Quantitative evaluation of myelosuppression after chemotherapy for patients with ovarian cancer using material decomposition in dual-energy computed tomography

Objective: To demonstrate feasibility of quantifying myelosuppression after chemotherapy for patients with ovarian cancer (OC) using fat density measurement on material decomposition (MD) images in dual-energy computed tomography (DECT). Materials and methods: Fifty-seven patients with OC after chemotherapy underwent DECT. MD using fat and hydroxyapatite (HAP) as basis material pair was performed. Regions of interest (ROIs) of 20 mm2 were placed on bone marrow of ilia and femoral shafts bilaterally at level of coronal hip joint and the third sagittal lumbar vertebra to measure fat density on FAT (HAP) MD images. Eight characteristics (age, pathological type, International Federation of Gynaecology and Obstetrics (FIGO) stage, unilateral/bilateral, chemotherapy protocol and cycle, days after therapy (DAT), and ROI location) were recorded. Fat densities in ilia and femoral shafts were compared bilaterally with paired sample t tests and among 3 ROI locations with Kruskal-Wallis tests. Regression and correlations were made between fat density and 8 characteristics. Regression equations, correlation coefficients, scatterplot and normal probability plot (P-P) are provided. Results: There were no statistically significant differences in fat density in ilia or femoral shafts bilaterally (p > 0.05). Average fat densities were significantly different with 916.93 +/- 9.3 in ilia, 927.04 +/- 11.86 in femoral shaft, and 932.18 +/- 8.45 in lumbar vertebra (p < 0.001). Regression equation was Y = 923.26 - 0.29 x age + 0.05 x pathological-type - 0.94 x FIGO stage + 0.82 x unilateral/bilateral - 0.54 x chemotherapy-protocol + 0.52 x chemotherapycycle + 0.01 x DAT + 7.63 x ROI location, with correlation coefficients with age, DAT and ROI location at -0.23, 0.16 and 0.53, respectively (p < 0.05). Conclusions: Myelosuppression after chemotherapy manifests as fat substitute and can be quantified by measuring fat density on FAT (HAP) images. Fat density was significantly correlated with patient age, DAT and measurement locations.