100 The Effect of Task-relevant Salient Signal on Visual Search Tasks

Faculty Mentor: Brennan Payne (Psychology, University of Utah)

Purpose

Visual search is a common activity we do in our daily lives. It is a type of search that requires attention and typically involves scanning for a particular object. However, attention is a limited resource; we cannot pay attention to every detail of our lives so we subconsciously select which element is important and worth paying attention to. This selection process can be hindered by distraction; Moher(2020) investigated the effect of target irrelevant salient signals (distractors) on visual search tasks. Target-relevant signals are the stimuli that convey no useful information for tasks, compared to the task-relevant stimuli that are sometimes helpful in completing tasks. Moher suggested that the target irrelevant salient signals cause early quitting effect: when salient distractors were present, observers quit tasks earlier and thus missed the target which they otherwise would not have.

Missing the target in daily life cannot be too dangerous; it can be as minimal as missing the celebrity you walked past on the street because you were too focused on responding to your friend on the phone. But for some others, it may be more detrimental. Radiologists, for example, have to go through a myriad of medical scans looking for possible abnormalities. Any distractions during their readings of the medical scans are unfavorable since they can lead to misdiagnosis or mistreatment of the patients. In order to assist radiologists in minimizing the risks of missed lesions, computer-aided detection, CAD was designed to highlight the potential abnormalities to encourage additional evaluation from the radiologists. Though previous studies have suggested that CAD increases the accuracy, however, CAD is not always accurate and thus causes false alarms CAD (2, 3).

The present study investigates the effect of task-relevant salient signals on visual search tasks; we utilize task-relevant salient signals since CAD is a task-relevant salient signal. More specifically, we will observe if the early quitting effect can also be seen in visual search tasks with task-relevant salient signals. The observations from this study may ultimately help answer the question of whether the benefit of CAD outweighs the cost of it.

Method

Stimulus Design Stimuli used for this study were created by mimicking the medical scans radiologists read; rotated Ls and Ts in a noisy background represented the object on those medical scans, and the red circle around one of the letters Ls and Ts represented the CAD.

Study Design A between-subject design is employed; half of the participants are assigned to a salient signal present condition, in which red circles sometimes appear around a letter and the other half of the participants are assigned to a no salient signal present condition. Our salient signal is inspired by CAD, therefore, for the purpose of this paper, the salient signal is going to be referred to as CAD. For the first group of subjects when the target is present, CAD would highlight the target 75% of the time. 25% of the time, it would highlight non-target which makes CAD task-relevant. Both CAD and no-CAD groups completed 300 trials: 100 targets present and 200 target-absent trials. Participants would go through an additional 50 practice trials at the beginning of the experiment, therefore, 350 trials in total. Each trial has 12 objects in total, and only one rotated T is present in target-present trials.

Task Participants were asked to search for a rotated letter T among rotated letter Ls and give a simple yes(target present) or no(target absent) response while our eye tracker recorded their eye movements.

Eye-tracking data There are key variables that are being measured through each trial. First is response time, recorded from the moment the participants start each trial until they give a response using response keys. As well as accuracy, from participants’ keypress responses; represents how accurately participants performed the visual search tasks. Third would be the number and location of fixation. Objects on the screen (letters Ls and Ts and CAD) were considered fixated if their pupils fell within a 5 ̊ angle from them. Search coverage is another key variable. It is an estimate of the total area of the display viewed by the participants on any given trial. Additionally, dwell time: the amount of time participants spent fixating on a particular object. Lastly, oculomotor capture; identify which objects they most paid attention to, and how many total objects on the screen were fixated.

Result

We analyzed the data from four piloting participants. Based on these data, when CAD highlighted the target, higher accuracy was observed compared to both trials when CAD did not highlight the target, as well as the trials with no CAD at all. The accuracy of trials with CAD highlighting the target was over 90 percent and the accuracy of trials where CAD is not highlighting the target was around 65 percent, lastly, the accuracy of trials with no CAD was around 60 percent.

The average reaction time in target-present trials was shorter when CAD highlighted the target in targetpresent trials (=2.5 secs) compared to that of trials where CAD highlighted the nontarget(=4.3 secs), and reaction time of trials with no CAD (= 3.5 secs). Additionally, in target-absent trials, the average reaction time in the trials where CAD was absent was slightly longer (= 4.6 secs) than in trials where CAD was absent(=4.5 secs), and also than trials in no CAD condition (=4.5 secs).

Conclusion

Moher (2020) research showed that task-irrelevant salient signals cause early quitting and lower accuracy in visual search experiments. This summer, we are testing out if task-relevant salient signals also cause the same effect. We observed this effect in within-subject comparison; the average reaction time on CAD present trials was shorter compared to the average reaction time on CAD absent trials. However, this effect is yet to be observed in between-subject comparison; the average reaction time between CAD present trials compared to no CAD trials was similar. This result may change as we collect more data.

References

1. Moher, J. (2020). Distracting objects induce early quitting in visual search. Psychol Sci, 31(1), 31-42. doi:10.1177/0956797619886809\

2. Birdwell, R. L. (2009). The preponderance of evidence supports computer-aided detection for screening mammography. Radiology, 253(1), 9-16. doi:10.1148/radiol.2531090611

3. Philpotts, L. E. (2009). Can computer-aided detection be detrimental to mammographic interpretation? Radiology, 253(1), 17-22. doi:10.1148/radiol.2531090689