Dies ist eine Übersichtsseite mit Metadaten zu dieser wissenschaftlichen Arbeit. Der vollständige Artikel ist beim Verlag verfügbar.
10048-CO-2 CAN NATURAL LANGUAGE PROCESSING AI BE AN APPROPRIATE “ADVISER” FOR BRAIN TUMOR PATIENTS? CLINICAL APPLICATION AND ISSUES ON CHATGPT
1
Zitationen
4
Autoren
2023
Jahr
Abstract
Abstract INTRODUCTION Brain tumors affect approximately 20,000 individuals annually in Japan, with some cases requiring urgent treatment. However, the medical knowledge of non-medical individuals (the general public) is often limited, resulting in delays in seeking treatment even when neurological symptoms are evident. Having appropriate advisors who can encourage medical consultations could potentially improve this situation. While Natural Language Processing AI (NLP-AI) is widely used, its full integration into healthcare remains a work in progress. ChatGPT, a prominent NLP-AI, possesses extensive knowledge and is accessible 24/7 online. We explored the potential of leveraging its advanced capabilities and accessibility in the field of brain tumor management. METHODS To evaluate its ability to guide hospital visits effectively, we simulated various symptoms associated with primary brain tumors and consulted ChatGPT(GPT3.5) as patients. We also assessed how ChatGPT handled unnecessary noise information (unrelated symptoms) during symptom consultations. RESULTS ChatGPT consistently recommended hospital visits for all simulated brain tumor patients and provided appropriate advice regarding the urgency and relevant medical department. Notably, even when patients expressed hesitation and refusal hospital visits, ChatGPT persistently persuaded them to visit the hospital despite the user(patient)'s intention. Despite the introduction of noise information, ChatGPT accurately identified and emphasized the importance of relevant symptoms, prompting appropriate medical attention. However, accuracy declined when excessive noise was present. Additionally, ChatGPT generated concise summaries of medical histories for presentation during hospital visits. DISCUSSION While AI is widely utilized in medical imaging, the comprehensive integration of conversational AI in healthcare is still in progress. ChatGPT, despite not being a healthcare professional, holds potential as an "advisor" with extensive medical knowledge, accessible for consultations at any time, to support brain tumor patients. However, challenges related to accountability and expanding knowledge capabilities remain to be addressed.
Ähnliche Arbeiten
Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI
2019 · 8.245 Zit.
Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead
2019 · 8.100 Zit.
High-performance medicine: the convergence of human and artificial intelligence
2018 · 7.466 Zit.
Proceedings of the 19th International Joint Conference on Artificial Intelligence
2005 · 5.776 Zit.
Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI)
2018 · 5.429 Zit.