Five simultaneous artificial intelligence data challenges on ultrasound, CT, and MRI.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Data Science

Background:

• The study addresses the need for structured data dynamics in medical imaging.
• It aims to integrate General Data Protection Regulation (GDPR) compliance into radiologic practices.
• Building a collaborative network among diverse stakeholders is crucial for advancing medical AI.

Purpose of the Study:

• To educate radiologists on GDPR regulations.
• To establish a prospective, multicentric database of ultrasound, CT, and MRI images.
• To foster collaboration between radiologists, researchers, industry, and students through data challenges.

Main Methods:

• Clinical questions were defined by the Société Francaise de Radiologie.
• The challenge adhered to French ethical and data protection standards.
• Multidisciplinary teams were formed, including radiologists, engineering students, and industry/research representatives.

Main Results:

• Five challenges were conducted: MRI meniscal tears, CT renal cortex segmentation, ultrasound liver lesions, MRI breast lesions, and CT thyroid lesions.
• 46 radiology services contributed 5,170 images over 4 months.
• Three challenges (meniscal tears, renal cortex, liver lesions) achieved >90% accuracy; breast lesions reached 82%, thyroid lesions 70%.

Conclusions:

• The data challenges rapidly convened a large community of professionals and students.
• High accuracy in three modalities indicates AI's significant promise in medical imaging.
• The initiative highlights the potential benefits and challenges of AI in radiology.