A research team from atlanTTic, the Telecommunications Technology Research Centre at the University of Vigo, has developed VisIA-Q, a dataset that compiles demographic and psychometric information from more than 200 adolescents aged between 12 and 17 years. The study, published in the journal Scientific Data by Nature, involved collaboration with the Psychiatry departments of hospitals in Ourense and Vigo, as well as the Galicia Sur Health Research Institute. This resource aims to facilitate research into adolescent mental health and advance the development of early detection tools.
The dataset includes information from three groups of adolescents: those who have recently experienced a suicidal crisis, patients receiving psychiatric care for other reasons, and adolescents with no known history. To create this dataset, seven validated psychometric instruments were employed, allowing for both individual responses to various items and overall scores from the tests. The evaluation covers suicidal thoughts, general mental health issues, depressive symptoms, bullying, cyberbullying, and problematic internet use.
VisIA-Q serves as a valuable resource for researchers in psychiatry, psychology, statistics, and machine learning, focusing on the psychological factors associated with suicidal tendencies in adolescents and developing predictive models for risk assessment. The atlanTTic team, led by principal investigator José Manuel Ramírez Sánchez, included Carmen García Mateo and Laura Docío Fernández from the Multimedia Technologies Group, and Manuel José Fernández Iglesias and Mario Manso Vázquez from the Telematic Systems Engineering Group.
Ensuring data security
One of the team’s key contributions was the design of a technological infrastructure to ensure data security throughout the collection and processing stages. They developed specific software and adapted data capture devices, such as cameras and laptops, to prevent data extraction. The storage systems incorporate a dual encryption system, using both a password and a physical key.
“Our work focused on two fronts. First, we designed the infrastructure that allowed for data collection with full security guarantees, in compliance with the GDPR and the requirements of the Galicia Research Ethics Committee. Then, we processed, refined, and extracted information from the data compiled by professionals at IIS Galicia Sur, Ourense Hospital, and Álvaro Cunqueiro Hospital in Vigo,”
explained the atlanTTic research team.
The team also defined the anonymisation procedure for the 207 participants, using specific identifiers to prevent direct links between the data and the individuals involved. Clinical information collection and participant recruitment were managed by healthcare staff from the Psychiatry services of the university hospitals in Ourense and Vigo, coordinated by psychiatrist Alejandro García Caballero.
Bridging AI and clinical practice
“VisIA-Q bridges the gap between artificial intelligence and real clinical practice. It is a rigorous, anonymised resource validated by psychiatry professionals that can help drive new research and early detection tools for potential risks in adolescence,”
stated José Manuel Ramírez Sánchez. The dataset allows for the joint analysis of the potential relationship between various psychological factors and compares results from adolescents with different clinical profiles. Furthermore, the availability of individual responses to the seven standardised instruments provides a reference base for studying risk-associated factors and developing machine learning models aimed at psychometric screening.
This work addresses a gap in available resources, as many datasets used in this line of research are derived from adults or information extracted from social media without formal clinical evaluation. In contrast, VisIA-Q incorporates validated assessments conducted in a healthcare context and supervised by professionals, along with three differentiated comparison groups.
“One of the values of this work is that it focuses on a particularly vulnerable and underrepresented population, adolescents, and allows for the joint analysis of factors such as depression, suicidal ideation, bullying, and internet use, based on validated clinical assessments. Additionally, the data can serve as a reference for developing and evaluating machine learning models aimed at psychometric risk screening,”
the team noted. However, they highlighted some limitations, including the sample size, group composition, and the cross-sectional nature of the study, which provides information from a specific moment and does not allow for the analysis of participants' evolution over time. Furthermore, the high-risk group has a predominance of female participants (88.37%), necessitating careful validation of future models to avoid potential biases.
VisIA-Q is thus a resource for research rather than a diagnostic or clinical prediction tool. The work results from collaboration between atlanTTic researchers and professionals from the Psychiatry services of the hospitals in Ourense and Vigo, as well as IIS Galicia Sur. This cooperation combined the clinical expertise of healthcare teams with the technological and research capabilities of the telecommunications and information processing fields.
The involvement of healthcare professionals was crucial for recruiting adolescents, collecting clinical information, and applying evaluation protocols. Their expertise helped establish the criteria and conditions under which the study should be conducted, considering the characteristics and special vulnerability of the participating population.
“The collaborative effort was key to the success of the publication and allowed several professionals to address a complex problem together. Without the guidance of clinical personnel, it would not have been possible to correctly identify what material and how it should be done (at what moment of the crisis, with what type of interaction, and under what criteria to stop the protocol),”
the researchers emphasised.
Data protection measures
Data protection was a central aspect of the project. The devices used for data collection remained disconnected from any network, and the storage systems incorporated dual encryption through passwords and physical keys. Additionally, the data were anonymised using UUID v5 identifiers associated with a private namespace, ensuring that participants could not be re-identified from public information. Access to the dataset is also restricted; requests must come from research groups that must justify the purpose of use and sign a data usage agreement prohibiting redistribution or attempts to re-identify participants. The information can only be used for research purposes and must be stored securely.
Looking ahead, the team is working on developing and refining artificial intelligence models for emotional screening in adolescents.
