Purpose / Introduction Prostate cancer (PCa) is the highest incidence malignancy among men in Europe[1]; however, there aren’t diagnostic tools to provide an accurate assessment of PCa aggressiveness non-invasively. We aimed to investigate the diagnostic ability of Diffusion Kurtosis Imaging (DKI) and Diffusion Tensor Imaging (DTI) in staging PCa with histogram analysis. Since prostate has a heterogeneous glandular structure, a non-Gaussian diffusion model could provide a better description of different tissue compartments[2]. Subjects and Methods In this retrospective study, 31 patients, with 5 different tumor grades, underwent a diffusion-weighted (DW) magnetic resonance imaging exam after two months from the first biopsy, by using a 3T clinical MR scanner (Philips Achieva) and a six-channel phased array SENSE torso coil. DW-images were acquired along 6 directions and with 5 different b-values up to 2500s/mm2, by using a single shot EPI sequence (TR/TE=3000/67ms,voxel size=1.56x1.56x3.3mm^3,NSA=4). The image pre-processing and the reconstruction of the Mean Diffusivity (MD) maps were performed in FSL 5.0. Maps of apparent Kurtosis (K) were obtained by using an in-house algorithm developed in Matlab R2012b. The mean, 25th and 90th percentile were evaluated from the entire-lesion histogram and correlated with the new grade group classification (GG=1,2 and GG=3,4,5 are considered low-risk and high risk PCa, respectively) proposed by ISUP[3] by using the Pearson’s test. Paired t-Test was performed to test statistical significance of differences in histogram-derived parameters between benign and PCa tissue; the ability of each parameter in discriminating low- and high-grade PCa was determined by ANOVA-test. Results Statistically significant differences were found between PCa and benign tissue for each parameter, kurtosis-derived 90th percentile presents the highest significance(p<10-5). A positive correlation was found between K-derived parameters and GG, while a negative correlation was found between MD-derived parameters and GG (Table-1). Among histogram derived parameters, 90th-percentile MD and mean K discriminate between low- and high-risk PCa with the highest significance (p<0.02). Discussion / Conclusion DKI and DTI exhibit similar perfermances in discriminating PCa from benign tissue and in stratifying different tumour grades, nevertheless mean-K showed the strongest correlation with GG. These results demonstrated that Gaussian and non-Gaussian diffusion are sensitive to tissue changes, occurring with tumour progression. Moreover, histogram analysis could represent a useful tool in PCa diagnosis to investigate micro-structural arrangements in PCa more deeply. References [1] Ferlay, J. et al., Int. J. Cancer, (2015): E359–E386. [2] Jensen JH, Helpern JA. NMR Biomed, (2010);23:698-710. [3] Nezzo, M. et al., European Journal of Radiology, (2016); 85:1794-1801.
Histogram analysis of low- and high-risk prostate cancer . a comparison between Gaussian and non-Gaussian diffusion / DI TRANI, M; Caporale, Alessandra; Nezzo, M; Miano, R; Mauriello, A; Bove, P; Manenti, G; Capuani, Silvia. - 30:(2017). (Intervento presentato al convegno ESMRMB CONGRESS 2017 tenutosi a Barelona).
Histogram analysis of low- and high-risk prostate cancer . a comparison between Gaussian and non-Gaussian diffusion
DI TRANI M
;CAPORALE, ALESSANDRA;CAPUANI, Silvia
2017
Abstract
Purpose / Introduction Prostate cancer (PCa) is the highest incidence malignancy among men in Europe[1]; however, there aren’t diagnostic tools to provide an accurate assessment of PCa aggressiveness non-invasively. We aimed to investigate the diagnostic ability of Diffusion Kurtosis Imaging (DKI) and Diffusion Tensor Imaging (DTI) in staging PCa with histogram analysis. Since prostate has a heterogeneous glandular structure, a non-Gaussian diffusion model could provide a better description of different tissue compartments[2]. Subjects and Methods In this retrospective study, 31 patients, with 5 different tumor grades, underwent a diffusion-weighted (DW) magnetic resonance imaging exam after two months from the first biopsy, by using a 3T clinical MR scanner (Philips Achieva) and a six-channel phased array SENSE torso coil. DW-images were acquired along 6 directions and with 5 different b-values up to 2500s/mm2, by using a single shot EPI sequence (TR/TE=3000/67ms,voxel size=1.56x1.56x3.3mm^3,NSA=4). The image pre-processing and the reconstruction of the Mean Diffusivity (MD) maps were performed in FSL 5.0. Maps of apparent Kurtosis (K) were obtained by using an in-house algorithm developed in Matlab R2012b. The mean, 25th and 90th percentile were evaluated from the entire-lesion histogram and correlated with the new grade group classification (GG=1,2 and GG=3,4,5 are considered low-risk and high risk PCa, respectively) proposed by ISUP[3] by using the Pearson’s test. Paired t-Test was performed to test statistical significance of differences in histogram-derived parameters between benign and PCa tissue; the ability of each parameter in discriminating low- and high-grade PCa was determined by ANOVA-test. Results Statistically significant differences were found between PCa and benign tissue for each parameter, kurtosis-derived 90th percentile presents the highest significance(p<10-5). A positive correlation was found between K-derived parameters and GG, while a negative correlation was found between MD-derived parameters and GG (Table-1). Among histogram derived parameters, 90th-percentile MD and mean K discriminate between low- and high-risk PCa with the highest significance (p<0.02). Discussion / Conclusion DKI and DTI exhibit similar perfermances in discriminating PCa from benign tissue and in stratifying different tumour grades, nevertheless mean-K showed the strongest correlation with GG. These results demonstrated that Gaussian and non-Gaussian diffusion are sensitive to tissue changes, occurring with tumour progression. Moreover, histogram analysis could represent a useful tool in PCa diagnosis to investigate micro-structural arrangements in PCa more deeply. References [1] Ferlay, J. et al., Int. J. Cancer, (2015): E359–E386. [2] Jensen JH, Helpern JA. NMR Biomed, (2010);23:698-710. [3] Nezzo, M. et al., European Journal of Radiology, (2016); 85:1794-1801.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.