Natural Language Processing and Digital Health Transformation, Health News, ET HealthWorld

Natural Language Processing and Digital Health Transformation

By Satish Kumar

Over the previous years, digital technologies have gone a widespread adoption across health organisations. It has helped healthcare organisations in reducing efficiencies, improving access and enhancing the quality of care. Digital Healthcare is also helping in preventing chronic diseases. One of the key technologies driving this transformation is Artificial Intelligence. It is transforming the entire value chain of healthcare delivery including the critical clinical and administrative processes. Many of these exciting AI applications use Natural Language Processing.

Natural Language Processing (NLP) is a sub-branch of AI which deals with algorithms and models to process language as human beings do. NLP uses deep learning algorithms to process unstructured data to structured data. Clinical notes and physician opinions even in Electronic Health Records (EHR) systems are usually in unstructured text. NLP systems can analyse these unstructured data and derive useful medical insights. NLP used with Record Linkage and other AI technologies like genome analytics can help in delivering personalised medicine.

NLP systems can shift through newly generated medical knowledge in articles and research papers can create useful summaries out of them. Availability of these summaries linked with EHR will help, physicians to take more informed clinical decisions and optimise treatment plans.

Healthcare is transitioning to value-based care and one of the critical success factors in this transition is patient experience. A medical virtual assistant can aid in collecting patient information like demographics, insurance, etc. It can navigate the patient to enter self-measured physiological parameters. In question-answer or chat format, these visual assistant asks the user about the symptoms or illness. Once assistant gets the answer, it matches it with possible diagnoses and creates probability scored outcomes for the next step – which might be another question, forwarding to a human agent or telling the patient what to do next. Building this intelligence requires that the agent use NLP to tokenise the answer, create the feature set and do the information matching using pre-trained deep learning models.

Training these deep learning models requires a huge amount of data about disease causes, symptoms, and diagnoses. Most of the major hospitals and healthcare providers have electronic health records, which must be de-identified before using them for model training and validation. Virtual assistants have improved on their accuracy to a great degree in the last few years, still the best results can be obtained when are paired with a human physician.

NLP-based assistants can also be used for routine administrative tasks like self-scheduling/re-scheduling/cancellation of appointments by the patient to a healthcare facility. For better adoption, they can be made as part of mobile applications and voice-enabled.

NLP and AI can also playing a big role in disease diagnosis and disease prevention. AI systems are used in surveillance and outbreak detection. These are core functions of public health and are critical for keeping the population safe from infections diseases and variants. AI systems can help in prioritising medication or vaccination when the capacities are limited.

In the pandemic, hospitals have used pre-hospital Digital Triage to direct the patient to appropriate care setting like home quarantine, respiratory clinics, testing sites, surge facilities, emergency rooms, etc.

NLP and other AI technologies are playing an important role in public health transitioning to Precision Public Health. This will help in delivering the right intervention to the right population at the right time and includes consideration of social determinants of health.

By analysing the structured/semi-structured data in EMR/EHR systems using AI, at risk and vulnerable populations can be identified. This can aid in delivering tailored public health messaging and other public-health measures.

Using NLP for the analysis of social media feeds, newspaper clips, and other media content can help in environmental scanning and situational awareness. Clustering and spatial analysis along with public health notifiable disease information will also help in identify the new disease or pathogen outbreak.

These two datasets can be cleaned, processed, and combined with other datasets to train context-aware machine learning models to estimate disease prevalence and burden. The disease prevalence and burden and their projected values in the future will help the authorities for planning surge capacities in medical supplies and hospital capacities.

NLP-powered question-answering platforms/chatbots can give personalised support or advice in the relevant context. They can also help in knowing patients social needs during hospital visits.

NLP Topic Modelling and sentiment analysis, can also help in measuring the adherence/acceptance of public health measures and the concerns of the public.

Development of these NLP models will require different datasets owned by many different organisations which required to be connected. Also the data needs to be exchanged from testing laboratories, and healthcare providers.

The data exchanges between different healthcare organisations and public health systems should happen through HL7/ISO data exchange protocols.

Changing disease prevalence and burden will require new public health and clinical guidelines. Deep learning models need to be continuously trained with new/updated data and new/updated public health guidelines.

For NLP systems to remain relevant, this entire process from data acquisition to training of deep learning models should be automated. Automation of these will ensure, that Question & Answer virtual assistants are always up to date.

AI adoption in organisations has increased by many folds in the last two years. For sure, AI is going to change healthcare jobs for the better. The insights and visualisations. It is going to amplify the healthcare workers decision making. Future applications of AI is going to be more towards Explainable AI (XAI). Bring XAI is critical for building trust in medical AI as humans whether patients or physicians need to understand the rationale behind the decisions. XAI is also going to accelerate adoption of AI in public health and healthcare organisations.

By Satish Kumar, CEO Suparna Systems

(DISCLAIMER: The views expressed are solely of the author and ETHealthworld does not necessarily subscribe to it. ETHealthworld.com shall not be responsible for any damage caused to any person / organisation directly or indirectly.)


Source link

Leave a Reply

Your email address will not be published.

%d bloggers like this: