This post has been republished via RSS; it originally appeared at: New blog articles in Microsoft Community Hub.
Azure AI Health Insights: New built-in models for patient-friendly and radiology insights
Azure AI Health Insights is an Azure AI service with built-in models that enable healthcare organizations to find relevant trials, surface cancer attributes, generate summaries, analyze patient data, and extract information from medical images.
Earlier this year, we introduced two new built-in models available for preview. These built-in models handle patient data in different modalities, perform analysis on the data, and provide insights in the form of inferences supported by evidence from the data or other sources.
The following models are available for preview:
- Patient-friendly reports model* This model simplifies medical reports and creates a patient-friendly simplified version of clinical notes while retaining the meaning of the original clinical information. This way, patients can easily consume their clinical notes in everyday language. Patient-friendly reports model is available in preview.
- Radiology insights model* This model uses radiology reports to surface relevant radiology insights that can help radiologists improve their workflow and provide better care. Radiology insights model is available in preview.
Simplify clinical reports
Patient-friendly reports is an AI model that provides an easy-to-read version of a patient’s clinical report. The simplified report explains or rephrases diagnoses, symptoms, anatomies, procedures, and other medical terms while retaining accuracy. The text is reformatted and presented in plain language to increase readability. The model simplifies any medical report, for example a radiology report, operative report, discharge summary, or consultation report.
The Patient-friendly reports model uses a hybrid approach that combines GPT models, healthcare-specialized Natural Language Processing (NLP) models, and rule-based methods. Patient-friendly reports also uses text alignment methods to allow mapping of sentences from the original report to the simplified report to make it easy to understand.
The system uses scenario-specific guardrails to detect hallucinations, omissions, and any other ungrounded content and does several steps to ensure the full information from the original clinical report is kept and no new additional information is added.
The Patient-friendly reports model helps healthcare professionals and patients consume medical information in a variety of scenarios. For example, Patient-friendly reports model saves clinicians the time and effort of explaining a report. A simplified version of a clinical report is generated by Patient-Friendly reports and shared with the patient, side by side with the original report. The patient can review the simplified version to better understand the original report, and to avoid unnecessary communication with the clinician to help with interpretation. The simplified version is marked clearly as text that was generated automatically by AI, and as text that must be used together with the original clinical note (which is always the source of truth).
Figure 1 Example of a simplified report created by the patient-friendly reports model
Improve the quality of radiology findings and flag follow-up recommendations
Radiology insights is a model that provides quality checks with feedback on errors and mismatches and ensures critical findings within the report are surfaced and presented using the full context of a radiology report. In addition, follow-up recommendations and clinical findings with measurements (sizes) documented by the radiologist are flagged.
Radiology insights inferences, with reference to the provided input that can be used as evidence for deeper understanding of the conclusions of the model. The radiology insights model helps radiologists improve their reports and patient outcomes in a variety of scenarios. For example:
- Surfaces possible mismatches. A radiologist can be provided with possible mismatches between what the radiologist documents in a radiology report and the information present in the metadata of the report. Mismatches can be identified for sex, age and body site laterality.
- Highlights critical and actionable findings. Often, a radiologist is provided with possible clinical findings that need to be acted on in a timely fashion by other healthcare professionals. The model extracts these critical or actionable findings where communication is essential for quality care.
- Flags follow-up recommendations. When a radiologist uncovers findings for which they recommend a follow up, the recommendation is extracted and normalized by the model for communication to a healthcare professional.
- Extracts measurements from clinical findings. When a radiologist documents clinical findings with measurements, the model extracts clinically relevant information pertaining to the findings. The radiologist can then use this information to create a report on the outcomes as well as observations from the report.
- Assists generate performance analytics for a radiology team. Based on extracted information, dashboards and retrospective analyses, Radiology insights provides updates on productivity and key quality metrics to guide improvement efforts, minimize errors, and improve report quality and consistency.
Figure2 Example of a finding with communication to a healthcare professional
Figure 3 Example of a radiology mismatch (sex) between metadata and content of a report with a follow-up recommendation
Get started today
Apply for the Early Access Program (EAP) for Azure AI Health Insights here.
After receiving confirmation of your entrance into the program, create and deploy Azure AI Health Insights on Azure portal or from the command line.
Figure 4 Example of how to create an Azure Health Insights resource on Azure portal
After a successful deployment, you send POST requests with patient data and configuration as required by the model you would like to try and receive responses with inferences and evidence.
Do more with your data with Microsoft Cloud for Healthcare
With Azure AI Health Insights, health organizations can transform their patient experience, discover new insights with the power of machine learning and AI, and manage protected health information (PHI) data with confidence. Enable your data for the future of healthcare innovation with Microsoft Cloud for Healthcare.
We look forward to working with you as you build the future of health.
- Learn more about Azure AI Health Insights.
- Learn more about Azure Health Data Services.
- Learn more about Microsoft Cloud for Healthcare.
- Learn more about how health companies are using Azure to drive better health outcomes.
- Learn more about Azure AI Health Insights and how to start working with this Azure resource in the Azure AI Health Insights documentation.
*Important
Patient-friendly reports models and radiology insights model are capabilities provided “AS IS” and “WITH ALL FAULTS.” Patient-friendly reports and Radiology insights aren’t intended or made available for use as a medical device, clinical support, diagnostic tool, or other technology intended to be used in diagnosis, cure, mitigation, treatment, or prevention of disease or other conditions, and no license or right is granted by Microsoft to use this capability for such purposes. These capabilities aren’t designed or intended to be implemented or deployed as a substitute for professional medical advice or healthcare opinion, diagnosis, treatment, or the clinical judgment of a healthcare professional, and should not be used as such. The customer is solely responsible for any use of Patient-friendly reports model or Radiology insights model.