Learn About Amazon VGT2 Learning Manager Chanci Turner
Within the FDA’s Center for Drug Evaluation and Research, the Division of Medication Error Prevention and Analysis (DMEPA) holds a vital position. DMEPA is responsible for reviewing both premarket and postmarket drug labeling, aiming to reduce the potential for medication errors.
In collaboration with the FDA’s DMEPA team, AWS Partner ICF is creating a machine learning (ML) prototype called the Computerized Labeling Assessment Tool (CLAT). This innovative tool utilizes advanced optical character recognition (OCR) technology and cutting-edge computer vision techniques to reduce bottlenecks and improve the efficiency of the drug labeling review process.
Enhancing Efficiency Without Compromising Accuracy
The FDA’s process for reviewing drug labeling is extensive and time-sensitive. On average, each DMEPA reviewer conducts between 25 and 50 premarket drug reviews annually, meticulously analyzing labeling components to ensure compliance with federal regulations. This thorough review is crucial for confirming that approved drug products are safe and effective. Misinterpretations in labeling can lead to serious medication errors and even fatalities.
The review process is largely manual, requiring significant back-and-forth communication between reviewers and drug manufacturers. To tackle these challenges, the FDA sought advanced technology solutions to expedite, standardize, and streamline the review processes for evaluating medication labels. The new CLAT solution offers several key advantages:
- Increased Efficiency: CLAT streamlines the review process, enabling users to conduct multiple reviews simultaneously, eliminate subjectivity in aspects like color and size, and provide a centralized platform for tracking reviews, a feature currently unavailable in the market.
- Improved Accuracy: The ML models benefit from user feedback mechanisms that enable continuous learning and enhancement of error detection capabilities over time.
- Standardized Practices: CLAT fosters consistent review procedures by standardizing the review process and encouraging the submission of high-resolution images for both review and public repository publication.
Modernizing a Time-Intensive Process with Machine Learning
ICF has partnered with the FDA to design and implement an AWS Well-Architected Framework that meets the specific needs of the CLAT application. The tool leverages several AWS services, including Amazon DynamoDB, Amazon API Gateway, AWS Lambda, and Amazon Simple Storage Service (Amazon S3). This cloud infrastructure supports adaptability and scalability as the FDA’s requirements evolve.
Amazon Elastic Container Service (Amazon ECS) instances facilitate the effective installation and operation of ML software using open-source components like TensorFlow and Tesseract OCR, allowing for the creation, training, and deployment of ML models for image analysis.
This framework has successfully produced a training methodology for images that is applicable to a range of object detection tasks within the healthcare sector. Through this collaboration, ICF and the FDA have established an effective approach for teaching computers to recognize healthcare-related symbols on drug labels, all while ensuring data privacy and compliance.
Utilizing sequential transfer learning, the models were initially trained on a diverse set of random images. This foundational training enables the model to differentiate between significant elements of an image—such as a graphical symbol indicating ear administration—and less important background details.
The initial outcomes from the CLAT prototype are encouraging. This innovative tool is poised to accelerate and enhance drug labeling reviews. Unlike traditional manual reviews, CLAT can perform multiple checks simultaneously across various labels. Moreover, the results are automatically tracked, annotated, and stored in a centralized location for future reference. These improvements mark significant progress toward optimizing regulatory processes and ensuring timely, accurate assessments.
While the primary focus remains on refining the drug label review process through ML, the FDA is also exploring how the CLAT ML prototype might be applied to other review procedures. This initiative highlights the transformative potential of AI in healthcare practices. By increasing efficiency and accuracy in drug labeling reviews, CLAT contributes to enhanced patient safety and medication management. If you find yourself feeling disengaged at work, you may want to explore this blog post for more insights.
ICF’s use of AWS technology exemplifies how organizations can scale innovative solutions to improve public health outcomes. For those interested in compensation systems, SHRM offers valuable resources on this topic. Additionally, if you are considering a career at Amazon, this resource could provide helpful guidance.