Changes in Visual Attention Patterns for Detection Tasks due to Dependencies on Signal and Background Spatial Frequencies

TL;DR

This work addresses how image background, signal properties, and lesion morphology influence visual attention and diagnostic errors in digital breast tomosynthesis (DBT). The authors combine phantom-based DBT simulations with two lesion types (3-mm spheres and 6-mm spiculated lesions), six observers, and eye-tracking to quantify gaze metrics and error signatures, modeling performance with $Gaussian CDF$ fits to derive $80\%$ success thresholds for lesion visibility, search, recognition, and decision. They find that later perceptual stages, especially the decision stage, require higher signal contrast and are strongly modulated by background complexity and target structure, with spiculated lesions showing robust salience across backgrounds while spherical lesions are more background-sensitive. These results illuminate the interplay between local signal features and global anatomical noise and suggest avenues for improving lesion conspicuity, imaging protocols, and gaze-guided decision support in dense-breast imaging.

Abstract

We aim to investigate the impact of image and signal properties on visual attention mechanisms during a signal detection task in digital images. The application of insight yielded from this work spans many areas of digital imaging where signal or pattern recognition is involved in complex heterogenous background. We used simulated tomographic breast images as the platform to investigate this question. While radiologists are highly effective at analyzing medical images to detect and diagnose diseases, misdiagnosis still occurs. We selected digital breast tomosynthesis (DBT) images as a sample medical images with different breast densities and structures using digital breast phantoms (Bakic and XCAT). Two types of lesions (with distinct spatial frequency properties) were randomly inserted in the phantoms during projections to generate abnormal cases. Six human observers participated in observer study designed for a locating and detection of an 3-mm sphere lesion and 6-mm spicule lesion in reconstructed in-plane DBT slices. We collected eye-gaze data to estimate gaze metrics and to examine differences in visual attention mechanisms. We found that detection performance in complex visual environments is strongly constrained by later perceptual stages, with decision failures accounting for the largest proportion of errors. Signal detectability is jointly influenced by both target morphology and background complexity, revealing a critical interaction between local signal features and global anatomical noise. Increased fixation duration on spiculated lesions suggests that visual attention is differentially engaged depending on background and signal spatial frequency dependencies.

Figures (7)

• Figure 1: Sample regions with spherical (left) and speculated (right) lesions.
• Figure 2: Setup of remote eye-tracking device attached to the monitor and the setup of in-house built graphical user interface (GUI).
• Figure 3: Differences between search patterns of an observer on an DBT slice of a 25% dense XCAT breast phantom (a), a 25% dense Bakic phantom (b), and a 50% dense Bakic phantom (c). The size of the circles are proportional to the fixation duration. Observer made fewer fixations and diagnosed quickly 25% dense and XCAT breast images than 50% dense and Bakic images respectively.
• Figure 4: The average amount of time spent and number of fixations made on each image, first hit time, lesion dwell time, and number of fixations on lesion were plotted for different breast densities and phantom types. Error bar lengths indicate twice the standard error of the six observers’ average gaze metrics. Observers made less number of fixations and diagnosed quickly 25% dense images than 50% dense images and XCAT breast backgrounds than Bakic backgrounds. Observers took longer to first fixate on lesion in 50% dense images than 25% dense images whereas no difference was observed due to change of phantom type. Observers made fewer fixations on lesion when inserted in 50% dense images than in 25% dense images whereas no difference was observed due to change of phantom type. * represents a statistical significant difference with 0.01< p-value < 0.05. ** represents a statistical significant difference with p-value < 0.01.
• Figure 5: The average amount of time spent on each image, number of fixation per image, first hit time, lesion dwell time, and number of fixations on lesion were plotted for change of signal type. Observers spent less time and made lesser number of fixations on sphere lesion compared to spicule lesion. No significant differences in other gaze metrics observed due to change of signal type.