Mining Multi-Center Heterogeneous Medical Data with Distributed Synthetic Learning

Abstract

Overcoming barriers of multi-center data analysis is challenging due to privacy protection and data heterogeneity in the healthcare system. In this study, we propose Distributed Synthetic Learning (DSL) architecture to learn across multi-medical centers without leaking sensitive personal information. DSL emphasizes the building of a homogeneous data center with entirely synthetic medical images via a form of GAN-based synthetic learning. In particular, DSL architecture is extensible with three key variances: multi-modality learning, missing modality completion learning, and continuous learning over time. We systematically evaluate the performance of DSL on different medical applications using cardiac computed tomography angiography (CTA), brain tumor MRI, and histopathology nuclei datasets. Extensive experiments demonstrate the superior performance of DSL as a high-quality synthetic medical image provider for various image tasks from an ideal synthetic image quality metric called Dist-FID. We show that our model can be adapted to heterogeneous data and remarkably outperforms the real misaligned modalities segmentation model by 55% and the temporal datasets segmentation model by 8%. The proposed DSL framework demonstrates its potential for integrating multi-center heterogeneous data to support downstream clinical decision making.

Multimodality Cardiovascular Imaging for Totally Video-Guided Thorascopic Cardiac Surgery

“…The clinical practice of TVTCS has benefited not only from individual imaging technical progress, but also from fusion technology from multiple imaging modes, including computational modeling and mobile detection modes [ 77 ]. Overcoming the barriers stemming from data heterogeneity among the sorts of detection instruments to establish the common standard for clinicians is a clinical challenge [ 78 ]. A realistic stereoscopically anatomical model from the visible imaging datasets permits the developments of custom-tailored procedure strategies.…”

Expertise-informed Generative AI Enables Ultra-High Data Efficiency for Building Generalist Medical Foundation Model

“…The acquisition of massive medical data is extremely expensive and time-consuming, especially in fields requiring high-precision equipment, such as MRI imaging, and long-term patient tracking, such as oncology [7] and neurodegenerative diseases [8]. In addition, the large-scale pre-training medical data are typically collected from multi-center into a centralized institution, which significantly increases the risk of exposing patient privacy as they have access to a rich set of personal patient information and routine data anonymization can not guarantee data privacy protection [9][10][11]. Given these challenges, it is critical to develop a new pre-training paradigm with high data efficiency for building medical foundation models from limited real-world pre-training datasets, which can effectively mitigate the issues of data scarcity, extensive resource requirements, and privacy concerns that currently impede the development of medical foundation models.…”