Articles

In the quest to advance cancer care, oncology researchers face many complex challenges. Leveraging medical imaging data across diverse populations can help researchers develop studies to meet these challenges. But in the face of siloed data, data shortages and stringent privacy regulations, that’s easier said than done. 

Here, we’ll uncover some of the top challenges oncology researchers currently face—and how enhancing your data management strategy can help. 

1. Tumor Heterogeneity  

No two tumors are exactly alike, regardless of pathophysiology. This variability complicates diagnosis and treatment decisions, particularly in cancers caused by rare mutations where access to sufficient data is prohibitive. In addition, imaging data may vary greatly across institutions due to differences in equipment, protocols and data formats. 

An enterprise Oncology data management strategy should support multi-modal imaging such as MRI, PET and CT, as well as relevant metadata from the EMR and genomics systems. This ensures your ability to curate large datasets representing diverse patient populations, with the end goal of enabling longitudinal research to track tumor changes over time. 

2. Early Detection and Accurate Diagnosis 

Early detection is key for successful cancer care, but it’s difficult for clinicians to detect tumors at early stages, especially in asymptomatic patients. In fact, only about 50% of cancers are detected at an early stage.  

A well-managed, centralized imaging data store empowers researchers to develop and train AI models aimed at detecting subtle early biomarkers. In addition, clean and well-annotated data accelerates these efforts. An example of this is the early identification of distribution and activity of cytotoxic T cells within tumors using CD8 imaging.  

3. Radiomics and AI 

Radiomics, or extracting quantitative features from medical images, are highly valuable in oncology space. These calculations help researchers and clinicians tap into the power of artificial intelligence (AI) to improve diagnosis, make more accurate prognoses and optimize clinical decisions to deliver precision medicine. However, radiomic pipeline development requires curated, high-quality imaging datasets with complete tumor segmentation and feature extraction.  

A robust imaging data platform can help researchers aggregate the large-scale imaging data needed to fuel development of AI models that can predict treatment response, disease progression and patient outcomes. The Flywheel medical imaging platform also enables radiomics with the Gear Exchange, which includes prepackaged algorithms that support feature extraction from 2D and 3D images. 

4. Data Privacy and Sharing 

Privacy compliance is paramount when handling imaging data with PHI (protected health information). But privacy and data sharing are often at odds, especially when multiple institutions need to collaborate on research. In fact, studies at three major NCI-designated Comprehensive Cancer Centers found 25%-50% error rates in compliance with clinical trial imaging protocols prior to implementing a comprehensive clinical trials imaging informatics platform. 

A secure, cloud-based imaging platform that allows you to automate de-identification tasks and enact role-based access to datasets can help you adequately meet data privacy regulations while enabling remote access and secure data sharing between institutions. A data platform such as Flywheel also can support 21 CFR Part 11 compliant workflows, with project locking, change logging and e-signatures, to enable submittal to the FDA and other regulatory bodies. 

5. Cost and Resource Efficiency 

Inefficiencies in data management drive up the cost of oncology research and at times make critical studies infeasible. Sifting through scattered imaging files, manually curating datasets and duplicating efforts waste valuable time and resources. 

A streamlined imaging data management platform like Flywheel is built to reduce redundancy and optimize resource allocation. For example, one enterprise-level cancer center has used Flywheel to index more than 2 million DICOM studies representing 328,000 patients so they can identify and create cohorts for tumor segmentation model training, dramatically reducing the time it would take to otherwise sift through all that data. 

What challenges are you currently facing in oncology imaging data management? 

At Flywheel, we’re always interested in chatting with folks in the field to learn what their greatest challenges are around imaging data management. We’ve been fortunate enough to work with five out of the top 10 top cancer centers, as ranked by Newsweek, who use Flywheel to streamline annotation and tumor segmentation, conduct blind reader studies, automate de-identification, and securely share data with internal and external collaborators.  

What challenges are you currently facing in your oncology research? We’d love to hear from you. Get in touch to start solving your greatest challenges around oncology imaging data management.