Prostate cancer segmentation with simultaneous estimation of Markov random field parameters and class

Automated prostate cancer localization without the need for peripheral zone extraction using multiparametric MRI

PURPOSE

Multiparametric magnetic resonance imaging (MRI) has been shown to have higher localization accuracy than transrectal ultrasound (TRUS) for prostate cancer. Therefore, automated cancer segmentation using multiparametric MRI is receiving a growing interest, since MRI can provide both morphological and functional images for tissue of interest. However, all automated methods to this date are applicable to a single zone of the prostate, and the peripheral zone (PZ) of the prostate needs to be extracted manually, which is a tedious and time-consuming job. In this paper, our goal is to remove the need of PZ extraction by incorporating the spatial and geometric information of prostate tumors with multiparametric MRI derived from T2-weighted MRI, diffusion-weighted imaging (DWI) and dynamic contrast enhanced MRI (DCE-MRI).

METHODS

In order to remove the need of PZ extraction, the authors propose a new method to incorporate the spatial information of the cancer. This is done by introducing a new feature called location map. This new feature is constructed by applying a nonlinear transformation to the spatial position coordinates of each pixel, so that the location map implicitly represents the geometric position of each pixel with respect to the prostate region. Then, this new feature is combined with multiparametric MR images to perform tumor localization. The proposed algorithm is applied to multiparametric prostate MRI data obtained from 20 patients with biopsy-confirmed prostate cancer.

RESULTS

The proposed method which does not need the masks of PZ was found to have prostate cancer detection specificity of 0.84, sensitivity of 0.80 and dice coefficient value of 0.42.

CONCLUSIONS

The authors have found that fusing the spatial information allows us to obtain tumor outline without the need of PZ extraction with a considerable success (better or similar performance to methods that require manual PZ extraction). Our experimental results quantitatively demonstrate the effectiveness of the proposed method, depicting that the proposed method has a slightly better or similar localization performance compared to methods which require the masks of PZ.

Weighted maximum posterior marginals for random fields using an ensemble of conditional densities from multiple Markov chain Monte Carlo simulations

The ability of classification systems to adjust their performance (sensitivity/specificity) is essential for tasks in which certain errors are more significant than others. For example, mislabeling cancerous lesions as benign is typically more detrimental than mislabeling benign lesions as cancerous. Unfortunately, methods for modifying the performance of Markov random field (MRF) based classifiers are noticeably absent from the literature, and thus most such systems restrict their performance to a single, static operating point (a paired sensitivity/specificity). To address this deficiency we present weighted maximum posterior marginals (WMPM) estimation, an extension of maximum posterior marginals (MPM) estimation. Whereas the MPM cost function penalizes each error equally, the WMPM cost function allows misclassifications associated with certain classes to be weighted more heavily than others. This creates a preference for specific classes, and consequently a means for adjusting classifier performance. Realizing WMPM estimation (like MPM estimation) requires estimates of the posterior marginal distributions. The most prevalent means for estimating these--proposed by Marroquin--utilizes a Markov chain Monte Carlo (MCMC) method. Though Marroquin's method (M-MCMC) yields estimates that are sufficiently accurate for MPM estimation, they are inadequate for WMPM. To more accurately estimate the posterior marginals we present an equally simple, but more effective extension of the MCMC method (E-MCMC). Assuming an identical number of iterations, E-MCMC as compared to M-MCMC yields estimates with higher fidelity, thereby 1) allowing a far greater number and diversity of operating points and 2) improving overall classifier performance. To illustrate the utility of WMPM and compare the efficacies of M-MCMC and E-MCMC, we integrate them into our MRF-based classification system for detecting cancerous glands in (whole-mount or quarter) histological sections of the prostate.

Elastic registration of multimodal prostate MRI and histology via multiattribute combined mutual information

PURPOSE

By performing registration of preoperative multiprotocol in vivo magnetic resonance (MR) images of the prostate with corresponding whole-mount histology (WMH) sections from postoperative radical prostatectomy specimens, an accurate estimate of the spatial extent of prostate cancer (CaP) on in vivo MR imaging (MRI) can be retrospectively established. This could allow for definition of quantitative image-based disease signatures and lead to development of classifiers for disease detection on multiprotocol in vivo MRI. Automated registration of MR and WMH images of the prostate is complicated by dissimilar image intensities, acquisition artifacts, and nonlinear shape differences.

METHODS

The authors present a method for automated elastic registration of multiprotocol in vivo MRI and WMH sections of the prostate. The method, multiattribute combined mutual information (MACMI), leverages all available multiprotocol image data to drive image registration using a multivariate formulation of mutual information.

RESULTS

Elastic registration using the multivariate MI formulation is demonstrated for 150 corresponding sets of prostate images from 25 patient studies with T2-weighted and dynamic-contrast enhanced MRI and 85 image sets from 15 studies with an additional functional apparent diffusion coefficient MRI series. Qualitative results of MACMI evaluation via visual inspection suggest that an accurate delineation of CaP extent on MRI is obtained. Results of quantitative evaluation on 150 clinical and 20 synthetic image sets indicate improved registration accuracy using MACMI compared to conventional pairwise mutual information-based approaches.

CONCLUSIONS

The authors' approach to the registration of in vivo multiprotocol MRI and ex vivo WMH of the prostate using MACMI is unique, in that (1) information from all available image protocols is utilized to drive the registration with histology, (2) no additional, intermediate ex vivo radiology or gross histology images need be obtained in addition to the routinely acquired in vivo MRI series, and (3) no corresponding anatomical landmarks are required to be identified manually or automatically on the images.

Contextual computer-aided detection: improving bright lesion detection in retinal images and coronary calcification identification in CT scans

Contextual information plays an important role in medical image understanding. Medical experts make use of context to detect and differentiate pathologies in medical images, especially when interpreting difficult cases. The majority of computer-aided diagnosis (CAD) systems, however, employ only local information to classify candidates, without taking into account global image information or the relation of a candidate with neighboring structures. In this paper, we present a generic system for including contextual information in a CAD system. Context is described by means of high-level features based on the spatial relation between lesion candidates and surrounding anatomical landmarks and lesions of different classes (static contextual features) and lesions of the same type (dynamic contextual features). We demonstrate the added value of contextual CAD for two real-world CAD tasks: the identification of exudates and drusen in 2D retinal images and coronary calcifications in 3D computed tomography scans. Results show that in both applications contextual CAD is superior to a local CAD approach with a significant increase of the figure of merit of the Free Receiver Operating Characteristic curve from 0.84 to 0.92 and from 0.88 to 0.98 for exudates and drusen, respectively, and from 0.87 to 0.93 for coronary calcifications.

Enhanced Multi-Protocol Analysis via Intelligent Supervised Embedding (EMPrAvISE): Detecting Prostate Cancer on Multi-Parametric MRI

Currently, there is significant interest in developing methods for quantitative integration of multi-parametric (structural, functional) imaging data with the objective of building automated meta-classifiers to improve disease detection, diagnosis, and prognosis. Such techniques are required to address the differences in dimensionalities and scales of individual protocols, while deriving an integrated multi-parametric data representation which best captures all disease-pertinent information available. In this paper, we present a scheme called Enhanced Multi-Protocol Analysis via Intelligent Supervised Embedding (EMPrAvISE); a powerful, generalizable framework applicable to a variety of domains for multi-parametric data representation and fusion. Our scheme utilizes an ensemble of embeddings (via dimensionality reduction, DR); thereby exploiting the variance amongst multiple uncorrelated embeddings in a manner similar to ensemble classifier schemes (e.g. Bagging, Boosting). We apply this framework to the problem of prostate cancer (CaP) detection on 12 3 Tesla pre-operative in vivo multi-parametric (T2-weighted, Dynamic Contrast Enhanced, and Diffusion-weighted) magnetic resonance imaging (MRI) studies, in turn comprising a total of 39 2D planar MR images. We first align the different imaging protocols via automated image registration, followed by quantification of image attributes from individual protocols. Multiple embeddings are generated from the resultant high-dimensional feature space which are then combined intelligently to yield a single stable solution. Our scheme is employed in conjunction with graph embedding (for DR) and probabilistic boosting trees (PBTs) to detect CaP on multi-parametric MRI. Finally, a probabilistic pairwise Markov Random Field algorithm is used to apply spatial constraints to the result of the PBT classifier, yielding a per-voxel classification of CaP presence. Per-voxel evaluation of detection results against ground truth for CaP extent on MRI (obtained by spatially registering pre-operative MRI with available whole-mount histological specimens) reveals that EMPrAvISE yields a statistically significant improvement (AUC=0.77) over classifiers constructed from individual protocols (AUC=0.62, 0.62, 0.65, for T2w, DCE, DWI respectively) as well as one trained using multi-parametric feature concatenation (AUC=0.67).

CADOnc©: An Integrated Toolkit For Evaluating Radiation Therapy Related Changes In The Prostate Using Multiparametric MRI

The use of multi-parametric Magnetic Resonance Imaging (T2-weighted, MR Spectroscopy (MRS), Diffusion-weighted (DWI)) has recently shown great promise for diagnosing and staging prostate cancer (CaP) in vivo. Such imaging has also been utilized for evaluating the early effects of radiotherapy (RT) (e.g. intensity-modulated radiation therapy (IMRT), proton beam therapy, brachytherapy) in the prostate with the overarching goal being to successfully predict short- and long-term patient outcome. Qualitative examination of post-RT changes in the prostate using MRI is subject to high inter- and intra-observer variability. Consequently, there is a clear need for quantitative image segmentation, registration, and classification tools for assessing RT changes via multi-parametric MRI to identify (a) residual disease, and (b) new foci of cancer (local recurrence) within the prostate. In this paper, we present a computerized image segmentation, registration, and classification toolkit called CADOnc^©, and leverage it for evaluating (a) spatial extent of disease pre-RT, and (b) post-RT related changes within the prostate. We demonstrate the applicability of CADOnc^© in studying IMRTrelated changes using a cohort of 7 multi-parametric (T2w, MRS, DWI) prostate MRI patient datasets. First, the different MRI protocols from pre- and post-IMRT MRI scans are affinely registered (accounting for gland shrinkage), followed by automated segmentation of the prostate capsule using an active shape model. A number of feature extraction schemes are then applied to extract multiple textural, metabolic, and functional MRI attributes on a per-voxel basis. An AUC of 0.7132 was achieved for automated detection of CaP on pre-IMRT MRI (via integration of T2w, DWI, MRS features); evaluated on a per-voxel basis against radiologist-derived annotations. CADOnc^© also successfully identified a total of 40 out of 46 areas where disease-related changes (both absence and recurrence) occurred post-IMRT, based on changes in the expression of quantitative MR imaging biomarkers. CADOnc^© thus provides an integrated platform of quantitative analysis tools to evaluate treatment response in vivo, based on multi-parametric MRI data.

MULTI-MODAL DATA FUSION SCHEMES FOR INTEGRATED CLASSIFICATION OF IMAGING AND NON-IMAGING BIOMEDICAL DATA

With a wide array of multi-modal, multi-protocol, and multi-scale biomedical data available for disease diagnosis and prognosis, there is a need for quantitative tools to combine such varied channels of information, especially imaging and non-imaging data (e.g. spectroscopy, proteomics). The major problem in such quantitative data integration lies in reconciling the large spread in the range of dimensionalities and scales across the different modalities. The primary goal of quantitative data integration is to build combined meta-classifiers; however these efforts are thwarted by challenges in (1) homogeneous representation of the data channels, (2) fusing the attributes to construct an integrated feature vector, and (3) the choice of learning strategy for training the integrated classifier. In this paper, we seek to (a) define the characteristics that guide the 4 independent methods for quantitative data fusion that use the idea of a meta-space for building integrated multi-modal, multi-scale meta-classifiers, and (b) attempt to understand the key components which allowed each method to succeed. These methods include (1) Generalized Embedding Concatenation (GEC), (2) Consensus Embedding (CE), (3) Semi-Supervised Multi-Kernel Graph Embedding (SeSMiK), and (4) Boosted Embedding Combination (BEC). In order to evaluate the optimal scheme for fusing imaging and non-imaging data, we compared these 4 schemes for the problems of combining (a) multi-parametric MRI with spectroscopy for prostate cancer (CaP) diagnosis in vivo, and (b) histological image with proteomic signatures (obtained via mass spectrometry) for predicting prognosis in CaP patients. The kernel combination approach (SeSMiK) marginally outperformed the embedding combination schemes. Additionally, intelligent weighting of the data channels (based on their relative importance) appeared to outperform unweighted strategies. All 4 strategies easily outperformed a naïve decision fusion approach, suggesting that data integration methods will play an important role in the rapidly emerging field of integrated diagnostics and personalized healthcare.

Prostate cancer localization with multispectral MRI using cost-sensitive support vector machines and conditional random fields

Prostate cancer is a leading cause of cancer death for men in the United States. Fortunately, the survival rate for early diagnosed patients is relatively high. Therefore, in vivo imaging plays an important role for the detection and treatment of the disease. Accurate prostate cancer localization with noninvasive imaging can be used to guide biopsy, radiotherapy, and surgery as well as to monitor disease progression. Magnetic resonance imaging (MRI) performed with an endorectal coil provides higher prostate cancer localization accuracy, when compared to transrectal ultrasound (TRUS). However, in general, a single type of MRI is not sufficient for reliable tumor localization. As an alternative, multispectral MRI, i.e., the use of multiple MRI-derived datasets, has emerged as a promising noninvasive imaging technique for the localization of prostate cancer; however almost all studies are with human readers. There is a significant inter and intraobserver variability for human readers, and it is substantially difficult for humans to analyze the large dataset of multispectral MRI. To solve these problems, this study presents an automated localization method using cost-sensitive support vector machines (SVMs) and shows that this method results in improved localization accuracy than classical SVM. Additionally, we develop a new segmentation method by combining conditional random fields (CRF) with a cost-sensitive framework and show that our method further improves cost-sensitive SVM results by incorporating spatial information. We test SVM, cost-sensitive SVM, and the proposed cost-sensitive CRF on multispectral MRI datasets acquired from 21 biopsy-confirmed cancer patients. Our results show that multispectral MRI helps to increase the accuracy of prostate cancer localization when compared to single MR images; and that using advanced methods such as cost-sensitive SVM as well as the proposed cost-sensitive CRF can boost the performance significantly when compared to SVM.

Supervised and unsupervised methods for prostate cancer segmentation with multispectral MRI

PURPOSE

Magnetic resonance imaging (MRI) has been proposed as a promising alternative to transrectal ultrasound for the detection and localization of prostate cancer and fusing the information from multispectral MR images is currently an active research area. In this study, the goal is to develop automated methods that combine the pharmacokinetic parameters derived from dynamic contrast enhanced (DCE) MRI with quantitative T2 MRI and diffusion weighted imaging (DWI) in contrast to most of the studies which were performed with human readers. The main advantages of the automated methods are that the observer variability is removed and easily reproducible results can be efficiently obtained when the methods are applied to a test data. The goal is also to compare the performance of automated supervised and unsupervised methods for prostate cancer localization with multispectral MRI.

METHODS

The authors use multispectral MRI data from 20 patients with biopsy-confirmed prostate cancer patients, and the image set consists of parameters derived from T2, DWI, and DCE-MRI. The authors utilize large margin classifiers for prostate cancer segmentation and compare them to an unsupervised method the authors have previously developed. The authors also develop thresholding schemes to tune support vector machines (SVMs) and their probabilistic counterparts, relevance vector machines (RVMs), for an improved performance with respect to a selected criterion. Moreover, the authors apply a thresholding method to make the unsupervised fuzzy Markov random fields method fully automatic.

RESULTS

The authors have developed a supervised machine learning method that performs better than the previously developed unsupervised method and, additionally, have found that there is no significant difference between the SVM and RVM segmentation results. The results also show that the proposed methods for threshold selection can be used to tune the automated segmentation methods to optimize results for certain criteria such as accuracy or sensitivity. The test results of the automated algorithms indicate that using multispectral MRI improves prostate cancer segmentation performance when compared to single MR images, a result similar to the human reader studies that were performed before.

CONCLUSIONS

The automated methods presented here can help diagnose and detect prostate cancer, and improve segmentation results. For that purpose, multispectral MRI provides better information about cancer and normal regions in the prostate when compared to methods that use single MRI techniques; thus, the different MRI measurements provide complementary information in the automated methods. Moreover, the use of supervised algorithms in such automated methods remain a good alternative to the use of unsupervised algorithms.

High-throughput detection of prostate cancer in histological sections using probabilistic pairwise Markov models

In this paper we present a high-throughput system for detecting regions of carcinoma of the prostate (CaP) in HSs from radical prostatectomies (RPs) using probabilistic pairwise Markov models (PPMMs), a novel type of Markov random field (MRF). At diagnostic resolution a digitized HS can contain 80Kx70K pixels - far too many for current automated Gleason grading algorithms to process. However, grading can be separated into two distinct steps: (1) detecting cancerous regions and (2) then grading these regions. The detection step does not require diagnostic resolution and can be performed much more quickly. Thus, we introduce a CaP detection system capable of analyzing an entire digitized whole-mount HS (2x1.75cm(2)) in under three minutes (on a desktop computer) while achieving a CaP detection sensitivity and specificity of 0.87 and 0.90, respectively. We obtain this high-throughput by tailoring the system to analyze the HSs at low resolution (8microm per pixel). This motivates the following algorithm: (Step 1) glands are segmented, (Step 2) the segmented glands are classified as malignant or benign, and (Step 3) the malignant glands are consolidated into continuous regions. The classification of individual glands leverages two features: gland size and the tendency for proximate glands to share the same class. The latter feature describes a spatial dependency which we model using a Markov prior. Typically, Markov priors are expressed as the product of potential functions. Unfortunately, potential functions are mathematical abstractions, and constructing priors through their selection becomes an ad hoc procedure, resulting in simplistic models such as the Potts. Addressing this problem, we introduce PPMMs which formulate priors in terms of probability density functions, allowing the creation of more sophisticated models. To demonstrate the efficacy of our CaP detection system and assess the advantages of using a PPMM prior instead of the Potts, we alternately incorporate both priors into our algorithm and rigorously evaluate system performance, extracting statistics from over 6000 simulations run across 40 RP specimens. Perhaps the most indicative result is as follows: at a CaP sensitivity of 0.87 the accompanying false positive rates of the system when alternately employing the PPMM and Potts priors are 0.10 and 0.20, respectively.