Using principal component analysis for the prediction of tumor response to transarterial chemoembolization

Abstract

Purpose: To quantitate the tumor blush of hepatocellular carcinoma (HCC) at the time of transarterial chemoembolization (TACE) using principal component analysis (PCA), and to correlate the quantitated tumor blush to response to therapy. Materials and methods: In this proof-of-concept study, 27 primary HCC tumors in 25 patients (18 men, 7 women; mean age 66 years ± 9) were analyzed. We conducted a retrospective analysis of TACE procedures that were performed during March through July of 2017. Digital subtraction angiography (DSA) was combined with PCA to condense spatial and temporal information into a single image. The tumor and liver contrast enhancements were calculated, and the ratio was used to determine the relative vascular enhancement of the tumor. Tumor response to therapy was determined at 1-month post procedure. Results: Using PCA-generated fluoroscopic imaging (PCA-FI), we quantitated the tumor blush and assigned a vascular enhancement value (VEV) to each tumor. Tumors that responded to treatment (N = 12) had statistically higher VEVs compared with the nonresponders (N = 15), with a mean value of 0.96 ± 0.455 vs. 0.57 ± 0.309, (p = 0.013). Conclusions: We developed a method for quantitating tumor blush using routine angiographic images. The VEVs calculated using these images may allow for the prediction of tumor response to therapy. This pilot study suggests that there is a correlation between tumor blush intensity and tumor response.