The International Radiomics Platform - An Initiative of the German and Austrian Radiological Societies - First...

Area of Science:

• Radiomics and Artificial Intelligence in Medical Imaging
• Cardiovascular Imaging and Diagnostics
• Public-Private Partnerships in Research Platforms

Background:

• The DRG-ÖRG IRP is a web/cloud-based platform fostering collaboration in radiomics.
• It supports data sharing, annotation, validation, and certification for AI and integrated diagnostics.
• Public-private partnerships are crucial for developing advanced research infrastructures.

Purpose of the Study:

• To evaluate the DRG-ÖRG IRP for automated myocardial segmentation and late gadolinium enhancement (LGE) detection.
• To assess the platform's utility in a proof-of-concept study using myocarditis datasets.
• To demonstrate the integration of AI methods within a collaborative radiomics framework.

Main Methods:

• Utilized a pre-trained neural network for myocardial segmentation of cardiac CINE and PSIR datasets.
• Applied radiomic features for segment-based detection of myocardial LGE.
• Leveraged the DRG-ÖRG IRP for multicenter data provision, quality assurance, and analysis.

Main Results:

• The neural network achieved a Dice coefficient of 0.813 for myocardial segmentation.
• Segment-based LGE detection yielded an AUC of 0.73, improving to 0.79 with uncertain segments excluded.
• The platform demonstrated advantages in data transparency, reliability, and collaborative evolution.

Conclusions:

• The DRG-ÖRG IRP serves as a valuable hub for generating and executing interdisciplinary radiomics projects.
• The platform simplifies quantitative AI analyses by enabling integration of pre-trained networks and scientific networking.
• Successful proof-of-concept demonstrates the platform's capability to implement strategic goals and foster collaborative research.