2075

Reduction of the variation in parameter estimation from atypical signal intensity decay or its variation near tumor in low b factors using a ROI-based analysis method in IVIM model for prostate diffusion imaging
In Chan Song1, Sang Youn Kim1, Jeong Yeon Cho1, and Seung Hyup Kim1

1Radiology, Seoul National University Hospital, Seoul, Korea, Republic of

Synopsis

In an IVIM technique, atypical signal intensity decay in low b factors or its variation at neighboring pixels at tumor lesion may be caused by tumor heterogeneity or spatial mismatch due to image distortion in EPI sequence, which can make estimated IVIM parameters be unreliable in conventional pixel-by-pixel method. Thus, to obtain more reliable IVIM outputs for prostate IVIM MR imaging, we suggest a new and simple ROI-based analysis method using all data of surrounding pixels in estimation and our study demonstrates a ROI-based analysis method decreased variation in IVIM parameters and can provide more reliable IVIM map images.

Introduction

Diffusion weighted images have been known as an important tool for the detection of prostate cancer (1). In these days, many literatures show that intravoxel incoherent motion (IVIM) imaging technique may be useful for the evaluation of prostate pathology in view of providing diffusion and perfusion components simultaneously (2). But, a variety of IVIM mathematical modeling and image quality problem of patients’ motion from long scan time by its usage of multiple b factors can make its clinical application be difficult in view of IVIM parameter precision. In addition, atypical signal intensity decay in low b factors or its variation at near tumor lesion may be caused by tumor heterogeneity or spatial mismatch due to image distortion from ultrafast imaging sensitivity to field inhomogeneity in EPI sequence. These factors can make estimated IVIM parameters be unreliable in conventional pixel-by-pixel method compared to simple diffusion exponential model. Thus, to obtain more reliable IVIM outputs, we suggest a new and simple ROI-based analysis method using all MR data of the surrounding pixels under the assumption that small areas of the neughboring areas at a given pixel may be homogeneous in pathology although tissue properties of tumor is known as to be heterogeneous.

Materials and Methods

In a total of 9 patients who have been histo-pathologically proven as prostate cancer, all MR data were acquired using an IVIM technique on a 3T MR unit. A total of 5 b factors of 0, 50,150, 300 and 1000 sec/mm2 were set in EPI diffusion sequence. IVIM parameters were estimated by a 2-step method based on the handling of diffusion and perfusion-related region separately (3). The b value of 300 s/mm2 was used as a threshold b factor in pseudo-diffusion, perfusion-related region. Diffusion, pseudo-diffusion and perfusion fraction parameters were obtained using the nonlinear least squares method based on IVIM model. The pixel-by-pixel and ROI-based analysis methods were performed depending on the inclusion of pixels in estimation. In pixel-by-pixel analysis, IVIM parameters were measured at only one pixel, that is to say, pixel-by-pixel conventionally. In ROI-based analysis, all signals of the center pixel and its surrounding 8 pixels were together included in estimation of IVIM parameters and its parameters were mapped at its center pixel, but inappropriate pixels on IVIM model determined in pixel-by pixel analysis in all 9 pixels were excluded in ROI-based analysis. The mean and variance in IVIM parameters of prostate cancer were evaluated at tumor in two analysis methods. Tumor lesions were identified from T2-weighted and diffusion-weighted images by two expert radiologists.

Results

In most prostate tumor lesions, smaller variances in pseudo-perfusion and perfusion fraction are found in ROI-based analysis than in pixel-by-pixel analysis methods. The mean values of diffusion were not significantly different between two methods, but other parameters related to perfusion such as pseudo-diffusion and perfusion fraction mostly decreased in ROI-based analysis method compared to pixel-by-pixel analysis method (Table 1). Many pixels showed the atypical signal decays or their variations in low b factors at mass in most patients (Figure 1). So, whereas in pixel-by-pixel analysis there were many areas with the failure regions in IVIM model, ROI-based analysis showed smaller failed areas (Figure 2). The blurring of map images from our analysis method was shown.

Conclusions and Discussion

Our findings indicates atypical signal intensity decay in low b factors or its variation at tumor lesion and its surrounding areas may cause IVIM parameters to be unreliable or to get large variance in pixel-by-pixel analysis method. However, our study demonstrates a ROI-based analysis method has decreased variation in IVIM parameters and can provide more reliable map images even if the blurring from the usage of multiple pixels in the estimation occurs

Acknowledgements

No acknowledgement found.

References

1. Döpfert JLA, et al. Magn Reson Imaging. 2011;29(8):1053–1058

2. Le Bihan D. Radiology.2008;249(3):748–752

3. Luciani A1, et al. Radiology. 2008 Dec;249(3):891-9.

Figures

Signal intensity decay curves along b factors in 9 pixels surrounding a representative pixel of mass with a bright region in DWI. Atypical signal intensity decays not suitable for IVIM model in low b factors is detected and pixels show variation in signal intensity decay pattern.

Diffusion, pseudo-perfusion and perfusion fraction map images from pixel-by-pixel and ROI-based analysis methods in patient with nodule in left peripheral zone on DWI and T2WI. Zero pixel (black) means it is not appropriate in IVIM model from the failure of the fitting method.

The mean and standard deviations of IVIM parameters by pixel-by-pixel and ROI-based analysis methods in tumor lesions from 9 patients

Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)
2075