Deep Learning Approaches for Robust Time of Arrival Estimation in Acoustic Emission Monitoring.

In this work, different types of artificial neural networks are investigated for the estimation of the time of arrival (ToA) in acoustic emission (AE) signals. In particular, convolutional neural network (CNN) models and a novel capsule neural network are proposed in place of standard statistical strategies which cannot handle, with enough robustness, very noisy scenarios and, thus, cannot be sufficiently reliable when the signal statistics are perturbed by local drifts or outliers. This concept was validated with two experiments: the pure ToA identification capability was firstly assessed on synthetic signals for which a ground truth is available, showing a 10× gain in accuracy when compared to the classical Akaike information criterion (AIC). Then, the same models were tested via experimental data acquired in the framework of a localization problem to identify targets with known coordinates on a square aluminum plate, demonstrating an overreaching precision under significant noise levels.

Full Wavefield Analysis and Damage Imaging Through Compressive Sensing in Lamb Wave Inspections.

One of the main challenges faced by the structural health monitoring community is acquiring and processing huge sets of acoustic wavefield data collected from sensors, such as scanning laser Doppler vibrometers or ultrasonic scanners. In fact, extracting information that allows the estimation of the damage condition of a structure can be a time-consuming process. This paper presents a damage detection and localization technique based on a compressive sensing algorithm, which significantly allows us to reduce the acquisition time without losing in detection accuracy. The proposed technique exploits the sparsity of the wavefield in different representation domains, such as those spanned by wave atoms, curvelets, and Fourier exponentials to recover the full wavefield and, at the same time, to infer the damage location, based on comparison between the wavefield reconstructions produced by the different representation domains. The procedure is applied to three different setups related to an aluminum plate with a notch, a glass fiber reinforced polymer plate with a notch, and a composite plate with a delamination. 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.

Complete band gaps in a polyvinyl chloride (PVC) phononic plate with cross-like holes: numerical design and experimental verification.

In this work the existence of band gaps in a phononic polyvinyl chloride (PVC) plate with a square lattice of cross-like holes is numerically and experimentally investigated. First, a parametric analysis is carried out to find plate thickness and cross-like holes dimensions capable to nucleate complete band gaps. In this analysis the band structures of the unitary cell in the first Brillouin zone are computed by exploiting the Bloch-Floquet theorem. Next, time transient finite element analyses are performed to highlight the shielding effect of a finite dimension phononic region, formed by unitary cells arranged into four concentric square rings, on the propagation of guided waves. Finally, ultrasonic experimental tests in pitch-catch configuration across the phononic region, machined on a PVC plate, are executed and analyzed. Very good agreement between numerical and experimental results are found confirming the existence of the predicted band gaps.

A retrospective study to reduce prostate biopsy cores by a real time interactive tool.

OBJECTIVE: Prostate carcinoma (PCa) is one of the most frequent neoplasms, with more than 110.000 new cases/year in Europe. As PCa is not clearly demonstrable at transrectal ultrasound (TRUS), guidelines on TRUS guided biopsy suggest to perform a random tissue sampling (at least 8-12 "cores" depending on gland volume). Although accuracy grows with core number, patient discomfort and adverse event probability grow as well. Thus it would be worth to aim to reduce the number of prostate biopsy cores without loss of diagnostic accuracy. MATERIALS AND METHODS: A retrospective study was performed to evaluate the feasibility of an improved version of a rtCAB tool developed at DEIS (University of Bologna) for the reduction of prostate biopsy cores. rtCAB is an innovative processing technique which enhances TRUS video stream by a live false color overlay image that helps the physician to perform the biopsy by guiding the sampling into target zones. In order to train rtCAB, a monocentric, single operator prostate gland adenocarcinoma database has been built. The database enlists 81 patients, for a total of 743 prostate byoptic (PBx) cores and 14860 ROI. For each patient we collected age, PSA levels, digital rectal examination (DRE) findings, presence or absence of focal lesions, and prostate volume. During TRUS, raw ultrasound data were acquired and associated to each PBx core. For each core we collected both the radio frequency (RF) signal and the histological outcome. RESULTS: The whole system was optimized for reducing the number of false positives while preserving an acceptable number of false negatives. Comparing to a classical PBx approach (8-12 cores), the estimated positive predictive value (PPV) of our method increased from 25% to 40%, with an overall sensitivity of 85%. CONCLUSIONS: Preliminary results show that the proposed tool can provide real-time feedback to the operator during TRUS. Sensitivity and PPV values suggest that a reduction of almost 50% the number of biopsy cores without losing in diagnostic accuracy is feasible. A prospective study is needed to further confirm these preliminary retrospective results.

A reduced complexity estimation algorithm for ultrasound images de-blurring.

In this paper we propose a deconvolution technique for ultrasound images based on a Maximum Likelihood (ML) estimation procedure. In our approach the ultrasonic radio-frequency (RF) signal is considered as a sequence affected by Intersymbol Interference (ISI) and AWG noise. In order to reduce the computational cost, the estimation is performed with a reduced-state Viterbi algorithm. The channel effect is estimated in two different ways: either measuring the transducer response with an experimental setting or with blind homomorphic techniques. We observed an enhancement in image quality with respect to different metrics. Extensive tests were made to estimate the quantization alphabet that gives the best performances.