Azure AI Health Insights: New built-in models for patient-friendly reports and radiology insights
New built-in models, available for preview, 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.
Empowering radiologists with clinical guidance, quality standards, and scoring and assessments
In March 2024, we announced general availability of Radiology Insights, a new model built in Azure AI Health Insights service. This model uses radiology reports to surface relevant insights that can help radiologists enhance the quality of their reports. Today, we are announcing three new capabilities added to the Radiology Insights model: clinical guidance, quality measures, and scoring and assessment. Each plays a crucial role in enhancing clinicians’ decision-making, improving healthcare quality, and standardizing evaluations in medical imaging. Clinical Guidance Clinical guidance uses an evidence-based approach based on industry guidelines (ACR Guidelines [1,2,3,4] and Fleischner Society Guidelines [5] ) to help radiologists make the most appropriate recommendations and timing for future actions, such as specific follow-up studies. Clinical guidance extracts clinical finding information from the documentation, retrieving the necessary evidence to support proposed recommendations. If no follow-up action can be proposed, clinical guidance will identify what information is missing from the documentation. Figure 1 – Clinical Guidance: In this example, only two findings are considered and highlighted, each serving as a trigger for the pulmonary nodule clinical guideline. Radiology Insights model response: The first finding in the report proposes two candidate recommendations, including their modality and anatomy. The details of the first finding, LOBE and SIZE, are present in the report and extracted by the model. The second finding does not lead to a recommendation proposal due to missing information (Size). The first finding is ranked higher because of its greater information depth. Quality Measures Quality measures guidelines are an essential tool in healthcare to monitor the quality of care by providing frameworks (MIPS QCDR measures [6]) for measuring, reporting, and continuously improving healthcare practices. These measures are now supported by the Radiology Insights model to ensure healthcare providers meet established quality standards. The model captures quality measures criteria explicitly documented in the report and checks if all criteria necessary to meet quality standards are included. For each quality measure there are three possible outcomes: The report meets the required criteria: the performance is 'met'. The report does not meet all the criteria: the performance is 'not met'. The report does not meet all the required criteria: the model states an ‘exception’. For example, if a patient is allergic to Chlorhexidine, a substance used to meet quality measures for Central Venous Catheter (CVC) insertion, the standard procedure cannot be followed. The model will recognize this as an exception. When quality measure criteria are missing from the report, the documentation could be updated to include this information, or a review by a healthcare professional could be conducted to understand why these important criteria were not documented. Quality measures play a vital role in ensuring healthcare providers adhere to high standards, ultimately improving patient care. By following these measures, healthcare providers can avoid complications and deliver better outcomes for their patients. Figure 2 – Quality Measures: In this example, findings about a new nodule are considered and highlighted by the model. The model triggers the quality measure for incidental pulmonary nodule. Radiology Insights model response: For the quality measure ‘Incidental Pulmonary Nodule’, the performance is 'met'. The model surfaces the criteria for having a follow-up recommendation in the report, which is the only criterion required for compliance. Scoring and Assessment Scoring and assessment systems [7,8] are used in medical imaging and diagnostics to help standardize the evaluation and reporting of findings and provide a structured approach to interpreting imaging studies, assessing disease risk, and guiding clinical management. The Radiology Insights model surfaces and highlights scoring and assessments with their classifications or values that were explicitly documented by radiologists in their reports. In the sample below, the model identifies two assessments with its values: the ASCVD (Atherosclerotic Cardiovascular Disease) risk and the Agatston Score, which measures the amount of calcium in the coronary arteries. In this example, there is a 17.6% chance of experiencing a cardiovascular event in the next 10 years (ASCVD) and Agatston Score of 0 suggests low short-term risk of heart attack. Figure 3 – Scoring and Assessment: In this example, two assessments, ASCVD and Agatston score, are reported with their values. Radiology Insights model response: The model surfaces two scoring and assessment instances, one of category ASCVD Risk with a value of 17.6% and one of category Calcium Score with value 0. 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 and how to start working with this Azure resource in the Azure AI Health Insights documentation Learn more about Radiology Insights Important Radiology Insights is a capability provided “AS IS” and “WITH ALL FAULTS.” Radiology Insights isn’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. This capability isn’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 the Radiology Insights model.
