Purpose To develop multiparametric magnetic resonance imaging (mpMRI) models for generating a quantitative, user-independent, voxel-wise composite biomarker score (CBS) for prostate cancer (PCa) detection utilizing co-registered correlative histopathology and comparing CBS-based detection performance against single quantitative MRI (qMR) parameters. Materials and Methods After providing informed consent to participate in an IRB approved protocol, patients with an initial diagnosis of PCa electing surgery as definitive therapy were imaged preoperatively with an mpMRI protocol from which quantitative MR parameters (qMR) were calculated. Per patient, all voxels in the prostate from an MRI imaging slice were classified as cancer or non-cancer based on deformably mapped histologically determined regions of disease to assess individual qMR parameters. Predictive models were developed using more than one qMR to generate CBS maps for cancer detection. Model development and evaluations of individual qMR and CBS were performed separately for the peripheral zone (PZ) alone and the whole gland (WG). Model accuracy was evaluated using the area under the receiver operating characteristic curve (AUC) and confidence intervals were calculated using the bootstrap procedure. The improvement in classification accuracy was evaluated by comparing the AUC for our multiparametric model and the single best performing qMR both at the individual level and in aggregate. Results T2TSE, ADC, Ktrans, kep and AUGC90 were significantly different between cancer and non-cancer voxels (p < 0.001 in all cases), with ADC exhibiting the best classification accuracy (AUC 0.82 for PZ and 0.74 for WG). A 4 parameter PZ-Model (AUC =0.85; p = 0.010 vs. ADC alone) and a 4 parameter WG-Model (AUC = 0.77; p = 0.043 vs. ADC alone) had the best performance of the multiparametric models considered. Based on individual-level analysis, we observed a statistically significant improvement in the AUC in 82% (23/28) and 71% (24/34) of patients when using the PZ and WG models, respectively, compared to ADC alone. Model-based CBS maps for cancer detection demonstrate improved visualization of cancer location and extent. Conclusions Co-registered correlative histopathology data was used as the ground truth for development of quantitative mpMRI models yielding voxel-wise CBS which outperform single qMR parameters for PCa detection especially when assessed at the individual level.

Diffusion MRI (dMRI) reveals microstructural features of the brain white matter by quantifying the anisotropic diffusion of water molecules within axonal bundles. Yet, identifying features such as axonal orientation dispersion, density, diameter, etc., in complex white matter fiber configurations (e.g. crossings) has proved challenging. Besides optimized data acquisition and advanced biophysical models, computational procedures to fit such models to the data are critical. However, these procedures have been largely overlooked by the dMRI microstructure community and new, more versatile, approaches are needed to solve complex biophysical model fitting problems. Existing methods are limited to models assuming single fiber orientation, relevant to limited brain areas like the corpus callosum, or multiple orientations but without the ability to extract detailed microstructural features. Here, we introduce a new and versatile optimization technique (MIX), which enables microstructure imaging of crossing white matter fibers. We provide a MATLAB implementation of MIX, and demonstrate its applicability to general microstructure models in fiber crossings using synthetic as well as ex-vivo and in-vivo brain data.

No abstract

Prostate cancer (PCa) is a major cause of cancer death among men. The histopathological examination of post-surgical prostate specimens and manual annotation of PCa not only allow for detailed assessment of disease characteristics and extent, but also supply the ground truth for developing of computer-aided diagnosis (CAD) systems for PCa detection before definitive treatment. As manual cancer annotation is tedious and subjective, there have been a number of publications describing methods for automating the procedure via the analysis of digitized whole-slide images (WSIs). However, these studies have focused only on the analysis of WSIs stained with hematoxylin and eosin (H&E), even though there is additional information that could be obtained from immunohistochemical (IHC) staining. In this work, we propose a framework for automating the annotation of PCa that is based on automated colorimetric analysis of both H&E and IHC WSIs stained with a triple-antibody cocktail against high-molecular weight cytokeratin (HMWCK), p63, and α-methylacyl CoA racemase (AMACR). The analysis outputs were then used to train a regression model to estimate the distribution of cancerous epithelium within slides. The approach yielded an AUC of 0.951, sensitivity of 87.1%, and specificity of 90.7% as compared to slide-level annotations, and generalized well to cancers of all grades.

No abstract