Transcript Document
Status and potential for further collaboration with RSNA QIBA QIN Meeting, March 28, 2014 D. Sullivan, MD Duke University; RSNA Premise • Variation in clinical practice results in poorer outcomes and higher costs. RSNA’s Perspective: • Extracting objective, quantitative results from imaging studies will improve the value of imaging in clinical practice. Quantitative Imaging Biomarkers Alliance (QIBA): Background • Started in 2007 • Mission: Improve value and practicality of quantitative imaging biomarkers by reducing variability across devices, patients, and time. – “Industrialize imaging biomarkers” QIBA Criteria for Biomarker Selection • Transformational – addresses a significant medical need • Translational – will likely result in significant improvement in the development, approval, or delivery of care to patients. • Feasible – end goals can likely be achieved in a specific timeframe • Practical – leverages preexisting resources (e.g., intellectual capital, personnel, facilities, specimens, reagents, data) wherever possible; warrants access to RSNA resources and support. • Collaborative – the biomarker has the support of the stakeholder community and the organizational impetus to sustain continued efforts. QIBA Committees Quantitative Magnetic Resonance Imaging [Q-MR] Perfusion, Diffusion, and Flow-MRI (PDF-MRI) Functional MRI (fMRI) Quantitative Computed Tomography [Q-CT] CT Volumetry in Solid Tumors and Lung Nodules CT Densitometry in COPD Airway Morphology in Asthma Quantitative Nuclear Medicine [Q-NM] FDG-PET SUV Amyloid-PET Quantitative Ultrasound [Q-US] Shear Wave Speed for liver fibrosis Imaging Assays Assays are characterized by their: • Technical Performance • Clinical Performance Clinical validation Clinical utility QIN Variability in imaging measurements is related to: 1. Image acquisition variability 2. Radiologist/Reader variability 3. Measurement method variability 8 QIBA Profiles A QIBA Profile describes a specific performance Claim and how it can be achieved. QIBA Claim Template • List Biomarkers/Measurand(s) • Specify: Cross-sectional vs. Longitudinal measurement • List Indices: – Bias Profile (Disaggregate indices) – Precision Profile • Test-retest Repeatability (Repeatability coefficient) • Reproducibility (Reproducibility coefficient; Intra-class Correlation Coefficient [ICC]; Concordant Correlation Coefficient [CCC]). – Specify conditions, e.g., » Measuring System variability (hardware & software) » Site variability » Operator variability (Intra- or Inter-reader) • Clinical Context True Biologic Change … • … is approximately twice the variability • Clinical Significance of that change needs to be determined by clinical studies. Topics for Collaboration Discussion: Reducing variability in imaging measurements is important to both QIN and QIBA: 1. Image acquisition variability a) Test objects – physical and virtual 2. Radiologist/Reader variability 3. Measurement method variability a) Algorithm comparisons 1 3 Measurement method variability How do we deal with the fact that different algorithms that purport to measure the same thing give different answers? Methodology for comparing algorithms Metrics of performance on same task Criteria for acceptability (compliance). 1 4 Thank you.