Enhancing patient education in cancer care: Intelligent cancer patient education model for effective communication

Background: The incidence of cancer is on the rise annually, whereas there exists a significant deficit of healthcare personnel. Inadequate communication between healthcare providers and patients may result in adverse emotional outcomes for the latter and interfere with their treatment progress. A viable solution to alleviate patient distress involves utilizing text generation models as an efficacious tool for delivering patient education. Materials and methods: In this study, we proposed an intelligent cancer patient education model (ICPEM) based on the pre-trained T5 model. Meanwhile, we presented a new method for optimizing the model's comprehension of the patient's intent through simulating the inquiries that the patient may ask. The datasets used include a doctor and patient dialogue dataset and a cancer patient education scenario dataset. After prompt-tuning, the model is capable of educating patients through four major aspects including medical examination, health care, radio-therapy, chemotherapy. Results: We conducted a comprehensive evaluation of the model by employing both automated and manual metrics. Our findings indicate that the responses generated by the model effectively catered to the requirements of patient education. Furthermore, our visualization analysis demonstrated the model's proficiency in processing sentences that are prone to confusion while maintaining a robust comprehension of human intent. Finally, several shortcomings of the model are presented, such as the inadequate amount of knowledge and the limited range of responses. Conclusion: We proposed a method for training and deploying medical language models in a low-resource environment. The proposed model facilitated the comprehension of cancer patients' intentions, resulting in suitable responses, which promotes effective patient-provider communication.