Variability in the interpretations of clinical imaging studies is a problem that has been recognized for decades and has been extensively documented in the radiology literature. It’s a problem that pervades all imaging modalities. Patients should get the same result if they go to the radiology department any day of the week. Sadly, that is too often not the case. The reasons for this day-to-day variability are complex and reflect the use of different scanners, software, technologists, local operating procedures, and different radiologists. However, to be in synch with the continuing emergence and maturity of precision medicine, and to meet the expectations of referring physicians and our patients, the field of radiology must more strongly strive to improve the reproducibility of clinical imaging results for each individual patient.

One of the generally accepted benefits of precision medicine is to monitor a patient’s response to therapy and adjust, tailor, or change the patient’s health care plan according to the degree of or lack of response. Unfortunately, the traditional subjective, qualitative interpretation of clinical imaging examinations, based on visual inspection of the images, results in marked interreader and intrareader variability. This frequently makes it difficult to be confident across serial scans when determining whether a given patient’s condition has improved, worsened, or stayed the same.

One strategy to reduce variability is to extract objective, reproducible, quantitative results from clinical imaging scans. Since all clinical imaging studies today are digital, this is feasible. One clinical setting where referring physicians particularly want objective measurements delineating change in the burden of disease is oncology. Oncologists want objective measurements of both tumor size (whether from CT or MRI) and metabolic activity (from FDG-PET scans).

Reproducible, quantitative standardized uptake value (SUV) results from FDG-PET scans are increasingly viewed as important in clinical oncology—both in routine clinical practice, as well as in clinical trials. In 2010, we surveyed several hundred oncologists at National Cancer Institute-funded cancer centers about tumor measurements. Ninety-four percent expected tumor size measurements to be provided routinely, and more than half also expected SUV to be provided from FDG-PET scans.

Recommendations to use FDG-PET scans as part of the staging workup for solid tumors have been included in several chapters of the eighth edition of the American Joint Committee on Cancer’s Cancer Staging Manual. These panels of expert oncologists recognize that SUV from FDG-PET scans likely conveys important diagnostic and/or prognostic information, but lack of standardization makes it impossible at present to determine appropriate thresholds or cut-points to guide clinical decision-making. They therefore recommend that all FDG-PET scan reports in breast cancer patients should contain SUV values for the primary tumor and SUV for the primary tumor, as well as hilar and mediastinal nodes in patients with lung cancer [6]. They further recommend that these SUV values be extracted from the medical record by cancer registrars, so that large databases can be developed to determine relevant thresholds. It is sobering to consider that another medical specialty is promulgating recommendations to collate imaging data to inform their clinical decision-making. One could argue that the radiology profession should be initiating such data collection and analysis.  

In 2018, the American College of Radiology (ACR) Quality Measures Technical Expert Panel, recognizing the increasing clinical importance of objective SUV measurement, approved a quality performance measure entitled “Use of Quantitative Criteria for Oncologic FDG PET Imaging”, which says, in part: “Final reports for FDG PET scans should include at a minimum…at least one lesional SUV measurement OR diagnosis of ‘no disease-specific abnormal uptake.’” In other words, providing an accurate SUV result for every patient with cancer is now an expected performance measure by the ACR. Obtaining accurate and reproducible SUV measures requires attention to a range of specifications that target hardware, software, personnel, and procedures. In 2007, the Radiological Society of North America (RSNA) formed the Quantitative Imaging Biomarkers Alliance (QIBA). QIBA now has more than 20 committees developing standards, called Profiles, for a variety of quantitative imaging biomarkers. One QIBA Profile deals with SUV from FDG-PET scans. Rigorous attention must be paid to all potential sources of variance to obtain reproducible, clinically meaningful SUV results.

• 

Adherence to these specifications is entirely possible in nuclear medicine departments that prioritize the quality of their results. Conformance to these specifications would lead to a significant improvement in the reproducibility of SUV measurements, thus greatly improving their clinical usefulness. This will translate into a major benefit to patients in this era of precision medicine.

The comments in this article are focused on the need for accurate and reproducible quantitative results in oncologic FDG-PET scans; however, the medical literature clearly supports the need for similar reproducible quantitative imaging in several other clinical areas ¹Wallis RS, Maeurer M, Mwaba P, et al. Tuberculosis—advances in development of new drugs, treatment regimens, host-directed therapies, and biomarkers. Lancet Infect Dis 2016; 16:e34–46
Loomba R. Role of imaging-based biomarkers in NAFLD: recent advances in clinical application and future research directions. J Hepatol 2018; 68:296–304
Schrantee A, Ruhé HG, Reneman L. Psychoradiological biomarkers for psychopharmaceutical effects. Neuroimaging Clin N Am 2020; 30:53–63
Mobasheri A, Saarakkala S, Finnilä M, Karsdal MA, Bay-Jensen AC, van Spil WE. Recent advances in understanding the phenotypes of osteoarthritis. F1000Res 2019; 12:2091
Thomas MR, Lip GY. Novel risk markers and risk assessments for cardiovascular disease. Circ Res 2017; 6:133–149. All current clinical images contain quantitative information. We must make use of the digital techniques available to us to extract that quantitative data and make radiology more precise.