Northwestern Medicine

Cardiology Neurology Oncology

Overview

Northwestern Medicine has developed a generative AI system, known as the Automated Radiology Interpretation Evaluation System (ARIES), designed to assist radiologists. This tool is intended for use by radiologists in various care settings, particularly those dealing with high volumes of imaging studies across an integrated health system.

ARIES integrates into the clinical workflow by analyzing X-rays and CT scans to generate draft radiology reports that are approximately 95% complete. Radiologists then review and finalize these reports. The system also includes a feature that monitors for critical findings and cross-references them with patient records, alerting radiologists to serious conditions.

Notable capabilities include its ability to process all types of X-rays and CT scans, rather than focusing on a single condition. Studies have indicated that the tool can increase report completion efficiency, with some radiologists experiencing significant gains. Northwestern Medicine is also working on expanding ARIES to support MRI and ultrasound imaging.

Reviewed by Pouyan Golshani, MD — Interventional Radiologist

Key Features

  • Accelerated image interpretation
  • Enhanced diagnostic accuracy
  • Generative AI capabilities
  • Potential for early disease detection
  • Streamlined radiology workflows

Use Cases

  • Assisting radiologists in image analysis
  • Improving diagnostic efficiency
  • Supporting clinical decision-making in radiology
  • Reducing radiologist workload
  • Enhancing patient care through faster diagnoses

What Physicians Need to Know

Evidence Base
The AI model was developed in-house by Northwestern Medicine physicians and engineers using clinical data from their 11-hospital network. It was built from scratch, rather than adapting existing large language models. The system also incorporates natural language processing (NLP) to analyze radiology reports.
Clinical Validation Studies
A major clinical study across an 11-hospital Northwestern Medicine network analyzed nearly 24,000 radiology reports over five months. The study showed an average 15.5% increase in report completion efficiency, with some radiologists achieving gains as high as 40%, without compromising accuracy. Unpublished follow-on work indicates efficiency gains of up to 80% for CT scans. The AI system also demonstrated the ability to flag serious conditions like pneumothorax with a sensitivity of 72.7% and specificity of 99.9% among almost 98,000 screened studies. Priority flags were available in a median of 24 seconds, compared to 24.5 minutes for radiologist notification.
Alert Fatigue Management
The AI system is designed to identify critical findings and cross-check them with patient records, issuing alerts to radiologists for serious detections. For incidental findings, the AI triggers alerts in the EHR to the ordering physician and tracks completion of recommended follow-ups, sending additional alerts if action isn't taken. This system aims to streamline care coordination and reduce the burden on clinicians.
Differential Diagnosis Support
The AI model takes a holistic approach, reviewing an entire X-ray or CT scan and producing a report that is approximately 95% complete and tailored to the patient and radiologist's preferred reporting style. It can identify serious problems like pneumothorax before a radiologist even opens the image. The team is also adapting the AI model to detect potentially missed or delayed diagnoses, such as early-stage lung cancer.
Clinical Workflow Integration
This generative AI radiology tool is integrated into the real clinical workflow, making it the first of its kind. The AI system drafts reports, which radiologists then review and finalize. For incidental findings, the AI integrates with the EHR to trigger Best Practice Advisories (BPAs) that display findings and recommended follow-ups directly in the ordering physician's workflow. The system also notifies patients via their online portal and involves a dedicated follow-up team for those without an established physician or portal access.
Decision Audit Trail
The system tracks the completion of recommended follow-ups for both incidental and expected findings. If follow-up action is not taken, another alert is sent.
Physician Tip

Leverage the AI-generated draft reports as a significant head start, allowing you to focus on critical review and personalization. Pay close attention to the AI's alerts for urgent findings, as these are designed to accelerate triage and potentially life-saving interventions. Remember that the AI is a tool to augment, not replace, your clinical judgment; your expertise remains the gold standard for final diagnosis and treatment decisions. Utilize the system's follow-up tracking for incidental findings to ensure comprehensive patient care and reduce the risk of missed diagnoses.

The AI system is deeply integrated with the Electronic Health Record (EHR) to provide alerts and notifications directly within the physician's workflow. It also cross-checks critical findings with patient records. The in-house development approach allows for tailored integration within the Northwestern Medicine network and its specific clinical data. Future expansions are planned to support MRI and ultrasound imaging.

Details

Category Clinical Decision Support & Reference, Radiology & Imaging AI
Pricing Unknown
DeploymentOn-premise (within their 11-hospital network)
Mobile App1
Compliance
BAA Available Yes AI-estimated
HIPAA Compliant Yes AI-estimated
FDA Status Unknown
GDPR Yes AI-estimated
Integrations
EHR Not specified
Specialties Cardiology, Neurology, Oncology, Radiology

Support & Reliability

Ratings & Reviews

No reviews yet. Be the first to review this tool!

Rate Northwestern Medicine

Clinical Value
Ease of Use
Integration
Support & Docs
Value for Money
No press coverage or publications on file yet. Know of some? Suggest an edit →

Videos

Product demos, reviews, and walkthroughs for Northwestern Medicine.

View all on YouTube

Frequently Asked Questions

Northwestern Medicine's CDS system is designed to integrate directly into our Epic EHR, providing real-time alerts and recommendations within the physician's existing workflow. We continuously monitor compliance rates through system analytics, and while specific numbers vary by specialty and alert type, we strive for high adoption through user-friendly interfaces and evidence-based recommendations.
While our primary CDS is integrated within Epic, some departments may utilize specialized third-party tools for specific clinical areas. These alternatives undergo rigorous vetting for efficacy and data security, adhering to all HIPAA regulations and Northwestern Medicine's stringent data governance policies. However, the integrated system offers the most comprehensive and unified approach to patient care across the enterprise.
There are no direct costs or fees passed on to physicians for utilizing Northwestern Medicine's core CDS tools. These systems are considered part of our standard clinical infrastructure and are funded centrally to support optimal patient care and operational efficiency. Any advanced features or specialized modules are typically integrated and covered under the same operational budget.
While our CDS is highly effective, limitations can include the inherent challenges of diagnostic accuracy in complex cases, particularly for very rare diseases where data may be limited. We actively work to mitigate alert fatigue through intelligent alert prioritization and customization options, and we continuously refine algorithms to improve applicability across a broader range of clinical scenarios.
Physicians can provide feedback through dedicated channels within the EHR system, as well as through their departmental leadership and our IT support teams. We have a structured process for feature requests and modifications, which involves review by clinical informatics committees to ensure alignment with evidence-based practices and overall system goals.
Northwestern Medicine provides comprehensive training through various modalities, including online modules, in-person workshops, and one-on-one support from clinical informaticists. For new updates or functionalities, targeted communications, quick guides, and refresher training sessions are regularly offered to ensure physicians can optimize their use of the CDS system.
Our CDS system is designed to incorporate the latest clinical guidelines and research through a continuous review process by our clinical informatics and medical staff committees. Updates to evidence-based recommendations are typically pushed out on a regular basis, often quarterly or as significant new evidence emerges, to ensure the most current and effective guidance is available to physicians.

Related Tools

Stanford University
Stanford University
Clinical Decision Support & Reference
Stanford Online's "AI in Healthcare" program provides a comprehensive overview of artificial intelligence's applications and implications within the medical field.
RADIN Health
RADIN Health
Clinical Decision Support & Reference
RADINA is an AI-powered platform by RADIN Health designed to optimize radiology workflows, enhance diagnostic accuracy, and streamline reporting.
Med-PaLM 2
Google
Clinical Decision Support & Reference
Med-PaLM 2 is a large language model from Google, fine-tuned for healthcare to provide high-quality answers to medical questions and support clinical decision-making.
Pacmed Critical Discharge
Pacmed
Clinical Decision Support & Reference
Pacmed Critical Discharge is an AI tool that predicts the probability of death or readmission within seven days of ICU discharge, providing explainable insights for physicians.
AiZtech Labs
AiZtech Labs
Clinical Decision Support & Reference
AiZtech Labs provides enterprise-grade AI solutions for healthcare, focusing on operational efficiency, diagnostic accuracy, and patient care workflows for providers and health systems.
Stethophone Pro
Sparrow Acoustics
Cardiology AI
Stethophone Pro transforms smartphones into advanced cardiac screening tools, enabling clinicians to capture, interpret, and visualize heart sounds for early detection of cardiovascular diseases.

See all Clinical Decision Support & Reference tools →

Suggest an Edit → | Report an issue | Last Verified: 2026-07-03 | First Added: 2026-06-21
AI Tool Finder
AI-powered search. Results may not be comprehensive.