We live in an era of unprecedented increases in the volume of health data. The digitization of medical records, medical imaging, genomic data, clinical notes, and more has contributed to an exponential increase in the amount of medical data. The potential benefits of taking advantage of this health data are enormous. However, with this growth in health data, new challenges emerge, including a focus on data privacy and security, and the need for data standardization and interoperability. Effective tools are needed to extract the information buried in this data and use it to derive deep insights, conclusions, and analyzes that can make sense of the data and support clinicians.
Today, I’m excited to announce the Project Health Insights Preview. Project Health Insights is a service that derives insights from patient data and includes pre-built models intended to power key high-value health scenarios. Models receive patient data in various ways, perform analysis, and enable clinicians to draw conclusions and insights with evidence from the input data. These insights can help healthcare professionals make sense of clinical data, such as patient descriptions, clinical trial matches, and more.
Project Health Insights – Leveraging patient data to support actionable insights
Project Health Insights supports pre-built models that receive patient data in multiple ways as input, and produce insights and conclusions that include:
- Confidence scores: The higher the confidence score, the more assertive the model is about the inference value provided.
- Evidence: Relates model output to specific evidence within the input provided, such as references to spaces of text that reflect the data that led to the insight.
Project Health Insights Preview includes two enterprise-grade AI models that can be provisioned and deployed in minutes: Oncology phenotype and clinical trial match.
Tumor phenotype It is a model that enables healthcare providers to quickly identify key cancer features among patients with their current cancer diagnosis. The model identifies cancer features such as tumor location, histology, clinical stage, tumor, nodules, metastasis (TNM) and pathologic stage categories of TNM from unstructured clinical documents.
Key features of the neoplastic phenotype model include:
- Cancer detection.
- Clinical text extraction of solid tumors.
- Rank the importance of evidence.
Matching clinical trials It is a form that matches patients with potentially appropriate clinical trials, according to trial eligibility criteria and patient data. The form helps to find relevant clinical trials, for which patients can be eligible, and also to find a pool of eligible patients who are eligible for a list of clinical trials.
Key features of the clinical trial matching model include:
- Support scenarios which are:
- Patient-centered: assisting patients in finding potentially appropriate clinical trials and assessing their eligibility according to trial criteria.
- Trial Center: Matching a trial with a database of patients to identify a group of patients who are potentially suitable.
- Interactive matching where the form provides insights into the missing information needed to narrow down the list of potential clinical trials via an interactive trial.
- Support for different approaches to patient data such as unstructured clinical notes, structured patient data, and Rapid Healthcare Interoperability Resources (FHIR)®) strap.
- Support search across embedded cognitive graphs for clinical trials from Clinictrials.gov as well as against a personalized trial protocol with specific eligibility criteria.
Simplifying clinical trial matching and cancer research
According to the World Health Organization, the number of registered clinical trials increased by more than 4,800 percent from 1999 to 2021. Today there are more than 82,000 clinical trials actively recruiting participants from around the world (based on Clinicaltrials.gov), with eligibility criteria for trials increasingly complex. . However, enrollment in clinical trials relies on manual screening of millions of patients, each with up to hundreds of clinical notes that require review and analysis by a healthcare professional, making it an unsustainable process. Given this, it is not surprising that up to 80% of clinical trials missed clinical trial enrollment schedules, and up to 48% failed to meet clinical trial enrollment goals according to data provided by Tufts University. The clinical trial matching model aims to solve exactly this problem by effectively matching patients with diverse conditions to clinical trials for which they are likely to qualify through analysis of patient data and complex clinical trial eligibility criteria.
The oncology phenotype model allows clinicians to effectively analyze cancer patient data based on tumor location, tumor histology, and cancer stages. These models provide essential building blocks for achieving the goals set out by the White House Cancer Initiative: developing and testing new therapies, sharing more data and knowledge, collaborating on tools that can benefit everyone, and making progress toward ending cancer as we know it.
Delivering value across the health and life sciences industry
Johns Hopkins University Medical Center is an early user of Project Health Insights. Dr. Srinivasan Yignasubramanian uses the oncology phenotype model to leverage unstructured data to accelerate cancer registry processing efforts for patients with solid tumors.
Pangea Data is a Microsoft partner working in the field of health artificial intelligence. “At Pangea Data, we help companies detect 22 times more undiagnosed, misdiagnosed, and deformed patients by profiling them by unlocking and summarizing clinically valid, actionable intelligence from patient records in a privacy-preserving, scalable, federal way. We are exploring using Project Health Insights to augment our advanced patient recognition capabilities. Vibhor Gupta, Director and Founder of Pangea Data
Akkure Genomics helps patients use their genomic data or DNA to improve their chances of finding a clinical trial. “At AKKURE GENOMICS, we leverage Project Health Insights, which enables the capabilities of our digital AI and DNA platform, to help patients match clinical trials based on individual medical diagnoses, thus enhancing enrollment, improving chances of finding an accurately matching trial and accelerating discovery of therapeutics. and new treatments. Professor Oran Rigby, Head of Engineering and Founder of Akkure.
Built with the end user in mind
Prototypes have been validated in a research environment through a strategic partnership between Microsoft and Providence to accelerate digital transformation in the health and life sciences. These models can enable oncologists to dramatically increase their micro-oncology capabilities and generate useful intelligence and insights for clinicians as well as patients.
“Microsoft’s ability to structure complex concepts using its natural language processing tools for cancer has greatly contributed to our ability to build research groups and discuss cancer treatment options— Dr. Carlo Bivolco, chief medical officer, Providence Genomics.
Microsoft will continue to expand capabilities within Project Health Insights to support additional health workloads and enable insights that will guide key healthcare decision-making.
Microsoft continues to grow its portfolio of AI services for health
Microsoft continues to invest in AI services in health and life sciences. Along with other new offerings in Microsoft Cloud for Healthcare, we are excited to announce new improvements in Text Analytics for Health (TA4H).
New improvements include:
- Social determinants of health (SDoH) and extracting ethnic information. Newly introduced SDoH and Ethnicity features make it possible to extract social, environmental, and demographic factors from an unstructured text. These factors will enable the development of more comprehensive healthcare applications. Read more about it on our blog.
- Time affirmations – past, present and future. The ability to set the temporal context of TA4H entities whether in the past, present or future.
- Customers can now extend TA4H to support custom entities based on their own data. Customers can now also extend entities extracted by the service.
We’re also excited to share that Azure Health Bot now has a new Azure OpenAI template in preview. The Azure Health Bot OpenAI template allows customers to extend their Azure Health Bot instance with the Azure OpenAI service to respond to unknown bots in a smarter way. This feature will be enabled by the Azure Health Bot Template Catalog. Customers can choose to import this template into their bot instance using an Azure OpenAI resource key and endpoint, enabling GPT-generated backup answers from trusted, medically viable sources that can be provided by customers. This feature provides a mechanism for customers to try this capability as a preview.1 Read more about this and how to apply responsible AI principles when implementing your own Health Bot example in this blog.
We look forward to what the coming years will bring to the health and life sciences industry powered by these new capabilities and the continued innovation we are seeing through AI and machine learning. The potential for improved precision care, faster and more efficient clinical trials, and thus the availability of drugs, treatment, and medical research is unparalleled. Microsoft looks forward to partnering with you and your organizations on this journey to improve the health of humanity.
learn more
1 At the moment, we are providing the preview for internal testing and evaluation purposes only.
FHIR® is a registered trademark of Health Level Seven International, registered with the US Trademark Office and used with their permission.