Hybrid Computerized EF Training Using Virtual Human Stimuli for Autism

Overview

The goal of this pilot clinical trial is to learn the preliminary feasibility and efficacy of a home-based hybrid computerized executive function (EF) training program using social stimuli for autistic individuals. The main questions it aims to answer are: 1. Does hybrid EF training using virtual human stimuli improve the EF and symptoms of autistic individuals? 2. What are the parents' attitudes toward the training program? Researchers will compare changes in EF task performance and parent-report autistic symptoms and investigate parents' responses on a feedback questionnaire after autistic individuals undergo two weeks of training. Participants will: 1. Undergo computerized training at home five 30-minute sessions a week, for two weeks 2. Visit the university campus before and after the training program to undertake an assessment

To facilitate the integration of executive functioning (EF) and socio-emotional cue processing, we developed a novel home-based EF training program utilizing virtual, culturally relevant human facial stimuli for autistic individuals. We will ask autistic participants aged 8-17 to engage in training using this newly developed application at home over a two-week period, with five 30-minute sessions per week. The newly developed hybrid computerized training application features three games targeting the three core components of executive function (EF): shifting, inhibition, and updating (Miyake et al., 2000). These games utilize 20 virtual Asian characters generated with Character Creator 4 (Reallusion Inc.). The Face-Judgment Game (Categorization or "分門別類") trains shifting and mental flexibility by involving sex and emotion judgments (de Vries \& Geurts, 2012). Each round has 33 trials. Trials begin with a blank period, followed by a cue period lasting between 2000 ms and a decreasing duration from 1000 ms to 200 ms as difficulty increases. The cue is either "Sex" or "Emotion." After these periods, a face displaying happiness or fear is shown at the center of the screen. The player judges the face's sex (male/female) or emotion (happy/scared) based on the cue, responding as quickly and accurately as possible. Trials are either repeat trials (same judgment as previous) or switch trials (different judgment). Performance indices include mean switch reaction time (RT) and switch cost RT, used for difficulty adjustment. The Gaze-Arrow Game (Point Out The Way or "指點迷津") trains inhibition and interference control by requiring processing gaze direction amid distractors (Eriksen \& Eriksen, 1974). Each round has 32 trials. Trials start with a blank period, then present a neutral or emotional face along with arrow flankers pointing left or right. The number of arrow pairs increases from one to five across levels, creating greater interference when target and flanker directions mismatch (Chanceaux et al., 2014). The player judges gaze direction. Trials are either congruent (gaze and arrows same direction) or incongruent (opposite directions). Performance is measured by the mean incongruent RT and the RT difference between incongruent and congruent trials (interference score), used for level adjustment. The Running Memory Game (Character Memorization or "人物記憶") trains updating working memory by processing face identities (Morris \& Jones, 1990). Each trial begins with an instruction cue, followed by a sequence of neutral or emotional faces. The cue prompts the player to remember a face from n trials earlier, with n increasing from 1 to 5 to impose greater working memory demands. The sequence length varies randomly from 6 to 10 faces, with no repeated identities. After the sequence, five recent faces are shown, and the player selects the face matching the cue. Accuracy serves as the performance index for level adjustment. Assessments of participants' EF and clinical symptoms will be conducted within one week before the first training session and within one week after the last training session. Additionally, we will examine parents' attitudes toward the program. The go/no-go task measures inhibition (Picton et al., 2007). It requires participants to press a button in response to one letter (go trials) and to withhold responses to other letters (no-go trials). Each trial consists of a letter appearing for 300 ms, followed by an interstimulus interval of 1,000 to 1,400 ms (mean = 1,200 ms). The go:no-go ratio is 3:1, with 240 go trials and 80 no-go trials. The false alarm rate is analyzed. The Sternberg working memory task assesses the storage and retrieval of items from short-item memory (Sternberg, 1966). It requires participants to remember sets of 3 or 6 capital English letters, one by one. The two set sizes are equiprobable and presented in a randomized order. On each trial, participants see a fixation cross for 500 ms, followed by the letter stimuli presented for 2,000 ms. This is followed by a 3,000-ms blank period, then a lowercase letter probe appears for 1,000 ms, followed by an interstimulus interval of 1,750 to 2,250 ms (mean = 2,000 ms). The response time window is 2,750 ms. There are28 trials for each set size. Accuracy at set size 6 is analyzed. The digit task-switching task measures shifting (Monsell, 2003). On each trial, a fixation cross is shown for 500 ms. Next, a cue indicating either "large/small" (magnitude judgment) or "odd/even" (parity judgment) is presented either at 0 ms or 1,000 ms before a single digit, ranging from 1 to 9 (excluding 5). Upon seeing the digit, participants are required to judge the digit according to the cue as fast as possible via a button press. A repeat trial occurs when the current judgment is the same as the previous one, and a switch trial is defined when the present judgment differed from the last one. These two trial types and the two cue-target intervals are equiprobable, with 32 trials for each combination. The reaction time (RT) switch cost (i.e., difference in mean RT between repeat and switch trials) at 0-ms cue-target interval, which allows no preparation, is analyzed. The Social Responsiveness Scale-2 (SRS-2) has 65 items measuring the social function deficits associated with autism (e.g., "has difficulty relating to peers") (Constantino \& Gruber, 2012). Each item is rated on a 4-point scale from "0" (never true) to "3" (almost always true). The questionnaire yields two DSM-V compatible subscale scores, including the Social Communication and Interaction (SCI) and Restricted Interests and Repetitive Behavior (RRB) scores. The raw scores of these two subscales are analyzed. A feedback questionnaire is completed by parents to evaluate their attitudes toward the training program and to collect suggestions for game improvement. They rate six items on a 7-point Likert scale, from -3 to +3. These items include 'interestingness of the training games' (from very boring to very interesting), 'difficulty of the training games' (from too easy to too difficult), 'duration of the training games' (from too short to too long), 'perceived change in the child following training' (from very negative to very positive), 'satisfaction with the overall arrangement of the training' (from very dissatisfied to very satisfied), 'willingness to have the child join similar cognitive training in the future' (from very unwilling to very willing). We hypothesize that both EF and clinical symptoms will improve due to enhanced integration of EF and social-cognitive processes. The hypothesis will be confirmed by conducting t-tests on the outcome measures.