ESTRO 2025 - Abstract Book

S2404

Interdisciplinary – Other

ESTRO 2025

Physics, Radiobiology and Radiation Protection Group, IPO Porto Research Center (CI-IPOP), Porto Comprehensive Cancer Center (Porto.CCC) & Rise@CI-IPOP (Health Research Network), Instituto Português de Oncologia do Porto Francisco Gentil, Porto, Portugal Purpose/Objective Effective communication between patients and healthcare providers must be simple, clear and prompt. Patients must be kept informed of their appointments, have the tools to raise questions about their medical condition and treatment and have access to reliable 24/7 resources to address their concerns. This initiative seeks to improve the quality of healthcare services through a direct communication interface designed to optimize the efficiency of healthcare professionals, reduce hospital visits, streamline treatment workflows, and enhance patient engagement. The main objective of the project was to deploy an AI-driven chatbot to facilitate seamless interactions between the patient and a healthcare institution, in line with the national digital strategy. Material/Methods This project was developed collaboratively by a multidisciplinary team of physicians, physicists, radiation therapists, and nurses. The chatbot was designed for integration into the institution's systems to enhance communication within the radiotherapy department, ensuring compatibility with institutional workflows and robust analysis of unstructured medical text. The multidisciplinary team curated an intent dictionary, comprising tags, patterns, and predefined responses. This dictionary served as the foundation for training a neural network classifier. Using the Natural Language Toolkit (NLTK) and Keras, the model was trained using a Stochastic Gradient Descent (SGD) optimizer and a cross-entropy loss function, conducted over 150 epochs with a batch size of 5. The chatbot's ability to process and analyse unstructured medical text enables it to respond effectively to a wide range of patient inquiries. A quality assurance mechanism, incorporating a star rating system, will be implemented to continuously monitor and improve the chatbot's performance. Results The chatbot demonstrated the value of a multidisciplinary approach, enhancing patient-centred communication and decision-making processes. Achieving a mean loss of 0.338 and a mean accuracy of 0.904, it shows potential to optimise care delivery by reducing hospital waiting times and facilitating triage and follow-up with automated interactions. These capabilities streamlined consultations and treatments by guiding patients to the most suitable care options. Although the chatbot doesn't replace healthcare professionals, it can effectively complement their expertise, simplifying and speeding up processes without compromising quality. This integration can improve efficiency and patient experience while maintaining a high level of care. Conclusion The chatbot implementation can strengthen trust between patients and the institution, improve its reputation for innovation and encourage collaboration between departments. This initiative aligns with modern care standards and can effectively contribute to the institution's structural and technological modernization, positioning it in the field of healthcare innovation. References [1] Babu A, Boddu SB. BERT-Based Medical Chatbot: Enhancing Healthcare Communication through Natural Language Understanding. Explor Res Clin Soc Pharm. 2024 Feb 15;13:100419 [2] G. Rajani and K. Ruparel, "Deep Learning based Chatbot Architecture for Medical Diagnosis and Treatment Recommendation," 2023 International Conference on Advanced Computing Technologies and Applications (ICACTA), Mumbai, India, 2023, pp. 1-6 [3] Chow JCL, Wong V, Li K. Generative Pre-Trained Transformer-Empowered Healthcare Conversations: Current Trends, Challenges, and Future Directions in Large Language Model-Enabled Medical Chatbots. BioMedInformatics. 2024; 4(1):837-852 Keywords: AI, chatbot, patient centered care