ABSTRACT
In clinical oncologic practice, positron emission tomography/computerized tomography (PET/CT) is used commonly to evaluate tumour response to therapy as well as for diagnosis, staging and prognosis. The correct interpretation of PET/CT images requires a knowledge of the possible pitfalls that may occur due to normal variation, artefacts and processes which may mimic pathology. Especially in the use of PET/CT in tumour response monitoring to treatment, standardization of technical and biological factors such as scanner calibration, imaging parameters, applied activity dose, plasma glucose levels are required for good image quality and accurate quantification. The standardized uptake value (SUV) is the most widely used semi-quantitative parameter for determining tumor uptake. Due to many factors that affect SUV, PET/CT scans by nonstandardized protocols in multicenter studies will result in unknown biases for reproducibilities of SUVs and SUV-based response measures. Several recommendations and guidelines have been proposed with the aims of improving the image quality and the quantitative accuracy of for quantitative F-18 fluorodeoxyglucose (FDG) PET studies. Recently, with the development of artificial intelligence algorithms, a more holistic image-based decision making mechanism has emerged with the help of tools such as multimodal imaging and data mining. This review provides an overview of the recommendations suggested in the guidelines, drawing attention to the underlying causes and remedies.