Simona Maggio

Predictive Deconvolution and Hybrid Feature Selection for Computer-Aided Detection of Prostate Cancer

Computer-aided detection (CAD) schemes are decision making support tools, useful to overcome limitations of problematic clinical procedures. Trans-rectal ultrasound image based CAD would be extremely important to support prostate cancer diagnosis. An effective approach to realize a CAD scheme for this purpose is described in this work, employing a multi-feature kernel classification model based on generalized discriminant analysis. The mutual information of feature value and tissue pathological state is used to select features essential for tissue characterization. System-dependent effects are reduced through predictive deconvolution of the acquired radio-frequency signals. A clinical study, performed on ground truth images from biopsy findings, provides a comparison of the classification model applied before and after deconvolution, showing in the latter case a significant gain in accuracy and area under the receiver operating characteristic curve.

A restoration framework for ultrasonic tissue characterization

Ultrasonic tissue characterization has become an area of intensive research. This procedure generally relies on the analysis of the unprocessed echo signal. Because the ultrasound echo is degraded by the non-ideal system point spread function, a deconvolution step could be employed to provide an estimate of the tissue response that could then be exploited for a more accurate characterization. In medical ultrasound, deconvolution is commonly used to increase diagnostic reliability of ultrasound images by improving their contrast and resolution. Most successful algorithms address deconvolution in a maximum a posteriori estimation framework; this typically leads to the solution of ℓ2-norm or ℓ1-norm constrained optimization problems, depending on the choice of the prior distribution. Although these techniques are sufficient to obtain relevant image visual quality improvements, the obtained reflectivity estimates are, however, not appropriate for classification purposes. In this context, we introduce in this paper a maximum a posteriori deconvolution framework expressly derived to improve tissue characterization. The algorithm overcomes limitations associated with standard techniques by using a nonstandard prior model for the tissue response. We present an evaluation of the algorithm performance using both computer simulations and tissue-mimicking phantoms. These studies reveal increased accuracy in the characterization of media with different properties. A comparison with state-of-the-art Wiener and ℓ1-norm deconvolution techniques attests to the superiority of the proposed algorithm.

A Sampling Theory Approach for Continuous ARMA Identification

The problem of estimating continuous-domain autoregressive moving-average processes from sampled data is considered. The proposed approach incorporates the sampling process into the problem formulation while introducing exponential models for both the continuous and the sampled processes. We derive an exact evaluation of the discrete-domain power-spectrum using exponential B-splines and further suggest an estimation approach that is based on digitally filtering the available data. The proposed functional, which is related to Whittles likelihood function, exhibits several local minima that originate from aliasing. The global minimum, however, corresponds to a maximum-likelihood estimator, regardless of the sampling step. Experimental results indicate that the proposed approach closely follows the Cramér-Rao bound for various aliasing configurations.

An expectation maximization framework for an improved ultrasound-based tissue characterization

Ultrasonic tissue characterization has been gaining increasing attention. This procedure is generally based on the analysis of the echo signal. As the ultrasound echo is degraded by the system Point Spread Function, deconvolution could be employed to provide a tissue response estimate, exploitable for a better characterization. In this context, we present a deconvolution framework expressively designed to improve tissue characterization. Thanks to a new model for tissue reflectivity the proposed framework overcomes limitations associated with standard ones. The performance was evaluated from several tissue-mimicking phantoms. Obtained results show relevant improvements in classification accuracy. From a comparison with standard schemes the superiority of the proposed algorithm was attested.

rtCAB: A tool for reducing unnecessary prostate biopsy cores

With more than 110.000 new cases/year in Europe, prostate cancer (PCa) is one of the most frequent neoplasy. When suspects arise from standard diagnostic methods (i.e. Digital Rectal Exam, Transrectal Ultrasonography (TRUS), PSA level) a prostate biopsy (PBx) is mandatory. As patient discomfort and adverse event probability both grows with core number, it is desirable to reduce the number of PBx cores without negative impinging on diagnose accuracy. The work describes an innovative processing technique called real-time Computer Aided Biopsy (rtCAB) which enhances TRUS video stream with a false color overlay image, and suggests the physician where to sample thus reducing the total number of cores. Our proposal consists in a real-time non-linear classifier which processes the output of an original Maximum Likelihood estimator of Nakagami parameters based on Pade´ Approximant. The resulting algorithm, implemented making full use of CUDA parallel processing capabilities, is capable to deliver frame rates as high as 30 fps. Classification model was trained on a prostate gland adenocarcinoma database (400 PBx cores, 8000 ROIs). Ground truth for each core was established by an expert physician, providing tissue description and illness percentage for each core. The system was tuned for reducing the number of false positives while preserving an acceptable number of false negatives. Comparing to a classical double sextant PBx, the positive prediction value (PPV) of our method is 65% better, with an overall sensitivity of 100%.

Continuous-domain ARMA modeling for ultrasound tissue characterization

In this study, we investigate the possibility of applying a continuous-time ARMA (CARMA) model to radio-frequency ultrasound signals. We consider the effect of the discretization process on the parameters of the continuous system, and we take into account the exponential nature of the autocorrelation function of the model to derive continuous-domain information from the parameters of the discrete ARMA process. We validate the effectiveness of the CARMA model parameters for the characterization of ultrasound tissues on a sequence of phantom images that represent various concentrations of scatterers. We also compare the proposed CARMA coefficients and the traditional ARMA parameters on the basis of their performance in discriminating between phantom tissues. We show that working in the continuous domain brings additional useful information to characterize the imaged materials.

Automatic Regions of Interest Segmentation for Computer Aided Classification of Prostate Trus Images

Transrectal ultrasound (TRUS) plays two central roles in prostate cancer diagnosis, prostate examination and measurement and biopsy guidance, but its sensitivity and specificity need improvement. This paper presents one possible method to improve TRUS detection and biopsy guidance using computer-aided diagnosis techniques for ultrasound images. The method uses automated segmentation of regions of interest followed by a supervised classifier. It was tested on a database of 37 prostate TRUS RF scans (22 with cancer). Average sensitivity was 78%, average specificity was 92% and average accuracy was 90% in discriminating normal from cancerous tissue.

Computer Aided Detection of prostate cancer based on GDA and predictive deconvolution

A Computer-Aided Detection (CAD) scheme to support prostate cancer diagnosis based on ultrasound images is presented. The approach described in this work employs a multifeature classification model. To indentify features highly correlated to the pathologic state of the tissue we use a Feature Selection algorithm based on mutual information. System-dependent effects are removed through predictive deconvolution and this operation results in increasing quality of images and discriminating power of features. A comparison of the classification model applied before and after deconvolution shows a gain in accuracy and area under the ROC curve. The use of deconvolution as preprocessing step in CAD schemes can improve prostate cancer detection.