Lawrence C. Toto

Recording and analyzing eye-position data using a microcomputer workstation

This paper presents a PC-based eye-position data collection and analysis system. Software routines are described that supplement hardware calibration procedures, improving data-collection accuracy and reducing the number of unusable trials. Collected eye-position data can be remapped over a displayed stimulus image and spatially and temporally represented by parameters such as individual fixations, clusters of fixations (Nodine, Carmody, & Kundel, 1978), cumulative clusters, and gaze durations. An important feature of the system is that the software routines preserve scan-path information that provides a sequential dimension to the analysis of eye-position data. A “hotspot” analysis is also described, which cumulates, across 1 or more observers, the frequency of eye-position landings or “hits” on designated areas of interest for a given stimulus. Experimental applications in the fields of radiology, psychology, and art are provided, illustrating how eye-position data can be interpreted both in signal detection and in information-processing frameworks using the present methods of analysis.

Searching for bone fractures: A comparison with pulmonary nodule search

RATIONALE AND OBJECTIVES We aimed to determine if the characteristics and principles of visual search described for the detection of pulmonary nodules apply to extremity fractures. METHODS The eye positions of staff orthopedic radiologists, radiology residents, and medical students were monitored as they searched hand and wrist X-ray images for fractures and a chest image for nodules. RESULTS More systematic scanning patterns were observed for experienced observers than inexperienced observers. Positive decisions for bone images were associated with prolonged gaze durations; prolonged gaze durations were significantly longer for false-negative versus true-negative decisions. Intercluster jump distances were found to be greater for chest images than bone images. CONCLUSIONS A search for bone fractures can be qualitatively characterized by classifying observer scan paths, dwell times, and jump distances. Gaze duration can be a useful predictor of bone image locations containing potential missed fractures. Perceptual feedback could aid observers in the detection of inconspicuous fractures.

Searching for lung nodules. A comparison of human performance with random and systematic scanning models.

The contrast sensitivity of the retina is greatest in the center and decreases rapidly toward the periphery. Therefore, the detection of low-contrast lung nodules depends upon the manner in which the image is sampled by retinal receptors as eye fixations jump across the image during scanning. The scanning performance of two radiologists was compared with two computed models, a systematic and a random scanner. Although radiologists do not seem to have random scanning patterns, their coverage of the image was matched more closely by the random model. This suggests that radiologists employ a scanning strategy that is designed to cover the image of the lungs in a minimum time using the smallest possible visual field. The visual field size that is most effective in detecting nodules during search has a radius of 3.5 degrees visual angle. Nodule detection may be limited by basic neurologic constraints on human scanning performance.

Nodule detection with and without a chest image.

The detectability of nodules displayed on uniform backgrounds was compared with their detectability on chest images by using receiver operating characteristic (ROC) curves. The images were displayed using a digital television system. Two conditions, chest image and uniform background, were compared at three different levels of added random noise. The viewing conditions were made as similar as possible by setting the luminance of the uniform background equal to the luminance of the lung in the chest image. Nodule detectability was significantly lower for the chest image than for the uniform background at each level of added random noise. The anatomic structure of the chest image interferes with the structural integrity of the nodule image resulting in lower detection performance.

Comparison scans while reading chest images. Taught, but not practiced.

Radiology instructors and residents were surveyed for their methods of instruction concerning viewing techniques. A similar group of radiologists had their eye activity measured as they viewed chest images. Image reading techniques are taught to be systematic and directive with comparisons of bilateral features. Yet, most images are read by a free search method; bilateral comparisons comprise less than 4% of the visual activity. Instructors and residents show this discrepancy between instructional techniques and reading practice.

Searching for lung nodules. The guidance of visual scanning.

Four subjects searched a chest image for lung nodules using an eye position-dependent display that presented the image of a nodule selectively to either the central or peripheral visual field, while preserving the appearance of the chest. The time required to scan the image and fixate the nodule was shortest for nodules that were both reported and accessible to the peripheral vision. A stepwise concentric reduction in the size of the peripheral field that could access the nodules only affected search performance when the field was less than 5 degrees. (A chest image subtends about 25 degrees.) These data support the hypothesis that the optimal scanning strategy for lung nodules consists of spacing fixation clusters 5 degrees apart, and that peripheral vision only acts as an adjunct in guiding the gaze to inconspicuous nodules.

Contrast gradient and the detection of lung nodules.

The effects of edge gradient and blur on the detection pulmonary nodules were studied, using a test series of 175 artificial nodules on 30 chest films. The nodules were of homogeneous texture, about 1 cm in diameter and had a mean contrast of .12. Six observers viewed the film set, scoring both location and confidence. The sensitivity or the probability of a true-positive response was linearly related to the blur and a modified ROC analysis resulted in a family of ROC curves with the index of detectability monotonically decreasing with increased blur.

Eye Movements and the Detection of Lung Tumors in Chest Images

Publisher Summary There is a large body of data showing that competent radiologists and chest physicians overlook about 30 percent of lung tumors on chest radiographs, tumors that are clearly visible in retrospect. These error rates which were first studied systematically in the late 1940s and early 1950s have not changed with improved technology. A recent report by Muhm et al . from the Mayo Lung Project, part of a long term prospective study of the value of cancer screening in heavy cigarette smokers over age 45, indicated that 45 of 50 peripheral lung cancers were visible retrospectively in chest radiographs. Clearly, lung cancer would be detected earlier by radiogram if these errors of mission were eliminated. It had been suggested that many of these errors could be attributed to faulty search and it is reasoned that if the progress of search could be followed by recording eye position then areas of the chest that were not “adequately” searched could be identified and given a second look. This was called “feedback assisted visual search.” This approach was predicated on the hypotheses that inadequate or incomplete search was responsible for errors and the first experiments were designed to test this hypothesis. This, in turn, required a definition of “adequate” search of a chest radiograph.

Densitometric measurements of lung nodules on chest radiographs.

A technique is described whereby embedded nodular densities appearing on chest radiographs can be densitometrically measured without the obscurring effects of anatomic structures. The technique consists of rear-projecting a light beam into a chest radiograph and an observe adjusting this beam until the lucency matches that of the nodules appearance. The radiography is then removed, and the lucency is scanned and analyzed with a microcomputer. Calibration data using simulated nodules show a good accuracy in measuring size, contrast, and edge gradient.

Eye-position study of the effects of a verbal prompt and pictorial backgrounds on the search for lung nodules in chest radiographs

Peripherally inconspicuous nodules on chest radiographs are frequently missed by competent readers. In order to find a peripherally inconspicuous nodule the reader must inspect the nodule site with the central vision and decide if the features at the site are sufficiently characteristic to report a nodule. The experiment reported here was carried out to examine the effect of a nodule prompt and distraction by unrelated native abnormalities on the location and recognition of inconspicuous lung nodules on chest x-ray images. On two occasions separated by 3 years, 4 radiologists had their eye- position recorded while viewing 24 chest radiographs, 12 with prominent native abnormalities and 12 with no abnormalities. An inconspicuous nodule was simulated in the lungs of half of the radiographs on the first viewing and in the other half on the second viewing. For the first viewing, the readers were instructed to report any abnormalities. For the second viewing the readers were told to report any abnormalities including nodules. A nodule prompt triggers a scanning strategy that sends the central vision to high probability nodule sites early in search and at the same time relaxes the criteria used to evaluate nodule features resulting in more true positives and false positives without a change in absolute detectability. Prominent native abnormalities, unrelated to nodules, do not affect the search strategy but competitively inhibit the nodule feature recognition mechanism.