Artificial Intelligence in Healthcare Competition (TEKNOFEST-2021): Stroke Data Set

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Healthcare Technology

Background:

• The first artificial intelligence (AI) competition in healthcare was held at the TEKNOFEST festival in Istanbul, Türkiye, in September 2021.
• This event focused on AI applications in medical diagnostics, specifically stroke detection using imaging data.

Purpose of the Study:

• To detail the preparation and processes of an AI competition focused on healthcare.
• To provide a valuable, anonymized, and annotated dataset for further AI research in stroke detection.

Main Methods:

• A dataset from 2019-2020 was collected from the Republic of Türkiye's Ministry of Health e-Pulse and Teleradiology System.
• The data was anonymized, curated, and annotated by 7 radiologists, then shared with participating teams under a non-disclosure agreement.
• The competition involved two stages: classifying 192 images into stroke/non-stroke categories and then classifying 97 images with hemorrhage, ischemia, or both.

Main Results:

• Teams utilized various AI methods, with Unet and DeepLabv3 being the most common.
• The competition facilitated the classification of medical images for stroke detection.

Conclusions:

• AI competitions in healthcare are effective for generating diverse and valuable annotated datasets.
• Expert-annotated data is crucial for advancing AI in medical problem-solving and diagnostics.