DATA STORIES | LLMs & AI | KNIME ANALYTICS PLATFORM
KNIME-Med-Chat-Bot: A Low Code Solution For AI Driven Conversational Information Extraction from Clinical Practice Guidelines
Bridging the gap between complex medical guidelines and everyday users with visual programming
Co-authors: Ekene Ben Agu, Jalynn Mabry, Austin Lyons, Hira Bhatti, Suad Alshammari.
Clinical practice guidelines, particularly those from organizations like the American Heart Association, play a pivotal role in shaping the landscape of patient care. Formulated through meticulous analysis of current research by expert panels, these guidelines provide evidence-based recommendations. As the bedrock of clinical decision-making, they promote consistency and excellence in healthcare practices. Offering granular guidance on diagnosis, treatment, and prevention, these guidelines not only foster enhanced patient outcomes but also keep healthcare professionals abreast of the latest medical advancements. Regularly updated to mirror evolving medical insights, these guidelines are key in steering treatment plans and bolstering patient comprehension of their health issues.
In the ever-evolving realm of healthcare, technological advancements are transforming our interaction with medical data. A notable innovation in this domain is the emergence of generative AI-powered chatbots, such as ChatGPT, Claude, Bard etc., being used for medical information extraction. However, these applications need to be grounded in most uptodate clinical practice guidelines to minimize hallucinations and provide updated information beyond the training cut off. The direct incorporation of entire medical texts, such as Clinical Practice Guidelines, into these models, however, faces practical challenges due to context window limitations and token costs. To circumvent these issues, a Retrieval Augmented Generation (RAG) approach is employed, enabling these AI tools to efficiently distill and provide pertinent information without the need to sift through voluminous documents.
The traditional creation of RAG-based applications, aimed at navigating medical guidance texts, often necessitates programming skills, such as Python proficiency, which is generally not included in the training of healthcare professionals. Identifying this skill gap, our team has crafted a solution: a chatbot built using KNIME Analytics Platform, an adaptable no-code/low-code data analytics platform. This breakthrough streamlines the intricate processes of coding and AI integration, making it accessible to healthcare practitioners of all backgrounds. Utilizing this intuitive platform, professionals can independently develop AI-based applications tailored to their unique needs, thereby enabling efficient engagement with any relevant medical guidance document. This tool democratizes access to sophisticated conversational information extraction technologies, significantly enhancing the capability of healthcare professionals to utilize and interact with clinical practice guidance documents in their specialized fields.
The Utility of the Medical Information Chat Bot
- Instant Access to Reliable Information: In critical situations, immediate access to accurate medical information can be a matter of life and death. This chat bot ensures that users receive reliable information promptly, aiding in informed decision-making.
- Empowering Patients: Patients are now empowered with the knowledge they need to engage in meaningful discussions with healthcare professionals. This newfound understanding fosters a sense of confidence and active participation in their healthcare journey.
- Supporting Healthcare Providers: Healthcare professionals benefit from the chatbot as well. By relieving them of basic inquiries, the bot allows medical staff to focus on more complex tasks, ultimately enhancing the efficiency of healthcare delivery.
The Power of KNIME: Unleashing the Potential of AI
KNIME stands at the heart of this innovative leap by empowering users to harness the utility of coding and software development without fully delving into the fundamentals of coding. The KNIME Community Hub provides users access to pre-made code condensed into what we call “nodes”. Each functional node performs a basic task but when integrated together they allow even inexperienced users to create complex data visualizations and analysis in the form of “workflows”. Taking advantage of the dynamic properties of KNIME, we uniquely configured a series of functional nodes to produce our interactive chatbot.
Bridging the Gap: Accessible Medical Information at Your Fingertips
One of the most significant advantages of this AI-powered chatbot lies in its ability to bridge the gap between complex medical guidelines and everyday users. Often, medical jargon and guidelines can be overwhelming, making it difficult for individuals to understand, interpret, and apply them to their specific situations. The chatbot acts as a mediator, simplifying intricate medical information into easily digestible and comprehensible responses. We used the American Heart Association’s most current guidelines on heart failure as the data retrieval source for our chatbot. This particular guideline contains 138 pages of dense medical information that can now be accessed and pulled from using natural language conversations with the chatbot.
Using KNIME to summarize clinical guidelines
Through KNIME we were able to create a chatbot that can analyze these guidelines. For the Python dependency we used these nodes from this workflow. For more information regarding the way we used Python dependency refer to this blog post written by Dayanjan Wijesinghe and Suad Alshammari. KNIMEZoBot: A Low Code Solution for Conversational Interrogation of Zotero Libraries with LLMs
Our workflow consists of two main components, one to read the provided guidelines and another to serve as the chat interface.
Chatbot in action — Heart failure
For the conversation, the KNIME-Med-Chat-Bot is given the name SLATE.
The American Heart Association (AHA) defines heart failure as a lifelong condition in which the heart muscles can’t pump enough blood to meet the body’s needs for blood and oxygen. Heart failure is a common condition that affects over 6 million adults in the United States. The prevalence increases significantly with age, affecting over 10% of people over the age of 80.
Heart failure can have a major impact on quality of life. Symptoms include shortness of breath, fatigue, fluid retention, and exercise intolerance. It is also associated with an increased risk of hospitalization and death. Heart failure is a leading cause of hospitalizations in older adults.
Properly managing heart failure is critical to improve symptoms, quality of life, and outcomes. Heart failure treatment is based on guidelines so it is critical that healthcare professionals are able readily access these guidelines to provide patient care. Fortunately, we have created a system that can do just that. Heart failure was used to showcase how our program works but treatment guidelines for any disease state can be used.
A video demonstration of the workflow in action can be seen below.
Conclusion: Shaping the Future of Healthcare
The healthcare field is ever changing with new guidelines seeming to come out every year. Healthcare workers find it very hard to keep up with the ever changing field. Our chatbot aims to help the workers make sense of new guidelines as quickly as possible so they can better treat their patients. This chatbot allows users to input any future guidelines and ask treatment based questions. The chatbot is able to retrieve the requested information from the provided guidelines and accurately provide an answer. One of the key features of our chatbot is its ability to give you quick reliable information based on whatever guidelines you need. It does not pull information from outside sources. It also saves time by eliminating the need to sift through individual guidelines to find information.
The creation of this AI-powered medical information chat bot using KNIME marks a pivotal moment in healthcare technology. By making intricate medical guidelines accessible, understandable, and applicable, it empowers individuals and transforms the way we navigate the complex world of healthcare.
As we move forward, the integration of AI and platforms like KNIME promises a future where healthcare information is not just a privilege for a select few but a fundamental right for all. Through innovation and accessibility, we are paving the way for a healthier, more informed society, where the power of knowledge is harnessed to its fullest potential.
Project Workflows
- The workflow is available to download at the KNIME Digital Healthcare page under the folder KNIME-Med-Chat-Bot. This is a repository for KNIME based digital health workflows built by us and others.
- The complete KNIME workflow for the KNIME-Med-Chat-Bot is also available at https://github.com/dayanjan-lab/KNIME-Med-Chat-Bot.git
- The workflow is distributed under the MIT license and is thus free to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the software, offering a wide scope of possibilities for both personal and commercial use.
