Increasing the Accuracy of Volume and ADC Delineation for Heterogeneous Tumor on Diffusion-Weighted MRI: Correlation with PET/CT

Biomarkers from in vivo molecular imaging of breast cancer: pretreatment 18F-FDG PET predicts patient prognosis, and pretreatment DWI-MR predicts response to neoadjuvant chemotherapy

“…In defining quantitative ADC-derived parameters, different studies do not deal with the separation of contribution of noise, necrotic or cystic tissue within the lesions, thus giving ADC values which may not reflect only proliferative cancer tissue. However, it is acknowledged by physicists, radiologists and oncologists that the different tissues could contribute differently to DWI b0 and AD maps signal intensities [37, 38], thus potentially giving the ability to distinguish the different contribution by image classification [28]. …”

“…Regarding MR, most works have assessed the role of ADC as surrogate biomarkers for characterization and treatment response prediction [39]. To our knowledge, only one study has proposed and validated a methodology to segment ADC maps and to obtain volume-like ADC parameters as candidate biomarkers for diagnosis and prognosis [28]. That study adopted a conservative approach based on the exclusion of non-carcinoma compartments from the segmented lesion volume rather than the direct classification of high-cellularity tumor tissues.…”

“…A three-dimensional (3D) functional volume was defined on the ADC maps ( V ADC ) by an adaptation of the segmentation method developed by Nan-Jie Gong et al [28], but taking advantage from the availability of the DCE acquisition to eliminate the need for manual lesion annotations. Nan-Jie Gong et al [28] proposed the measure of the lesion functional volume by overlapping the parts of lesion tissue obtained from DWI images with b value = 0 s/mm2 (DWI b0 image) and ADC maps, after classifying and excluding probable non-carcinoma compartments from both the images.…”

“…Nan-Jie Gong et al [28] proposed the measure of the lesion functional volume by overlapping the parts of lesion tissue obtained from DWI images with b value = 0 s/mm2 (DWI b0 image) and ADC maps, after classifying and excluding probable non-carcinoma compartments from both the images. In their procedure, taking ADC maps as references and on DWI b0 images, ROIs were drawn manually along the lesion contour by radiologists and used as input for segmentation.…”

“…1c). This step allows locating and limiting, in a semi-automatic way, the spatial extension for lesion signal segmentation on the DWI b0 images and ADC maps, thus improving the efficiency of volume estimation and avoiding inter-observer variability with respect to the manual delineation on DWI b0 images proposed by Nan-Jie Gong et al [28]. The extracted binary mask is applied to both DWI b0 images and ADC maps, thus defining the enhanced lesion on these images (see Fig.…”