QIBA Newsletter

In My Opinion

Quantitative Imaging in the Era of Artificial Intelligence

By CAROLINE CHUNG, MD, MSc, FRCPC  

Despite the shared aspirations to enable personalized health care delivery and precision medicine, the progress toward this goal has been slow and challenging. Precision medicine, an approach that aims to customize health care decisions, including prevention, treatments and procedures to subgroups of patients or even individual patients, is built on the assumption of precision measurement. With imaging becoming a central tool for evaluating the baseline state and response assessment of medical treatments, quantitative imaging is a critical component in enabling precision measurement in medicine. As stated on the QIBA website, “Quantitative imaging is the extraction of quantifiable features from medical images for the assessment of normal or the severity, degree of change, or status of a disease, injury, or chronic condition relative to normal.”

However, amidst the exponential rise in the generation of medical data each year, with medical imaging data making up approximately 30% of this volume, the small fraction of information extracted for clinical impact from this growing collection of health care data remains largely inconsistent and unstructured. The human-centered workflows for data use have largely remained unchanged during periods of major digital transformation. As a result, there are growing demands to maximize insights and challenges to scale-up our approach to address the tsunami of data.

AI is a promising technology that can be utilized to support the immediate demands of clinical workflow, enhancing human performance while also raising the potential to support additional discoveries and insights from the advanced data we are gathering at each imaging session. Radiology has been at the forefront of computer assisted diagnosis and early demonstration of the potential utilization of AI in health care has utilized radiological imaging data.

As the world gains experience and the hype of AI and big data starts to settle, there is growing awareness of the need for high quality and consistent data both for training and successful deployment. Beyond the models developed from the imaging data itself, it has been recognized that the uncertainty of the imaging-derived measurements integrated into predictive AI models can impact the uncertainty in the predictions themselves. To this end, QIBA can play a key role in accelerating and enhancing AI development because the steps taken to ensuring quantitative imaging also generate consistent, high quality imaging data along with the estimated uncertainty.

Imaging acquisition and processing is also being impacted by AI technologies, and it is unclear how these innovations to accelerate image acquisition will impact the QIBA Profiles and claims. As emerging AI-driven image processing algorithms may help facilitate normalization, harmonization or calibration of images to facilitate aggregation of imaging data, the coordinated efforts across QIBA, industry partners and clinical sites will be critical to aligning and updating QIBA Profiles to realize benefits to clinical impact.

With mutually beneficial and common motivations to improve the consistency and quality of image acquisition in support of quantitative imaging and AI, the QIBA community is well-positioned to collaboratively connect with the clinical, research and data science communities to realize the full power of AI and support advances toward precision medicine and personalized health care.

QIBA Resources:

• About QIBA 
• QIBA Webpage
• QIBA Wiki
• QIBA Biomarker Committees
• QIBA Organization Chart
• QIBA LinkedIn page
• QIBA Twitter page

Please contact qiba@rsna.org for more information. We welcome your participation.

Please note that the list of references has been migrated to EndNote.
To obtain access to the RSNA EndNote citations, please send an email request to: qiba@rsna.org.

The list of references showcases articles that mention QIBA, quantitative imaging, or quantitative imaging biomarkers. In most cases, these are articles published by QIBA members or relate to a research project undertaken by QIBA members that may have received special recognition.

New submissions are welcome and may be directed to qiba@rsna.org.