What is Radiology Insights?
Radiology Insights is a model that aims to provide quality checks as feedback on errors and inconsistencies (mismatches). The model ensures that critical findings are identified and communicated using the full context of the report. Follow-up recommendations and clinical findings with measurements (sizes) documented by the radiologist are also identified.
Important
The Radiology Insights model is a capability provided “AS IS” and “WITH ALL FAULTS”. The Radiology insights service isn't intended, designed, or made available: (i) as a medical device, (ii) to be used in the diagnosis, cure, mitigation, monitoring, treatment or prevention of a disease, condition or illness, and no license or right is granted by Microsoft to use the healthcare add-on or online services for such purposes, and (iii) to be a substitute for professional medical advice, diagnosis, treatment, or judgment and should not be used to replace or as a substitute for professional medical advice, diagnosis, treatment, or judgment. The customer is solely responsible for testing and evaluating whether Radiology Insights is fit for purpose and identifying and mitigating any risks or harms to end users associated with its use. Output from the Radiology insights service does not reflect the opinions of Microsoft. The accuracy and reliability of the information provided by the Radiology insights service may vary and are not guaranteed.
Radiology Insights features
To remain competitive and successful, healthcare organizations and radiology teams must have visibility into trends and outcomes. The focus is on radiology operational excellence and performance and quality. The Radiology Insights model extracts valuable information from radiology documents for a radiologist.
Identifying Mismatches: A radiologist is provided with possible mismatches. These inferences are identified by comparing what the radiologist has documented in the radiology report and the information that was present in the metadata of the report.
Mismatches can be identified for sex, age, and body site laterality. Mismatches identify potential discrepancies between the dictated text and the provided metadata. They also identify potential inconsistencies within the dictated/written text. Inconsistencies are limited to gender, age, laterality, and type of imaging.
This information enables the radiologist to rectify any potential inconsistencies during reporting. The system isn't aware of the image the radiologist is reporting on.
This model doesn't provide any clinical judgment of the radiologist's interpretation of the image. The radiologist is responsible for the diagnosis and treatment of patient and the correct documentation thereof.
Providing Clinical Findings: The model extracts as structured data two types of clinical findings: critical findings and actionable findings. Only clinical findings that are documented in the report are extracted. Clinical findings produced by the model aren't deduced from pieces of information in the report nor from the image. These findings merely serve as a potential reminder for the radiologist to communicate with the provider.
The model produces two categories of clinical findings, Actionable Finding and Critical Result, and is based on the clinical finding, explicitly stated in the report, and criteria formulated by ACR (American College of Radiology). The model extracts all findings explicitly documented by the radiologist. The extracted findings may be used to alert a radiologist of possible clinical findings that need to be clearly communicated and acted on in a timely fashion by a healthcare professional. Customers may also utilize the extracted findings to populate downstream or related systems (such as EHRs or autoschedule functions).
Communicating Follow-up Recommendations: A radiologist uncovers findings for which in some cases a follow-up is recommended. The documented recommendation is extracted and normalized by the model. It can be used for communication to a healthcare professional (physician). Follow-up recommendations aren't generated, deduced, or proposed. The model merely extracts follow-up recommendation statements documented explicitly by the radiologist. Follow-up recommendations are normalized by coding to SNOMED.
Reporting Measurements: A radiologist documents clinical findings with measurements. The model extracts clinically relevant information pertaining to the finding. The model extracts measurements the radiologist explicitly stated in the report.
The model is simply searching for measurements reviewed by the radiologist. This info is extracted from the relevant text-based record and structures. The extracted and structured measurement data may be used to identify trends in measurements for a particular patient over time. Alternatively a customer could search a set of patients based on the measurement data extracted by the model.
Reports on Productivity and Key Quality Metrics: The Radiology Insights model extracted information can be used to generate reports and support analytics for a team of radiologists.
Based on the extracted information, dashboards and retrospective analyses can provide insight on productivity and key quality metrics. The insights can be used to guide improvement efforts, minimize errors, and improve report quality and consistency.
The Radiology Insights model isn't creating dashboards but delivers extracted information. The information can be aggregated by a user for research and administrative purposes. The model is stateless.
Quality Measures: Compliance with Reimbursement Criteria in Healthcare Programs: A healthcare organization needs to ensure compliance with evolving reimbursement programs, such as the Merit-based Incentive Payment System (MIPS) established under the MACRA act. The organization must meet specific performance criteria to qualify for reimbursement incentives or avoid penalties.
The organization aggregates the appropriate Quality Measure criteria from its clinical and administrative data. The organization automates the tracking and reporting of these criteria, the system supports compliance with reimbursement requirements, reducing manual effort and ensuring accurate reporting to Medicare. The RI model is not creating dashboards but delivers extracted information, not deduced, to be aggregated by the customer. The model is stateless.
Scoring and Assessment: Population Health Management through Radiology Insights: A healthcare provider seeks to improve population health management by identifying individuals or patient groups in need of additional screening or preventative services. The organization wants to extract valuable insights from radiology reports to improve patient outcomes.
Scores and values are then aggregated to identify trends, establish baselines, and flag individuals or groups that could benefit from further screening or preventative care. The system helps clinicians proactively manage patient populations, improving preventative care while reducing the likelihood of future complications.
Clinical Guidance for the Radiologist: Radiologists frequently need to reference specific clinical guidelines when documenting findings in their reports. However, missing or incomplete information can delay access to these guidelines and its recommendations, potentially impacting the accuracy of their reports.
Radiology Insights automatically highlights documented findings in the radiology report that are relevant to the applicable clinical guidelines and its candidate recommendations. If key information is missing, which could affect the recommendation outcome of a clinical guideline, the system flags these gaps. This proactive approach ensures that radiologists have immediate access to the necessary guidelines information, enhancing the accuracy and completeness of their reports.
Language support
The service currently supports the English language.
Limits and quotas
For the Public Preview, you can select the Free F0 SKU. The official pricing will be released after Public Preview.
Next steps
Get started using the Radiology Insights model: