Human and computer attention in assessing genetic conditions

Area of Science:

• Biomedical imaging
• Medical genetics
• Artificial intelligence in healthcare

Background:

• Artificial intelligence (AI) and deep learning (DL) are increasingly applied in biomedical fields, particularly in genetics for diagnostic image evaluation.
• Understanding how AI image assessment compares to human visual perception is crucial for integrating AI into clinical practice.

Approach:

• This study employed eye-tracking analysis on geneticist clinicians and non-clinicians evaluating medical images.
• The human visual attention data was then compared against deep learning (DL)-based saliency maps.

Key Points:

• Human visual attention patterns during image assessment significantly differ from the regions prioritized by DL models.
• Distinct image inspection patterns were observed among individuals, with notable differences between clinicians and non-clinicians.

Conclusions:

• AI and human visual perception in medical genetics image analysis are not aligned.
• Further research is needed to bridge the gap between AI and human expertise for reliable diagnostic tools.