Digital Health & AI

🌟 ConectaSalud (LLUM) - Remote Triage and PTSD Detection in Children (2025 - Ongoing)

Partnership: LabLENI (UPV) + Consorcio Hospital General Universitari | Agency: Consorcio Hospital General Universitari

Digital solution enabling effective remote triage for identifying children at risk of Post-Traumatic Stress Disorder (PTSD) following traumatic events. The system was deployed in response to the 2024 DANA natural disaster in Valencia, collecting psychological questionnaires from children in Paiporta. Natural language processing analyses children's descriptions in their own words to evaluate psychological impact and facilitate early detection of mental health risks.

• Real-World Deployment: Psychological assessment of children affected by the 2024 DANA flooding disaster in Valencia (Paiporta community), collecting trauma narratives and experiences in natural language
• NLP for Trauma Assessment: Processing of children's free-text descriptions to extract psychological distress indicators, emotional language patterns, trauma markers, and PTSD risk factors
• Remote Triage System: Digital platform enabling scalable psychological screening without requiring in-person visits, facilitating rapid assessment in post-disaster scenarios
• Early Detection Focus: Identifying children at high risk of developing PTSD through linguistic analysis of their trauma descriptions, enabling timely psychiatric intervention
• Clinical Integration: Direct communication channels with psychiatric units for critical cases, ensuring seamless referral and follow-up care
• Technologies: Natural Language Processing (NLP), sentiment analysis, trauma-specific linguistic markers, psychological questionnaire digitisation, secure paediatric mental health data management
• Scalability: Designed to be adaptable to other disaster scenarios and healthcare settings, promoting comprehensive children's mental health management beyond Valencia
• Innovation: First NLP-based remote triage system for paediatric PTSD screening in natural disaster contexts, combining digital health with emergency psychological response, enabling large-scale mental health assessment whilst maintaining clinical accuracy
• Current Status: Active project - ongoing data collection and analysis following the 2024 Valencia DANA disaster

💬 Empathetic Virtual Assistant for Mental Health (MindCompanion)

Conversational chatbot based on LLMs with emotional detection capabilities that provides first-level psychological support and professional referral when necessary.

• Technologies: GPT-4 fine-tuned, sentiment analysis, crisis detection algorithms
• Capabilities: Automated cognitive-behavioural therapy, mindfulness techniques, emotional diary
• Safety: Escalation protocol to professionals in risk situations
• Validation: Study with 500 users showing significant reduction in anxiety symptoms

🎤 Early Depression Detection System through Voice Analysis

Mobile application that analyses acoustic characteristics of the user's voice during phone calls to detect early signs of depressive episodes.

• Technologies: librosa, openSMILE, XGBoost, Flutter mobile app
• Features: 88 acoustic parameters (pitch, jitter, shimmer, energy, MFCCs)
• Accuracy: 83% detection of depressive episodes 2 weeks in advance
• Privacy: On-device processing, no audio storage

🏥 Hospital Readmission Prediction Platform

Machine learning system that predicts the risk of patient readmission within the first 30 days post-discharge, enabling personalised preventive interventions.

• Technologies: Python, scikit-learn, LightGBM, SHAP explanations, FastAPI
• Data: Electronic health records, medication, sociodemographic data, comorbidity indices
• Results: AUC 0.79, 23% reduction in readmissions after implementation
• Deployment: Integrated into hospital system across 3 healthcare centres

📱 Medication Adherence App with Conversational AI

Mobile application that uses intelligent reminders and natural conversations to improve therapeutic adherence in patients with chronic diseases.

• Platform: React Native, Node.js backend, MongoDB, push notifications
• AI: Medication schedule optimisation algorithm, chatbot with NLP
• Gamification: Achievement and reward system to maintain motivation
• Impact: 35% improvement in adherence in study with 200 diabetic patients

🧘 24/7 Digital Emotional Crisis Intervention Platform

Multichannel system (chat, voice, video) that provides immediate intervention in emotional crises through AI and connection with professionals when necessary.

• Architecture: Microservices, WebRTC, Twilio API, AWS auto-scaling
• AI: Crisis severity classification, intelligent routing, real-time transcription
• Team: Integration with network of psychologists available 24/7
• Usage: 15,000+ users served, average response time <2 minutes

🩺 AI-Assisted Diagnostic System for Dermatology

Convolutional neural network that assists dermatologists in classifying skin lesions, including early melanoma detection.

• Technologies: PyTorch, EfficientNet-B7, Grad-CAM visualisation, DICOM integration
• Dataset: Trained with 100,000+ dermatoscopic images (HAM10000, ISIC)
• Accuracy: 96% sensitivity in melanoma detection, 92% specificity
• Certification: CE marked as Class IIa medical device