We are sorry, but NCBI web applications do not support your browser and may not function properly. More information

## Results: 9

1.

**Total scanpath for different image category**. This parameter measures the actual distance covered by the eye’s movements. When the EM algorithm is employed, the scanpath increases significantly for every image category, but there was no significant difference between SCZ and HC groups. Error bars indicate SEM (*

*p*< 0.05 or less, for two-sample

*t*-test).

2.

**Number of saccades for different image category**. Graph shows the average, for each group, of the total number of saccades performed while viewing the ten images of each category. When the SR algorithm is employed, significant differences are found for constructions, landscapes, and fractals images only. Employing the EM algorithm significantly increases saccades for all image categories, but the number for saccades is similar between SCZ and HC groups. Error bars SEM (*

*p*< 0.05 or less, for two-sample

*t*-test).

3.

**Median fixations duration for different image category**. We plotted the median fixation duration for each experimental group while viewing every image category. Here, only significant differences for the SCZ and HC groups are found for the constructions images when the SR algorithm is employed. When SR is compared with EM, significant differences are found for all image categories in SCZ, but only presentation of blank images elicited significant differences between SCZ and HC groups under EM algorithm. Error bars indicate SEM (*

*p*< 0.05 or less, for two-sample

*t*-test).

4.

**Number of fixations for different image category**. This graph shows the average, for each group, of the total number of fixations performed while viewing the ten images of each category. As occurred with the saccades, when the SR algorithm is employed, significant differences are found for constructions, landscapes, and fractals images only. Employing the EM algorithm significantly increases the number of fixations for the same image categories and also for the blank. With EM, the number for fixations is similar between SCZ and HC groups. Error bars SEM (*

*p*< 0.05 or less, for two-sample

*t*-test).

5.

**Examples of eye movement’s recordings showing the SR and EM algorithms implementation**.

**(A)**Horizontal and vertical eye traces for a control subject while freely viewing natural images. Traces in blue depict the segments identified as saccades by the SR algorithm.

**(B)**Same vertical and horizontal eye traces shown in

**(A)**. Here, traces in green and red depict the segments identified by the EM algorithm as saccades and microsaccades, respectively.

**(C)**Main sequence of saccades and microsaccades for HC (in orange) and SCZ (in green) using the EM algorithm (

*n*= 6867 and 6431, respectively). The vertical black line depicts the 1° boundary between saccades and microsaccades.

6.

**Cumulative density function (CDF) for fixation durations for each image category**. These graphs show the shift in fraction fixation with short and large duration. The brackets depict the 95% confidence interval for each fixations duration bin. Employing the EM algorithm resulted in a left-shift of the distribution, both for the HC and SCZ subjects. Moreover, differences between these groups that were observed for images with high complexity were no longer significant when the EM algorithm was employed. Nevertheless, blank images still show significant differences in the distribution of fixation duration between SCZ and HC groups.

7.

**Total Euclidean distance for different image category**. The measure for visual exploration adds the minimal distances between fixation locations (see Materials and Methods) measured with the SR and EM algorithms. The total average distance for each subject is shown for each image category. In both groups of subjects, the amount of distance decreases with image complexity, independent of the algorithm used. SCZ patients showed significantly reduced visual exploration. The EM algorithm still shows significant differences for constructions, landscapes, and blank images. Error bars indicate SEM (*

*p*< 0.05 or less, for two-sample

*t*-test).

8.

**Probability density function (PDF) for fixation durations for each image category**.

**(A–D)**These plots are a smoothed representation of the binned histograms. HC subjects show a narrower distribution for fixation durations using the SR algorithm for image categories with the largest complexity. The distribution for construction was significantly different from all others; landscapes and fractals were similar but different to the other categories which were similar to each other (Kolmogorov–Smirnov,

*p*= 0.05). When we employed the EM algorithm, we found that the mode of their distribution changes slightly, with a largest proportion of the fixation duration seen for durations below 200 ms. However, statistical differences between these curves remain identical to those observed for the SR algorithm.

9.

**Visual exploration of images with different complexity**. Here are examples for a healthy control (HC, left column) and a patient affected with schizophrenia (SCZ, right column). The red lines and circles depict saccades and fixations, respectively. The example images are sorted in decreasing complexity measured as variances of their spatial frequency power spectrum.

**(A)**Construction,

**(B)**landscape,

**(C)**fractals,

**(D)**pink noise,

**(E)**white noise,

**(F)**gray, and

**(G)**blank images. Both subjects reduce the area covered by visual exploration as image complexity falls, but SCZ subjects typically exhibit more reduced areas than HC participants.