1.

Context: Chatbots have gained popularity over the years, especially in the healthcare sector, due to their ability to improve patient engagement and satisfaction. This project aims to contribute to this field by suggesting a healthcare system chatbot that depicts patient summary based on the European guidelines. The primary challenge addressed is the efficient delivery of the complex patient information included in the patient summary, in order to find crucial information quickly and easily. To address that we are implementing a prototype of the suggested system. This chatbot will provide patients and doctors with a comprehensive summary, encompassing medical history, medications, allergies, and vaccinations. Preliminary findings indicate that the rule-based chatbot serves as a valuable tool, enhancing patient engagement, reducing waiting times, and contributing to overall healthcare efficiency. This research contributes to the healthcare domain by introducing a functional chatbot prototype and envisions a future where advanced technologies revolutionize patient-doctor interactions and healthcare delivery.

2.

Subject and objectives: In recent years, technology has become essential to healthcare, significantly influencing patient outcomes and satisfaction. A pivotal concept in this technological evolution is eHealth, encompassing the use of electronic communication and information technology to enhance healthcare. Specifically, the European Union, through its EU4Health program has undertaken a commitment to reinforce crisis preparedness and address long-term health challenges by building stronger, more resilient, and more accessible health systems. The EU4Health program pursues four general objectives and ten specific objectives, aligning with our research focus on Strengthening healthcare systems. Within the broader scope of eHealth, which includes applications such as telemedicine, electronic health records (EHRs), mobile health, and health information exchange , our work contributes to the specific objectives outlined in the EU4Health program. EHRs, a cornerstone of eHealth, grant healthcare providers access to crucial patient information, contributing to improved diagnosis, treatment, and overall healthcare quality. Particularly, the patient summary emerges as a standardized document designed for shared use among healthcare providers, fostering better coordination of care and reducing medical errors. The patient summary, providing details on medical conditions, medications, allergies, vaccinations, and care teams, empowers both healthcare providers and patients. By offering a comprehensive overview, it enhances coordination and puts patients in control of their medical information, fostering improved healthcare delivery. Moving to the realm of chatbots, these artificial intelligence-driven programs play a pivotal role in healthcare. Rule-based chatbots, characterized by predefined responses, find utility in providing information about health conditions, medications, treatment options, healthcare services, and insurance coverage. Machine learning chatbots, a more sophisticated variant, leverage artificial intelligence to enhance responses over time, offering versatility in handling complex queries and patient engagement. While rule-based chatbots are simpler to implement, their limitations include an inability to provide personalized recommendations or accurate responses to complex, context-dependent queries. Nonetheless, both rule-based and machine learning chatbots contribute to improving patient engagement, reducing wait times, and enhancing overall healthcare delivery.

3.

Methodology and Work: The first step was writing down the requirements of the system. The functional requirements of the system are the following:

• Login – The users must be able to enter the system using their ID, in order to be identified.
• Provide assistance – The chatbot must be able to provide assistance to the user at any point in the conversation flow and direct them to find what they are looking for easily and quickly.
• Questions – The chatbot must be able to answer all questions about the Patient Summary.
• Exit – The chatbot should terminate the chat flow in case the user asks to log out.

As for the questions that the chatbot can answer in relation to the user’s Patient Summary, these have been grouped into categories according to the information they request. The categories are the following:

• Allergies and Intolerances
• Vaccinations
• Past diseases
• Medical Surgeries
• Current problems
• Medical devices and implants
• Summary of Prescription

Then we create use case scenarios. For example, imagine a scenario where a user, in need of emergency surgery abroad, requires immediate access to vital health information. The doctor overseeing the case must quickly and efficiently obtain details about the patient’s allergies, current health problems, and most recent surgeries to ensure a safe and informed medical procedure. A user, faced with a medical emergency requiring surgery in a foreign location, urgently needs to provide pertinent health information to the attending doctor. The doctor must swiftly access crucial details about the user’s medical history to ensure the safety and success of the procedure. The doctor, equipped with the user’s unique ID, connects to the healthcare chatbot through a secure platform. Using the unique identifier ensures privacy and quick retrieval of the user’s specific health information. The doctor initiates a conversation with the chatbot, asking for information about the user’s allergies. The chatbot promptly runs queries on the patient summary database, extracting and presenting the relevant allergy information in a user-friendly format. Following the allergy inquiry, the doctor proceeds to ask the chatbot about the user’s current health problems. The chatbot efficiently retrieves and organizes the information from the patient summary, providing a concise summary of the user’s existing health issues. The doctor can then continue asking for any other relevant and important information he wants from the patient’s summary. By utilizing the chatbot to access the patient summary, this use case exemplifies how our healthcare system can facilitate quick and accurate retrieval of essential health information, providing crucial support in emergency medical situations, and ensuring seamless communication between the user and healthcare professionals.

We implement a prototype using BotPress: In the initial phase of our project, we strategically adopted the BotPress platform, an open-source machine learning framework, to craft a pioneering prototype for our healthcare chatbot. This implementation marks our exploration into the feasibility, utility, and effectiveness of this innovative concept.  Our choice of BotPress was guided by its adaptability and robust capabilities in natural language processing (NLP) and chatbot development. In the backend, JavaScript seamlessly integrates with BotPress and enables dynamic functionalities within our healthcare system. Our prototype lies on the BotPress engine, excelling in NLP and the construction of interactive chatbot flows. Leveraging BotPress’s capabilities, our chatbot adeptly interprets user queries, comprehends natural language inputs, and intelligently responds based on pre-defined rules and interactions. Although currently operating as a rule-based chatbot, it effectively understands user inputs in a tag format. This implementation leverages the capabilities of JavaScript, JSON files, and a robust MySQL database as reliable data sources. Together, these elements contribute to effective data management, ensuring seamless information retrieval crucial for the chatbot’s functionality. In the backend, customized JavaScript actions enhance the chatbot’s functionality. Crafted to seamlessly integrate with test databases, these actions facilitate a smooth flow of information retrieval, contributing to an enhanced overall user experience. JavaScript’s versatility allows us to create dynamic responses tailored to the unique needs of our healthcare chatbot. Through the chatbot interface, we introduced an efficient connectivity mechanism, enabling patients and doctors to effortlessly connect using a unique ID. This innovative feature ensures swift access to crucial information within a patient’s summary, covering essential details such as allergies and vaccination history. Our initial implementation is grounded in a user-centric approach, prioritizing the accessibility and functionality of the healthcare chatbot. As we progress, our vision includes refining and expanding this prototype by incorporating advanced artificial intelligence algorithms and machine learning models, elevating the chatbot’s intelligence to further enhance the user experience

4.

Results and Conclusions: In the initial phase of our project, we successfully implemented a rule-based chatbot using Botpress and a MySQL database, with a primary focus on delivering patient summaries adhering to European guidelines. While the initial user feedback was positive, we identified certain limitations that prompted us to embark on a new phase of development. One of the challenges encountered related to security issues and a user desire for increased flexibility in customization. Recognizing the significance of addressing these concerns, we initiated the development of an enhanced chatbot version, pivoting to Python as the backend technology. A significant evolution in the new version lies in the Natural Language Processing (NLP) aspect. We meticulously fine-tuned a BERT model for intent classification, enabling the chatbot to discern user intentions with greater accuracy. Furthermore, we incorporated the spaCy library for Named Entity Recognition (NER), a pivotal step in extracting specific details from user inputs[6]. The new version has successfully advanced to categorizing user input and providing patient summary information based on predicted intent. The robust intent classification capability is already deployed, while the entity extraction component is currently under active development. This entails refining the system’s ability to recognize specific details, such as allergy types (e.g., drug or food allergies), and tailoring responses accordingly. The ongoing development efforts are centered on optimizing the efficiency and accuracy of understanding user inputs. The goal is to empower the chatbot with an advanced capability to recognize more specific information, ensuring that it evolves into a smarter and more secure tool for healthcare interactions. Of particular note is the commitment to implementing a feature that allows the chatbot to filter responses based on user-provided information. This personalized approach ensures that the chatbot caters to individual needs, providing tailored and relevant information in alignment with user-provided details. The overarching objective is to elevate the chatbot into a smarter and more secure tool for healthcare interactions. By continually refining and expanding its capabilities, we aim to ensure that users experience an enhanced and personalized engagement with the chatbot, thereby optimizing its utility in diverse healthcare scenarios. This ongoing development phase reflects our dedication to addressing limitations and proactively shaping the chatbot into an intelligent, secure, and user-friendly healthcare companion.