Development of Robust AI Based X-Ray Analysis WorkFlow Based on Real-Time out of Distribution Detection

Szabó, Lóránt and Weltsch, Zoltán (2024) Development of Robust AI Based X-Ray Analysis WorkFlow Based on Real-Time out of Distribution Detection. MŰSZAKI TUDOMÁNYOS KÖZLEMÉNYEK (EN), 20. pp. 74-77. ISSN 2601-5773

Preview

Text MTK20-EN-14-SzaboL.pdf - Published Version Download (1MB) | Preview

Abstract

In medicine, the role of artificial intelligence (AI) is becoming increasingly common. An important example here is the application of AI in X-Ray analysis, as a known aspect of medical imaging and finding-detection. However, the effectiveness of AI image analysis may be challenging due to the out-of-distribution (OOD) re cords, i.e. data that significantly differ from the data set used to train the model. These OOD data may result from symptoms, that the model is not prepared for, or even from unpredictable tool behaviour, environmen tal changes or new errors that have not occurred during the data-gathering phase. This paper shows that with proper OOD analysis the AI-based tool may be prepared for handling “unknown” input data.

Actions (login required)

Edit Item