Guido Masetti

Model-based compressive sensing for damage localization in lamb wave inspection

Compressive sensing (CS) has emerged as a potentially viable technique for the efficient compression and analysis of high-resolution signals that have a sparse representation in a fixed basis. In this work, we have developed a CS approach for ultrasonic signal decomposition suitable to achieve high performance in Lamb-wave-based defect detection procedures. In the proposed approach, a CS algorithm based on an alternating minimization (AM) procedure is adopted to extract the information about both the system impulse response and the reflectivity function. The implemented tool exploits the dispersion compensation properties of the warped frequency transform as a means to generate the sparsifying basis for the signal representation. The effectiveness of the decomposition task is demonstrated on synthetic signals and successfully tested on experimental Lamb waves propagating in an aluminum plate. Compared with available strategies, the proposed approach provides an improvement in the accuracy of wave propagation path length estimation, a fundamental step in defect localization procedures.

Guided wave expansion in warped curvelet frames

Lamb wave testing for structural health monitoring (SHM) often relies on analysis of wavefields recorded through scanning laser Doppler vibrometers (SLDVs) or ultrasonic scanners. Damage detection and characterization with these techniques requires isolation of defect-induced reflections in the wavefield from the injected wave packet and from scattering events associated with structural features such as boundaries, rivets, joints, etc. This is a challenging task when dealing with complex structures and multimodal, dispersive propagation regimes, whereby various wave contributions in both the time/space and the frequency/wavenumber domain overlap. A new mathematical tool named warped curvelet frames (WCFs) is proposed to effectively decompose the recorded wavefields. The presented technique results from the combination of two operators, i.e., the curvelet transform (CT) and the warped frequency transform (WFT). The CT provides an optimally sparse representation of nondispersive wave propagators. Combining the CT with the WFT allows for a flexible analysis of multimodal wave propagation in dispersive media. Exploiting the spatial and temporal localization of curvelets, as well as the spectro-temporal adaptation of the analysis frame to the characteristics of each propagating mode, provided by frequency warping, a convenient decomposition of guided waves is achieved and relevant contributions can be effectively isolated. The proposed approach is validated through dedicated simulations and further tested experimentally to demonstrate the effectiveness of the method in separating guided wave modes corresponding to acoustic events in close spatial proximity.

Best basis compressive sensing of guided waves in structural health monitoring

A novel signal compression and reconstruction procedure suitable for guided wave based structural health monitoring (SHM) applications is presented. The proposed approach combines the wavelet packet transform and frequency warping to generate a sparse decomposition of the acquired dispersive signal. The sparsity of the signal in the considered representation is exploited to develop data compression strategy based on the Best-Basis Compressive sensing (CS) theory. The proposed data compression strategy has been compared with the transform encoder based on the Embedded Zerotree (EZT), a well known data compression algorithm. These approaches are tested on experimental Lamb wave signals obtained by acquiring acoustic emissions in a 1 m 2 aluminum plate with conventional piezoelectric sensors. The performances of the two methods are analyzed by varying the compression ratio in the range 40-80%, and measuring the discrepancy between the original and the reconstructed signal. Results show the improvement in signal reconstruction with the use of the modified CS framework with respect to transform-encoders such as the EZT algorithm with Huffman coding. Compressive Sensing based on wavelet analysis and frequency warping operator.Frequency warping Wavelet analysis.Compression of Ultrasonic Lamb Waves.Acoustic emission localization in plates with dispersion and reverberation.Procedure is tested with a passive network of three piezo-sensors.

Principal Component Analysis for the Classification of Cardiac Motion Abnormalities Based on Echocardiographic Strain and Strain Rate Imaging

Clinical value of the quantitative assessment of regional myocardial function through segmental strain and strain rate has already been demonstrated. Traditional methods for diagnosing heart diseases are based on values extracted at specific time points during the cardiac cycle, known as ‘techno-markers’, and as a consequence they may fail to provide an appropriate description of the strain (rate) characteristics. This study concerns the statistical analysis of the whole cardiac cycle by the Principal Component Analysis (PCA) method and modeling the major patterns of the strain (rate) curves. Experimental outcomes show that the PCA features can outperform their traditional counterparts in categorizing healthy and infarcted myocardial segments and are able to drive considerable benefit to a classification system by properly modeling the complex structure of the strain rate traces.

Machine learning of the spatio-temporal characteristics of echocardiographic deformation curves for infarct classification

The aim of this study was to analyze the whole temporal profiles of the segmental deformation curves of the left ventricle (LV) and describe their interrelations to obtain more detailed information concerning global LV function in order to be able to identify abnormal changes in LV mechanics. The temporal characteristics of the segmental LV deformation curves were compactly described using an efficient decomposition into major patterns of variation through a statistical method, called Principal Component Analysis (PCA). In order to describe the spatial relations between the segmental traces, the PCA-derived temporal features of all LV segments were concatenated. The obtained set of features was then used to build an automatic classification system. The proposed methodology was applied to a group of 60 MRI-delayed enhancement confirmed infarct patients and 60 controls in order to detect myocardial infarction. An average classification accuracy of 87% with corresponding sensitivity and specificity rates of 89% and 85%, respectively was obtained by the proposed methodology applied on the strain rate curves. This classification performance was better than that obtained with the same methodology applied on the strain curves, reading of two expert cardiologists as well as comparative classification systems using only the spatial distribution of the end-systolic strain and peak-systolic strain rate values. This study shows the potential of machine learning in the field of cardiac deformation imaging where an efficient representation of the spatio-temporal characteristics of the segmental deformation curves allowed automatic classification of infarcted from control hearts with high accuracy.

Automatic detection of ischemic myocardium by spatio-temporal analysis of echocardiographic strain and strain rate curves

Interpretation of ultrasonic deformation traces for making a diagnosis on local myocardial function has been known to be a challenging task in daily clinical practice. A traditional approach is to use values extracted at specific time points during the cardiac cycle which has the main drawback of not taking the temporal information of the deformation traces into account. This paper presents a framework for the automatic detection of ischemic myocardium by statistical analysis of the entire segmental strain and strain rate curves using principal component analysis (PCA). Having the PCA-derived parameters of the regional temporal profiles at hand, a spatio-temporal representation of the global left ventricle (LV) function is established to train a classification system. Experimental outcomes show that the proposed deformation representation of the LV can outperform its traditional counterpart in categorizing healthy from ischemic myocardium.

Automatic Detection of Myocardial Infarction Through a Global Shape Feature Based on Local Statistical Modeling

This paper presents a local-to-global statistical approach for modeling the major components of left ventricular LV shape using its 3-D landmark representation. The rationale for dividing the LV into local areas is bi-fold: 1 to better identify abnormalities that lead to local shape remodeling and, 2 to decrease the number of shape variables by using a limited set of landmark points for an efficient statistical parametrization. Principal Component Analysis PCA is used for the statistical modeling of the local regions and subsets of the learned parameters that provide significant discriminatory information are taken from each local model in a feature selection stage. The selected local parameters are then concatenated to form a global representation of the LV and to train a classifier for differentiating between normal and infarcted LV shapes.

Improving prostate biopsy protocol with a computer aided detection tool based on semi-supervised learning

Prostate cancer is one of the most frequently diagnosed neoplasy and its presence can only be confirmed by biopsy. Due to the high number of false positives, Computer Aided Detection (CAD) systems can be used to reduce the number of cores requested for an accurate diagnosis. This work proposes a CAD procedure for cancer detection in Ultrasound images based on a learning scheme which exploits a novel semi-supervised learning (SSL) algorithm for reducing data collection effort and avoiding collected data wasting. The ground truth database comprises the RFsignals acquired during biopsies and the corresponding tissue samples histopathological outcome. A comparison to a state-of-art CAD scheme based on supervised learning demonstrates the effectiveness of the proposed SSL procedure at enhancing CAD performance. Experiments on ground truth images from biopsy findings show that the proposed CAD scheme is effective at improving the efficiency of the biopsy protocol.

Damage imaging through compressed wavefield recovery in Lamb wave inspections

This research presents a damage detection technique based on a compressive sensing (CS) algorithm applied to full wavefield data. The aim is to realize an effective tool for damage detection and localization which allows to reduce the acquisition time. The proposed technique exploits the compressive sensing framework to infer the damage location and entity from the comparison between the wavefield reconstructions produced by the different representation domains such as those spanned by Wave atoms (WA), Curvelets (CT) and Fourier (FT) exponentials. The procedure was applied to an aluminum plate with a notch cut. The propagation of Lamb waves on such structure was simulated numerically. The results show that the technique can be applied in a variety of structural components to reduce acquisition time and achieve high performance in defect detection and localization by removing up to 80% of the Nyquist sampling grid.

A fully automated method for carotid plaques segmentation in ultrasound images based on motion estimation and level-set

Ultrasound (US) guided carotid atherosclerosis diagnosis is based on the evaluation of the stenosis degree due to the presence of carotid plaques (CPs) and on CPs composition study. Accurate and automated CPs segmentation is essential to enable these evaluations. In this work, we present a fully automated CPs segmentation method based on level-set with an innovative initialization procedure that makes the approach completely user-independent by exploiting the carotid walls motion analysis. Performance were tested on 10 US image sequences of carotid artery and compared with manual contouring from an expert physician. Results show the effectiveness of our method at performing accurate CPs segmentation in US images without requiring any user intervention.