# Digital screening of children with ASD: diagnostic accuracy of emotion recognition and visual preference tasks

## TL;DR

This study explores how digital tools can help identify autism in children by analyzing their ability to recognize emotions and visual preferences.

## Contribution

The study introduces a digital screening method using emotion recognition and visual preference tasks to detect ASD in children.

## Key findings

- Children with ASD showed significant differences in recognizing fear and sadness emotions.
- Digital screening achieved 81.25% accuracy in distinguishing children with and without ASD.
- Gaze patterns in video stimuli helped differentiate between groups in visual preference tasks.

## Abstract

In Germany, there is a need to improve care for suspected autism spectrum disorder (ASD) cases, as the time between parents’ initial suspicion and an official clinical diagnosis can reach three years. New technologies for digital screening promise relief in addressing children’s difficulties in recognizing emotions and social attention. This study investigated the diagnostic validity of tablet-based screening to differentiate between children with and without ASD via emotion recognition and visual preference tasks involving prior calibration.

This study involved 24 boys with ASD and a matched control group of 24 typically developing (TD) boys aged 6–11 years. Mixed logistic models were applied for the emotion recognition task, while mixed linear models were used for the visual preference task, along with decision trees for both tasks.

The results indicate significant group differences in recognizing the emotion “fear” and naming an example for “sadness”. The emotion recognition of fear and sadness was relevant to the decision tree to differentiate between the groups, with an accuracy of 81.25%, a sensitivity of 91.67%, and a specificity of 70.83%. For the visual preference task, no significant group differences were found between groups; however, significant differences emerged between social and non-social image stimuli. Gaze fixation and gaze changes in video stimuli were relevant to the decision tree to differentiate between the groups. The accuracy was 81.25%, with a sensitivity of 70.83% and specificity of 91.67%.

Overall, this study suggests that automated digital screening might provide support and relief to families and clinicians, as it can distinguish between children with and without ASD using a combination of selected emotion recognition and visual preference tasks.

Not applicable.

The online version contains supplementary material available at 10.1186/s12888-025-07725-z.