Combined deep and reinforcement learning with gaming to promote healthcare in neurodevelopmental disorders: A new hypothesis

Fabrizio Stasolla ^1* Anna Passaro ¹ Enza Curcio ¹ Mariacarla Di Gioia ² Antonio Zullo ² MIRELLA DRAGONE ¹ Elvira Martini ¹

• ¹ Faculty of Law, Giustino Fortunato University, Benevento, Italy
• ² Mercatorum University, Rome, Lazio, Italy

Children and adolescents with neurodevelopmental disorders (NDD) may experience significant problems dealing with daily activities and/or everyday life environmental requests. Besides intellectual disabilities, communication disorders and challenging behaviors may occur. Commonly, isolation and passivity are acknowledged. Accordingly, social interactions may be relevantly compromised. NDD usually has an early onset, a variable clinical manifestation, a wide range of severity, and a recognized comorbidity (Howner et al., 2018;Malik et al., 2023). Their clinical conditions may have negative outcomes on their quality-of-lifefamilies' and caregivers' burden may be meaningfully increased with negative consequences on their overall management and healthcare (Kanniappan et al., 2024;Lefton-Greif et al., 2024;Materula et al., 2023). To tackle this issue an early assessment is crucial (Chorna et al., 2024;Henry et al., 2016). Thus, either standard tests or technology-based solutions are available (Ceruti et al., 2024;Niu et al., 2022;Woodcock and Blackwell, 2020). Traditional assessment relies on neuropsychological evaluation (Haddad et al., 2024;Hamadelseed et al., 2023). Technology-aided options may represent a functional bridge between personal skills and environmental requests by enhancing self-determination and positive occupation accordingly (Passaro et al., 2024). Recently, artificial intelligence-based programs (AI) emerged (Boubakri and Nafil, 2024). Both assessment and rehabilitative goals are targeted (Anbarasi et al., 2024;Climent-Pérez et al., 2024).Deep learning (DL), as part of machine learning solutions (ML), has been growingly used to evaluate the normal brain functioning and differentiate between individuals who have a normal development and individuals who are at risk of developmental disorders (Kucewicz et al., 2023;Li et al., 2024;Swinckels et al., 2024). For example, DL algorithms as convolutional neural networks (CNN) have progressed allowing future learning of a significant amount of data patterns which enables the subjectivity in future extraction procedure. Successful implementations of DL and CNN have been documented. Thus, brain functioning has been positively investigated through functional magnetic resonance (fMRI) in main domains (Hu et al., 2023).Reinforcement Learning (RL) as a further part of ML has been adopted for rehabilitative purposes.That is, an artificial intelligent agent continuously interacting with a participant dealing with a cognitive task is positively reinforced by such interaction and is capable of learning by it. Based on the interaction, it will provide the participant with an optimal task. Consequently, the artificial intelligent agent ensures the participants with highly customized and tailored solutions along all the working sessions and an ideal learning process will be ensured (Zini et al., 2022). Recently, RL-based principles have been adopted for emotional regulation in neurodevelopmental disorders and neurodegenerative diseases (Stasolla et al., 2024;Stasolla and Di Gioia, 2023).Gamification may be considered an advanced technological cornerstone in neurodevelopmental disorders for both assessment and rehabilitative purposes. Educational and recovery goals may be targeted. Education, healthcare, and rehabilitation objectives have been pursued. Significant improvements have been reported in individuals with developmental disabilities. Self-determination, independence, and fulfillment may be fostered by embedding features such as challenges, competitions, and rewards. Thus, gamification can help persons with neurodevelopmental disorders by enhancing an active role and constructive engagement (Boubakri and Nafil, 2024).A literature overview was performed in Scopus. Neurodevelopmental disorders, quality of life, DL, RL, gamification, assessment, and rehabilitation were merged as keywords. Although the detailed solutions were widely adopted in neurodevelopmental disorders (Alves et al., 2020a;Bakır et al., 2023;Brzosko et al., 2019;Nahar et al., 2024;Ouyang et al., 2024;Pandya et al., 2024;Rahman et al., 2024;Rodulfo-Cárdenas et al., 2023;Wyatt et al., 2024;Zhao et al., 2024), no records were found on their integration. In line with the above, the aims of the current opinion paper were (a) to provide the reader with a concise framework on the use of DL, RL, and gaming in developmental DL in neurodevelopmental is a widely adopted strategy of assessment with 63 records found including five reviews published between 2020 and 2024 (Kucewicz et al., 2023;Li et al., 2024;Soybilgic and Avcin, 2020;Swinckels et al., 2024;Wang et al., 2023). Li et al. (2024) conducted a comprehensive review on the use of electroencephalography (EEG) as a method that records changes in brain activity which may represent a marker for the identification of autism spectrum disorders (ASD). The review included DL and ML methods. Future perspectives and challenges were highlighted to automatically diagnose ASD through EEG signals to emphasize ASD automated identification. Kucewicz et al. (2024) RL in neurodevelopmental disorders is also represented in the literature with 60 documents available in Scopus and five reviews in the last two decades (Brzosko et al., 2019;Meyer et al., 2005;Rodulfo-Cárdenas et al., 2023;Swan et al., 2016;Wyatt et al., 2024). By further inspection, the contribution of Meyer et al. (2005) investigated an animal model for the early detection of schizophrenia, was considered as irrelevant for the current work, and was not detailed accordingly. Conversely, exploitation decision-making was associated with default network brain regions. Data were interpreted in the context of a network architecture useful to support the flexible switching between externally and internally directed cognitive processes, mandatory for adaptive and purposeful behaviors. Moreover, they surveyed studies involving neurodevelopmental, neuropsychological, and neuropsychiatric disorders as well as lifespan development and neurodegenerative diseases. Significant differences between exploring-exploiting decision-making were observed across the populations corroborating that the two decision-making modes are supported by independent neural circuits. Comprehensive neural circuity mapping and neural behavioral correlates associated with exploration-exploitation in humans were warranted. A new trans-diagnostic approach to assessment, surveillance, and intervention for cognitive decline and dysfunction in mental health for normal development and clinical population was putted forward.Gamification and neurodevelopmental disorders is represented in Scopus with 3 reviews (Alves et al., 2020b;Bakır et al., 2023;Boubakri and Nafil, 2024). Alves et al. (2020) Boubakri and Nafil (2024) explored the potential impact of gamification on accessibility issues for persons with disabilities. According to PRISMA guidelines, a literature review was performed across seven databases. Fifty-three studies were selected. Data revealed that gamification was suitable for neurodevelopmental disorders, blindness, and visual impairments, relevantly improving the learning and rehabilitation processes of individuals with disabilities. Nevertheless, significant gaps remained to be filled including the need for a comprehensive assessment framework and a more accurate integration with emerging technologies like AI-based programs to customize optimal solutions for individuals with disabilities. Although gamification was proven to be effective and suitable for targeting challenges in persons with disabilities, further balanced perspectives were required.Specifically, a focused approach to gamification to identify its potential benefits as well as to address the existing gaps and claim for its integration with emerging technologies to generate more inclusive and tailored experiences.An illustrative example might be represented by a combined immersive system including AR, VR, and gamification. One can argue on a highly immersive and funny system which may promote executive functions, communicative skills, and social interactions. Different rigorously customized solutions may be designed. Based on RL principles, one may envisage different tasks properly tailored on the participant's capacities and continuously adapted to his/her performance. Both cognitive skills and behavioral responses may be recorded, monitored, and tracked. Recently, Stasolla et al. (2025) proposed a suitable scoping review on this specific topic.Considering the above, a new hypothesis on a combined and integrated technological solution based on DL, RL, and gamification principles was proposed. Matched technological options have been recently outlined (Stasolla et al., 2023). A three-step hierarchical solution was recommended in individuals with NDD for assessment and rehabilitative purposes (see figure 1). One may design a first step in which through DL principles the artificial system will be capable of differentiate between individuals who may have a normal development and individuals who are at risk of NDD through exploring and mapping the brain activity during cognitive tasks. Once differentiated, with a combined solution of RL and gamification (i.e., second step), one may plan funny, educational, and rehabilitative tasks focused on promoting cognitive and communicative skills as well as redirecting challenging behaviors into adaptive, functional or occupational activities (Chiapparino et al., 2011).In a third step, the clinical validity may be assessed involving expert external raters in social validation procedures (Stasolla et al., 2019). Different solutions may be adapted and customized depending on the participant's functioning.For instance, for individuals with severe to profound intellectual and multiple disabilities, one may suppose to design a basic discrimination between positive and negative emotions or situations eliciting specific emotions. For individuals with a moderate level of intellectual disabilities, one may plan combined AR, virtual reality (VR), and gamification opportunities to enhance different social interactions. For individuals with a high intellectual functioning and ASD or for persons estimated as borderline between the normal range of intellectual functioning and mild intellectual disabilities, one may implement access to literacy integrated with occupational and/or leisure opportunities (Chiapparino et al., 2011;Lancioni et al., 2007). Finally, one may envisage combined rehabilitative programs focused on supporting academic skills, personal needs, and access to literacy (Stasolla et al., 2022). Different technological solutions may be considered to provide a fully immersive environment (Bennewith et al., 2024;Panzeri et al., 2024). Individuals with NDD and different levels of disabilities may experience significant difficulties in different daily contexts and settings. Because they commonly present intellectual, motor, communicative, and sensorial impairments, they constantly rely on families' and caregivers' assistance. This condition may be deleterious for their social image, status, and desirability. In fact, it may seriously hamper their quality of life. To overcome this issue, technology-aided solutions may be helpful as previous findings demonstrated (Kinsella et al., 2017;Paul et al., 2023;Pham et al., 2019). Recently, AI-based solutions emerged as a valid avenue to promote healthcare of children and adolescents with NDD (Mengi and Malhotra, 2022). Here is a new hypothesis on a combined technological solution was proposed. The following considerations were suggested.First, self-determination, constructive engagement, independence, and social inclusion were fostered. Thus, an active role, positive occupation, and functional activities were supported with a reduction of isolation and passivity. Assessment and rehabilitation were critically targeted issues (Harris, 2020;Kwan et al., 2021;Lancioni et al., 2008Lancioni et al., , 2004a;;Stasolla et al., 2014bStasolla et al., , 2014a;;Zimmer and Dunn, 2021).Second, different behaviors depending on the individuals' levels of functioning may be outlined.For example, one may envisage a simple discrimination between positive and negative emotions for individuals with a very low functioning and limited behavioral repertoire. Otherwise, one may design to promote communication and positive social interactions for individuals with a moderate level of disabilities. Furthermore, more complex programs focused on different hierarchical managed opportunities may be considered (Michalski et al., 2021).Third, one may include the assessment of brain activity and a rehabilitative intervention in an unique program. Thus, both evaluation and recovery issues may be embedded. Through DL systems one may differentiate between individuals with a normal development and people who are at risk of neurodevelopmental disorders. Gamification may play a dual crucial role with educational and rehabilitative purposes (Gao et al., 2024;Liu et al., 2024;Shariat et al., 2024).Fourth, families' and caregivers' burden may be reduced. The clinical validity of the proposed intervention may be assessed. One may argue that being constructively engaged, positively occupied, with highly customized and tailored solutions, individuals with NDD may be more easily involved and included in daily settings with a burden reduction and a more effective and suitable solution was adopted (Song et al., 2022).Despite the promising postulated outcomes, some relevant issues should be acknowledged. First, no empirical data was available. Systematic reviews, metanalysis, and single-subject comparisons matched with longitudinal studies should be carried out. Second, sustainability issues should be carefully considered. Human, financial, and environmental resources should be targeted to investigate the affordability and the suitability of such proposed combination. Ethical issues should be additionally targeted. Third, a differentiation between the NDD is currently lacking with a large body of literature focused on ASD. Other disorders such as rare genetic diseases should be included in future studies. Accordingly, future research perspectives should deal with the following topics (a) empirical and/or experimental data to be collected, (b) systematic comparisons between groups or longitudinal investigations based on single subjects should be sought, (c) families' and caregivers' burden should be targeted, and (d) the clinical validity with a focus on sustainability and suitability in daily settings should be prioritized.