orGAN: A Synthetic Data Augmentation Pipeline for Simultaneous Generation of Surgical Images and Ground Truth Labels

2025-06-17

AI Summary

High-quality annotated intraoperative hemorrhage data are scarce due to high annotation costs, ethical constraints, and clinical access limitations. Method: We propose the first end-to-end synthetic data generation framework—termed orGAN—that jointly synthesizes hemorrhage images and corresponding pixel-level coordinate labels using a StyleGAN-based architecture. To ensure anatomical plausibility, we introduce relational positional learning to model spatially coherent hemorrhage patterns; additionally, we integrate a LaMa inpainting module to reconstruct hemorrhage-free baseline images, enabling precise ground-truth supervision. The framework requires only a small set of biomimetic organ images for training. Results: When combined with limited real data, orGAN-synthesized data boost hemorrhage detection accuracy to 90% and frame-level classification accuracy to 99% in surgical settings—significantly reducing reliance on scarce real annotations and establishing a new paradigm for low-resource medical imaging tasks.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Web data generation and simulation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Deep learning in medical imaging faces obstacles: limited data diversity, ethical issues, high acquisition costs, and the need for precise annotations. Bleeding detection and localization during surgery is especially challenging due to the scarcity of high-quality datasets that reflect real surgical scenarios. We propose orGAN, a GAN-based system for generating high-fidelity, annotated surgical images of bleeding. By leveraging small"mimicking organ"datasets, synthetic models that replicate tissue properties and bleeding, our approach reduces ethical concerns and data-collection costs. orGAN builds on StyleGAN with Relational Positional Learning to simulate bleeding events realistically and mark bleeding coordinates. A LaMa-based inpainting module then restores clean, pre-bleed visuals, enabling precise pixel-level annotations. In evaluations, a balanced dataset of orGAN and mimicking-organ images achieved 90% detection accuracy in surgical settings and up to 99% frame-level accuracy. While our development data lack diverse organ morphologies and contain intraoperative artifacts, orGAN markedly advances ethical, efficient, and cost-effective creation of realistic annotated bleeding datasets, supporting broader integration of AI in surgical practice.

Problem

• Generates synthetic surgical images with bleeding annotations
• Addresses limited data diversity and ethical concerns
• Improves bleeding detection accuracy in surgical settings

Innovation

• GAN-based system generates annotated surgical images
• StyleGAN with Relational Positional Learning simulates bleeding
• LaMa-based inpainting module restores pre-bleed visuals