The Challenge

Pneuma Healthcare wanted to compare online and face-to-face therapy effectiveness, determine optimal session frequencies, and maximise insights from their extensive data. Additionally, they sought to develop a scalable, AI-driven platform integrating wearable technology data (e.g., sleep, activity, diet) to provide real-time, personalised feedback for clients. Their vision was to enhance client engagement and streamline therapists’ workflows using AI-powered mood tracking, predictive analytics, and large language models (LLMs).

The Solution

With support from Hartree Centre Northern Ireland Hub, Pneuma explored advanced AI and cloud-based data solutions to build a future-proof digital mental health platform.

Key recommendations included:

• Data Management & Integration: Implementing a cloud-based system (AWS S3, Azure Data Lake) to integrate structured and unstructured data (session recordings, psychometric assessments, wearable data). Microservices architecture and real-time ETL pipelines (Apache Kafka) were suggested for scalable data processing.
• Data Analysis: Using AI-driven time-series models (TensorFlow, PyTorch, LSTMs) to track long-term trends in mood, therapy responses, and treatment effectiveness. Dashboards in Power BI or Tableau were proposed for clear trend visualisation.
• Predictive Analytics for Therapy Outcomes: Deploying machine learning models (Random Forest, XGBoost, Keras neural networks) to anticipate therapy success rates based on historical client data, allowing personalised treatment plans.
• Large Language Models (LLMs) for Session Support: Exploring GPT-4 integration to analyse therapist-client interactions in real time, providing AI-powered insights on client engagement, emotional state, and therapist communication style.
• Real-Time Feedback System: Designing an interactive web-based dashboard for both therapists and clients to receive live AI-driven insights, with AWS Lambda-powered event-driven analysis ensuring low-latency responses.

Project Impact

Since our work with Hartree we have completed the prototype of an online group therapy / psychoeducational platform. This collects extensive data in a way that that we can readily employ machine learning techniques in timely fashion to understand client behaviour and progress. TH second phase of this project will begin to use AI to support the client’s journey between sessions.

In addition, we have continued to develop the ideas that were generated with Hartree in exploring the use of LLMs in therapist training and as a therapist support assistant, and have received a Marie Curie Sklodowska Horizon Europe grant to work with European experts in LLM to develop. This project, known as the CRYSTAL project will supPnport in developing our own secure LLM and will open up new markets in education, with mental health practitioners outside psychology and build on our ‘human centred’ model to enhance productivity and efficiencies.

• Scalable AI integration: Lays the foundation for future automated therapy support.
• Enhanced client engagement: Real-time feedback provides immediate insights and therapy adjustments.
• Optimised treatment strategies: Data-driven predictive analytics help therapists personalise sessions more effectively.
• Greater accessibility: AI-powered tools improve the effectiveness of remote therapy, making mental health support more inclusive.

 Looking Ahead: Pneuma Healthcare is exploring further funding opportunities and technical partnerships to advance AI-driven therapy solutions. Their next steps include testing the AI models in clinical settings and expanding wearable data integration to refine personalised mental health interventions.