Prognosis Prediction of Pancreas Cancer using Textural Analysis Parameters derived from F-18 FDG PET/CT

Objectives We evaluated the prognosis prediction ability of textural analysis parameters derived from preoperative F-18 FDG PET/CT in operable pancreas cancer patients.

Methods Sixty pancreas cancer patients (m:f=36:24, age=67±9 yrs) who had undergone F-18 FDG PET/CT prior to the surgery were enrolled. Surgico-pathologic parameters, postop treatment, PET parameters (SUVmax, glucose-incorporated SUVmax=GI-SUVmax, metabolic tumor volume=MTV, total-lesion glycolysis=TLG), and a number of textural parameters were compared with survival. 3D textural analysis was performed using imaging analysis software (MaZda version 4.6).

Results All the patients had R0 resection. 37 (m:f=26:11, age 66 ± 11 yrs) patients died at 16.73±8.81 months after the operation, whereas 23 (m:f=10:13, age 69 ± 10 yrs) survived with the follow-up duration of 20.02±11.74 months. In the univariate analysis, concomitant operation (p=0.0003), tumor size (p=0.0019), GI-SUVmax with a cutoff of 600 (p=0.0030), blood loss amount (p=0.0212), portal invasion (p=0.0280), postop radiotherapy (p=0.0337) and number of metastatic lymph nodes (p=0.0481) were significant parameters for survival. Among the textural parameters, VolumeS(4,-4,0) (p=0.0479), Kurtosis3D (p=0.0518), VolumeS(5,0,0) (p=0.0575), Skewness3D (p=0.0635), VolumeS(4,0,0) (p=0.080), VolumeS(5,-5,0) (p=0.0986), and VolumeS(5,5,0) (p=0.0986) were potentially significant predictors in the univariate analysis. However, all the textural parameters having significancy in the univariate analysis did not remain in the multivariate analysis. Only GI-SUVmax with a cutoff of 600 (p=0.0001), tumor size (p=0.0060), and concomitant operation (p=0.0043) were significant predictors of survival.

Conclusions The current research result indicates that the textural analysis of F-18 FDG PET/CT has a limited performance for survival prediction of operable pancreas cancer.