Quantitative Image Marker Beneficial for Identifying Response to Postsurgical Chemotherapy in Ovarian Cancer
Among gynecologic malignancies, ovarian cancer is associated with the highest mortality rate.
Integrating image features from both spatial and frequency domains may be beneficial in predicting early response to chemotherapy in patients with advanced-stage ovarian cancer, according to a study published in Physics in Medicine & Biology.
A research team established a computed tomography (CT) image dataset that included 120 ovarian cancer patients enrolled in clinical trials for new chemotherapy development. Patients were included in the study if they had a diagnosis of recurrent ovarian, peritoneal, or tubal carcinoma of high-grade histology, received treatment with systemic chemotherapy after the primary cytoreduction, and had their pre- and post-therapy CT examinations available for review. A computer-assisted tumor segmentation module was utilized to identify metastatic tumors of interest.
The researchers applied a computer-assisted scheme to analyze 133 features in both the spatial and frequency domains; these features were classified into 5 groups: shape and density and gray level difference method (GLDM) in the spatial domain, and fast Fourier transform (FFT), discrete cosine transform (DCT), and Wavelet in the frequency domain. The shape and density features described entire tumor-containing images; however, the FFT, DCT, Wavelet, and GLDM features were block-based localized features created by dividing the image into 8 x 8 or 9 x 9 pixel blocks and fed into the FFT, DCT, Wavelet or GLDM algorithms to construct a feature vector.
Areas under the ROC curve (AUC values) of the initial feature group above 0.65 were identified to generate the optimal feature cluster using a particle swarm optimization (PSO) algorithm.
In the slice-by-slice tumor strategy, the central slice was found to contain the largest tumor area, with the first, second, and third adjacent slices decreasing gradually. The tumor area disappeared in the fourth adjacent slice. Among all 133 features, 48 (36%) achieved an AUC higher than 0.65 among which the skewnessGLDM4 (0.740 ± 0.092), RmsGLDM4 (0.731 ± 0.094) and meanGLDM4 (0.720 ± 0.092) feature clusters were the top 3 performers. Further analysis indicated that DCT- and FFT-related features yield the highest and lowest average AUC at 0.634 ± 0.098 and 0.565 ± 0.094, respectively.
“This study demonstrates the potential of our proposed new quantitative image marker fused with the features computed from both spatial and frequency domain for a reliable prediction of tumor response to postsurgical chemotherapy,” the authors concluded.
Zargari A, Du Y, Heidari M, et al. Prediction of chemotherapy response in ovarian cancer patients using a new clustered quantitative image marker. Phys Med Biol. 2018;63(15):155020.