transfer to hospital systems relies on DICOM. We propose a self-supervised pipeline to extract the US image from a photograph of the monitor. This removes the DICOM bottleneck and enables rapid testing and prototyping of new algorithms. In a proof-of-concept study, the rectified images retained enough visual fidelity to classify cardiac views with a balanced accuracy of 0.79 with respect to the native DICOMs.

Self-Supervised Ultrasound Screen Detection

TL;DR

The paper tackles the bottleneck of accessing ultrasound data by proposing a self-supervised pipeline that detects US screen content in photographs, corrects perspective, and reconstructs usable frames for downstream analysis. It introduces synthetic, self-annotated data generation and a multi-task CNN that localizes screen corners and predicts screen presence, coupled with a homography-based rectification and basic post-processing. The approach is evaluated on synthetic and real data, showing strong screen-detection performance and reasonable image reconstruction, though real-world performance declines due to reflections and labeling ambiguities. Overall, the method enables rapid testing and prototyping of US-image analysis without hardware changes, potentially accelerating development of real-time imaging workflows.

Abstract

Ultrasound (US) machines display images on a built-in monitor, but routine transfer to hospital systems relies on DICOM. We propose a self-supervised pipeline to extract the US image from a photograph of the monitor. This removes the DICOM bottleneck and enables rapid testing and prototyping of new algorithms. In a proof-of-concept study, the rectified images retained enough visual fidelity to classify cardiac views with a balanced accuracy of 0.79 with respect to the native DICOMs.

Paper Structure

This paper contains 15 sections, 2 equations, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Contributions
  4. Materials and methods
  5. Datasets
  6. Synthetic data generation
  7. Screen detection model design and training
  8. Geometric correction and post-processing
  9. Experiments and results
  10. Evaluation of screen detection and localization
  11. Image quality assessment
  12. Evaluation in echo view classification
  13. Conclusion
  14. Compliance with Ethical Standards
  15. Conflicts of Interest

Figures (7)

  • Figure 1: Steps in the creation of a synthetic image showing an echo screen with realistic reflection artifacts.
  • Figure 2: Detection
  • Figure 3: Reference
  • Figure 4: Reconstruction
  • Figure 6: Examples of original and reconstructed images, and matching MSE and SSIM values. (a), (b) are from the synthetic dataset, and (c), (d) are from the real dataset.
  • ...and 2 more figures