Comparison of quantitative PET/CT and textural features in differentiating benign from malignant lesions and predicting response to chemotherapy in interim PET/CT in Soft Tissue Sarcomas.

Introduction: Soft tissue sarcomas(STS) are a heterogeneous group of tumors, which can range in their aggressiveness from benign to malignant. The treatment options include surgery, chemotherapy and radiotherapy. 18F-FDGPET/CT is commonly used to assess treatment response, through qualitative and quantitative assessment of tumor metabolism. Texture analysis by Gray Level Cooccurrence Matrix GLCM) techniques are being increasingly used in many tumors, to evaluate those features in a lesion, which may not otherwise be discernible in visual inspection. These features in STS lesions, may show a relationship with lesion biology and thus treatment response.

Aim: To assess the utility of quantitative PET/CT parameters and textural features in PET, CT and PET/CT, in patients with histopathologically proven STS; in differentiating benign from malignant lesions and in predicting response to chemotherapy in interim PET/CT. Method: Forty seven patients(mean age 32.8years;SD15.8years) with histopathologically proven STS were prospectively included in the study. Baseline 18F-FDGPET/CT was done in all patients; 40-60 min post IV-injection of 259-370MBq of radiotracer. Seven patients were lost to follow up. Of the 40 patients (F:M::16:24); 6 had benign lesions. Of the 34 patients with malignant lesions; 21 were given chemotherapy and 13 were operated. Interim PET/CT was done in 21 patients receiving chemotherapy. The baseline Images of these patients were analyzed and quantitative parameters - lesion size, SUVmax, SUVmax-Tumor to mediastinum ratio(T/M); Metabolic Tumor Volume were documented. Post segmentation of lesions using semiautomatic adaptive thresholding; Haralick statistical texture analysis (a type of GLCM) was done (using horisontal direction - 0degrees; range of 1-nearest neighbour) in CT, PET and PET/CT images. Response assessment in interim PET/CT was done using PERCIST criteria and patients were grouped in those with Progressive disease (PD) and those with non progressive disease (NPD= Stable disease and Partial response).

Results: Parameters which showed statistically significant difference, between benign and malignant lesions were - Median SUVmax(p=0.008; Mann Whitney Test); SUVmaxT/M ratio(p=0.027); CT texture features - “correlation”(p=0.019), “sum entropy”(p=0.015), “information measures of correlation1”(p=0.007), “information measures of correlation2”(p=0.005); PET texture features - “contrast”(p=0.054), “difference variance”(p=0.054); PET/CT texture features - “contrast”(p=0.013), “entropy”(p=0.037), “difference variance”(p=0.013), “difference entropy”(p=0.002). Among these maximum area under the curve was found for PET/CT parameter -“difference entropy”(AUC=0.873; p=0.004). Of 21 patients, who received chemotherapy; 7 showed PD in interim PET/CT. Baseline parameters which showed significant difference between the PD and Non-PD groups were - PET texture features - “contrast”(p=0.031), “variance”(p=0.046), “difference variance” (p=0.031); PET/CT texture features - “contrast”(p=0.002), “variance”(p=0.012), “inverse difference moment”(p=0.002), “sum entropy”(p=0.020), “entropy”(p=0.010), “difference variance”(p=0.002), “difference entropy”(p=0.001). Maximum AUC was obtained for PET/CT texture feature “difference entropy”(AUC=0.918; p=0.002; PD if <0.453 giving 100% specificity and 71.4% sensitivity). Conclusion: Semiquantitative parameters in PET/CT and textural features in CT, PET and PET/CT can differentiate benign from malignant lesions in patients with STS. Additionally, textural features in baseline PET and PET/CT can predict response to chemotherapy in interim PET/CT; while, baseline PET/CT semiquantitative parameters and CT textural features cannot. In both these scenarios, the single best textural parameter was found to be “Difference Entropy”.