Tumor segmentation and quantitative assessment of glucose metabolism for different histological types of NSCLC using dynamic ¹⁸F-FDG PET

Objectives Combined radiochemotherapy is the standard treatment for irresectable stage III non-small cell lung cancer (NSCLC). For patients suspected of NSCLC, the role of ¹⁸F-FDG PET includes characterization of lung opacities as tumor or atelectasis, staging of lymph nodes and reduction of inter-observer variation of tumor delineation for radiotherapy planning. With 4D dynamic ¹⁸F-FDG PET pharmacokinetic information of tissue metabolism can be obtained and a distinction can be made between aspecific free ¹⁸F-FDG and bound ¹⁸F-FDG-6-phosphate (1). The aim of this prospective cohort study was dual. Firstly, to answer whether 4D dynamic is superior to static ¹⁸F-FDG PET for tumor delineation using pathology volumes as Gold standard (2). Secondly, to describe tumor regional variation in pharmacokinetic rate constants of ¹⁸F-FDG metabolism (K₁-k₃) and fractional blood volume (V_B) in regions with different levels of glucose metabolic rate (MR_glc) and compare these between the major histologies (3-5).

Methods One-hour dynamic PET/CT scans were acquired after a standardized injection of ¹⁸F-FDG in 38 non-diabetic, therapy-naïve patients with (suspicion of) primary resectable NSCLC of at least 30 mm in smallest diameter. Acquisition was performed as close to surgery as possible. After image reconstruction, two datasets were obtained: a static image 50-60 min post injection and a parametric image of Patlak MR_glc values. Lesions were delineated on both datasets using three strategies: (i) a threshold of 50% of the maximum voxel value, (ii) an adaptive threshold method of 40% of the background subtracted maximum voxel value and (iii) using the fuzzy locally adapted Bayesian (FLAB) algorithm (6). The pathological volume was calculated from three orthogonal tumor diameters using the equation for an ellipsoid. For the second aim, the tumor was divided in three equal volumes of increasing MR_glc, in which, K₁-k₃ and V_B were computed using non-linear least squares regression based on an irreversible 2-tissue compartment model (1).

Results 35 patients having 36 primary malignancies were available for analysis (exclusion: benign (n=1) or disappeared (n=1) lesion or severe mismatch between PET and CT (n=1)). These included 20 squamous cell carcinomas (SCC), 12 adenocarcinomas (AC) and 4 lesions with another malignant histology. Pathological volumes were significantly larger than image volumes, except for FLAB volumes in a subset of 23 patients without macroscopic necrosis. Static PET volumes were significantly larger than volumes determined on MR_glc images. The volume determined by FLAB on the static PET corresponded best with the pathological volume (median difference 8.7 cm³ (all lesions) and 3.7 cm³ (23 lesions without macroscopic necrosis)). In SCC compared to AC, lesion MR_glc and k₃ were significantly higher and V_B was significantly lower (supplemental figure). AC showed less heterogeneity relative to SCC in terms of mean MR_glc, k₃ and V_B. In SCC, a significant higher value for k₃ and lower value for V_B was found in regions with higher values of MR_glc.

Conclusions FLAB delineation on static ¹⁸F-FDG PET scans resulted in volumes in best agreement with pathology. There was significant difference in glucose metabolic rate, phosphorylation rate (k₃) and fractional blood volume between the major pathological subtypes of NSCLC. In SCC, tumor regions with higher MR_glc showed higher phosphorylation rate and lower fractional blood volumes compared to regions with less intense metabolism. AC demonstrated a less heterogeneous metabolism, but with similar trends. These differences in metabolism between NSCLC histologies might be relevant for targeting therapies and radiotherapy dose escalation.

• Download figure
• Open in new tab
• Download powerpoint