Eye Tracking in Medical Image Interpretation

Eye tracking has become an invaluable technique in the study of medical image interpretation, offering quantitative insights into the interplay between visual perception and diagnostic decision-making. By recording where, when and how the eye moves across radiological images, researchers can distinguish between expert and novice search patterns, identify moments of uncertainty or omission, and assess the usability of decision-support tools. Applications range from mapping global-to-focal search strategies in chest radiographs and CT volumes to evaluating the integration of artificial intelligence in routine workflows. Beyond characterising perceptual expertise, eye‐tracking metrics are now used to guide the design of educational interventions, optimise interface layouts and predict potential diagnostic errors before they occur. The global significance of this work lies in its capacity to enhance diagnostic accuracy, reduce interpretation times and ultimately improve patient outcomes through a deeper understanding of the visual and cognitive processes that underpin medical image reading.

Research from Nature Portfolio

A seminal investigation demonstrated that experienced radiologists can extract a rapid “gist” of malignancy from mammograms within a half-second exposure, distinguishing future cancer cases from normal screens at above-chance levels even before lesions are overtly visible. This global impression varied with reader expertise and breast density, suggesting that implicit visual signatures precede explicit pathology. The study’s findings point to potential strategies for early risk stratification and tailored screening intervals, and they underscore the role of holistic processing—where parafoveal and peripheral cues are integrated to inform subsequent detailed search.

Eye Tracking in Medical Image Interpretation publication trend

The graph below shows the total number of articles in eye tracking in medical image interpretation across all publications each year (not limited to Nature Index journals).

Technical terms

Fixation: A period during which the eye remains stationary, typically indicating focused attention on a specific region of an image.

Saccade: A rapid eye movement between fixations that allows the viewer to scan different parts of an image.

Fixation mask: A spatial map derived from aggregated fixation points, used to highlight areas of concentrated visual attention.

Region of interest (ROI): A predefined area within an image deemed diagnostically relevant for analysis.

Multimodal deep learning: A method that integrates multiple data types—such as image and gaze information—into a unified predictive model.

References

1. Expert gaze as a usability indicator of medical AI decision support systems: a preliminary study. npj Digital Medicine (2024).
2. EyeXNet: Enhancing Abnormality Detection and Diagnosis via Eye-Tracking and X-ray Fusion. Machine Learning and Knowledge Extraction (2024).
3. Prediction of radiological decision errors from longitudinal analysis of gaze and image features. Artificial Intelligence in Medicine (2024).
4. Radiologists can detect the ‘gist’ of breast cancer before any overt signs of cancer appear. Scientific Reports (2018).