Estimation of the most appropriate threshold for lesion delineation and volumetric quantification in 18F-FDG PET

Abstract EN

The function formulation to estimate the most appropriate percentage value of maximum Standardized uptake value as a threshold for automatic lesion delineation in PET/CT exam considering lesion to background ratio and the lesion volume determined by computed tomography "CT" tissue segmentation software was investigated.

The phantom consists of two parts: a hollow cylinder of a diameter (18.6 cm), length (21.6 cm), and multiple spheres.

The sphere assembly contains 4 spheres of different volumes.

The phantom was filled with water containing 18F-Fluorodeoxyglucose (18F-FDG) forming the sphere versus background activity concentration (μCi/ml) ratios 3:1, 5:1, 8:1, 10:1, 12:1, 14:1 and15:1.

For each scan, the sphere set is aligned in the center of the field of view "FOV" and the time per bed position is similar to that of the routine work 3min.

Graphs were plotted of standardized uptake value threshold versus lesion-size at different sphere/background ratio using MS Excel.

There was a logarithmic decrease in threshold with an increase in lesion size.

The most appropriate SUV threshold for each lesion at different lesion/background was estimated from the plotted graphs.

Four tests namely, (5:1, 8:1, 12:1 and 15:1) were used to generate a 3D general function to estimate threshold using the contrast ratio and lesion volume as inputs and the other three tests (3:1, 10:1 and 14:1) were used for the function validation.

The Normalized mean square error (NMSE) is used to measure the convergence between estimated and measured data set.

The created logarithmic functions offer an easy and simple way to estimate the percentage value of SUV max as a threshold using the lesion/background ratio and sphere volume(or Diameter) from CT images as inputs to be used for lesion delineation and volume determination in PET images that were verified and the NMSE=6.2.

It could be concluded that the created formula is useful and easy to estimate the most appropriate SUV threshold for automatically definite lesion volume.