In real-world applications, audio surveillance is often performed by large models that can detect many types of anomalies. However, typical approaches are based on centralized solutions characterized by significant issues related to privacy and data transport costs. In addition, the large size of these models prevented a shift to contexts with limited resources, such as edge devices computing. In this work we propose conv-SPAD, a method for convolutional SPectral audio-based Anomaly Detection that takes advantage of common tools for spectral analysis and a simple autoencoder to learn the underlying condition of normality of real scenarios. Using audio data collected from real scenarios and artificially corrupted with anomalous sound events, we test the ability of the proposed model to learn normal conditions and detect anomalous events. It shows performances in line with larger models, often outperforming them. Moreover, the model's small size makes it usable in contexts with limited resources, such as edge devices hardware.

The report accounts for the activities of TC16 on "Algebraic and Discrete Mathematical Techniques in Pattern Recognition and Image Analysis" of the International Association for Pattern Recognition during the term 2021-2022. It has been submitted to Prof. Lele Akarun and evaluated by IAPR ExCo for the renewal of the Technical Committee 16 which was granted on July 14, 2022.

A new algorithm is presented to discriminate reconstructed hadronic decays of tau leptons (? h) that originate from genuine tau leptons in the CMS detector against ? h candidates that originate from quark or gluon jets, electrons, or muons. The algorithm inputs information from all reconstructed particles in the vicinity of a ? h candidate and employs a deep neural network with convolutional layers to efficiently process the inputs. This algorithm leads to a significantly improved performance compared with the previously used one. For example, the efficiency for a genuine ? h to pass the discriminator against jets increases by 10-30% for a given efficiency for quark and gluon jets. Furthermore, a more efficient ? h reconstruction is introduced that incorporates additional hadronic decay modes. The superior performance of the new algorithm to discriminate against jets, electrons, and muons and the improved ? h reconstruction method are validated with LHC proton-proton collision data at s = 13 TeV.

Illegal archaeological excavations, generally denoted as looting, is one of the most important damage factors to cultural heritage, as it upsets the human occupation stratigraphy of sites of archaeological interest. Looting identification and monitoring are not an easy task. A consolidated instrument used for the detection of archaeological features in general, and more specifically for the study of looting is remote sensing. Nevertheless, passive optical remote sensing is quite ineffective in dense vegetated areas. For these type of areas, in recent decades, LiDAR data and its derivatives have become an essential tool as they provide fundamental information that can be critical not only for the identification of unknown archaeological remains, but also for monitoring issues. Actually, LiDAR can suitably reveal grave robber devastation, even if, surprisingly, up today LiDAR has been generally unused for the identification of looting phenomenon. Consequently, this paper deals with an approach devised ad hoc for LiDAR data to detect looting. With this aim, some spatial visualization techniques and the geomorphon automatic landform extraction were exploited to enhance and extract features linked to the grave robber devastation. For this paper, the Etruscan site of San Giovenale (Northern Lazio, Italy) was selected as a test area as it is densely vegetated and was deeply plundered throughout the 20th century. Exploiting the LiDAR penetration capability, the prediction ability of the devised approach is highly satisfactory with a high rate of success, varying from 85-95%.

The report accounts for the activities of TC16 on "Algebraic and Discrete Mathematical Techniques in Pattern Recognition and Image Analysis" of the International Association for Pattern Recognition during the term 2019-2020.

Cancer research has significantly advanced in recent years mainly through developments in medical genomics and bioinformatics. It is expected that such approaches will result in more durable tumour control and fewer side effects compared with conventional treatments such as radiotherapy or chemotherapy. From the imaging standpoint, non-invasive imaging biomarkers (IBs) that assess angiogenic response and tumor environment at an early stage of therapy are of utmost importance, since they could provide useful insights into therapy planning. However, the extraction of IBs is still an open problem, since there are no standardized imaging protocols yet or established methods for the robust extraction of IBs. DCE-MRI is amongst the most promising non-invasive functional imaging modalities while compartmental pharmacokinetic (PK) modelling is the most common technique used for DCE-MRI data analysis. However, PK models suffer from a number of limitations such as modelling complexity, which often leads to variability in the computed biomarkers. To address these problems, alternative DCE-MRI biomarker extraction strategies coupled with a profound understanding of the physiological meaning of IBs is a sine qua non condition. To this end, a more recent model-free approach has been suggested in the literature for DCE-MRI data analysis, which relies on the shape classification of the time-signal uptake curves of image pixels in a selected tumour region of interest. This thesis is centred on this classification approach and the clinical question of whether model-free DCE-MRI data analysis has the potential to provide robust, clinically significant biomarkers using pattern recognition and image analysis techniques.

Nowadays, uterine fibroids can be treated using Magnetic Resonance guided Focused Ultrasound Surgery (MRgFUS), which is a non-invasive therapy exploiting thermal ablation. In order to measure the Non-Perfused Volume (NPV) for treatment response assessment, the ablated fibroid areas (i.e., Region of Treatment, ROT) are manually contoured by a radiologist. The current operator-dependent methodology could affect the subsequent follow-up phases, due to the lack of result repeatability. In addition, this fully manual procedure is time-consuming, considerably increasing execution times. These critical issues can be addressed only by means of accurate and efficient automated Pattern Recognition approaches. In this contribution, we evaluate two computer-assisted segmentation methods, which we have already developed and validated, for uterine fibroid segmentation in MRgFUS treatments. A quantitative comparison on segmentation accuracy, in terms of area-based and distance-based metrics, was performed. The clinical feasibility of these approaches was assessed from physicians' perspective, by proposing an integrated solution.

The main purpose of this study is to analyze the intrinsic tumor physiologic characteristics in patients with sarcoma through model-free analysis of dynamic contrast enhanced MR imaging data (DCE-MRI). Clinical data were collected from three patients with two different types of histologically proven sarcomas who underwent conventional and advanced MRI examination prior to excision. An advanced matrix factorization algorithm has been applied to the data, resulting in the identification of the principal time-signal uptake curves of DCE-MRI data, which were used to characterize the physiology of the tumor area, described by three different perfusion patterns i.e. hypoxic, well-perfused and necrotic one. The performance of the algorithm was tested by applying different initialization approaches with subsequent comparison of their results. The algorithm was proven to be robust and led to the consistent segmentation of the tumor area in three regions of different perfusion, i.e. well- perfused, hypoxic and necrotic. Results from the model-free approach were compared with a widely used pharmacokinetic (PK) model revealing significant correlations.

The paper summarizes the main results achieved during the three-year European FP7 ARROWS project (ARchaeological RObot systems for the Worlds Seas). ARROWS concluded at the end of August 2015 and proposed to adapt and develop low-cost Autonomous Underwater Vehicle (AUV) technologies to reduce the operational cost of typical underwater archaeological campaigns. The methodology used by ARROWS researchers identified archaeologists requirements for all the phases of a campaign. These were based on guidelines issued by the project Archaeology Advisory Group (AAG), which comprised of many European archaeologists belonging to the consortium. One of the main goals of the ARROWS project was the development of a heterogeneous team of cooperating AUVs; these comprised of prototypes developed in the project and commercially available vehicles. Three different AUVs have been built and tested at sea: MARTA, characterized by flexible hardware modularity for easy adaption of payload and propulsion systems, U-CAT, a turtle inspired bio-mimetic robot devoted to shipwreck penetration and A-Size AUV, a small light weight vehicle which is easily deployable by a single person. The project also included the development of a cleaning tool for well-known artefacts and maintenance operations. Results from the official final demonstrations of the project, held in Sicily and in Estonia during Summer 2015, are presented in the paper as an experimental proof of the validity of the developed robotic tools.

The huge amount of data, generated by daily-life data sources, represents a big opportunity for the development and advancement in several fields: scientific research, social life and industry. At the same time, analyzing these big repositories is a hard challenge, since the overload of information can overwhelm our capability of reading and understanding data, making finding useful pieces of information a difficult task. In this discussion we give a general overview about Knowledge Discovery in Databases as a scientific discipline that provides methodologies, techniques and tools for dealing with Big Data in order to find underlying knowledge that can be exploited in decision making processes.

Over the last decade, toxic events along the Mediterranean coast associated with exceptional harmful blooms of the dinoflagellate Ostreopsis cf. ovata have increased in frequency and distribution, causing not only the death of marine organisms and human health problems, but also economic loss on the tourism and aquaculture industries. In order to reduce the burden of routine algal counting, an innovative automated, low-cost, opto-electronic system called OvMeter was developed. It is able to speed up the monitoring process and therefore it enables early warning of incipient harmful algal blooms. An ad-hoc software tool provides automated cell recognition, counting and real-time calculation of the final algal concentration. The core of dinoflagellate recognition relies on a localization step which takes advantage of the synergistic exploitation of 2D bright-field and quantitative phase microscopy images, and a classification phase performed by a machine learning algorithm based on Boosted Trees approach. The architectural design of the OvMeter device is presented here, together with a performance evaluation on sea samples.

We demonstrate a 3D holographic tracking method to investigate particles motion in a microfluidic channel while unperturbed while inducing their migration through microfluidic manipulation. Digital holography (DH) in microscopy is a full-field, label-free imaging technique able to provide quantitative phase-contrast. The employed 3D tracking method is articulated in steps. First, the displacements along the optical axis are assessed by numerical refocusing criteria. In particular, an automatic refocusing method to recover the particles axial position is implemented employing a contrast-based refocusing criterion. Then, the transverse position of the in-focus object is evaluated through quantitative phase map segmentation methods and centroid-based 2D tracking strategy. The introduction of DH is thus suggested as a powerful approach for control of particles and biological samples manipulation, as well as a possible aid to precise design and implementation of advanced lab-on-chip microfluidic devices.

The integration of digital holography (DH) imaging and the acoustic manipulation of micro-particles in a microfluidic environment is investigated. The ability of DH to provide efficient 3D tracking of particles inside a microfluidic channel is exploited to measure the position of multiple objects moving under the effect of stationary ultrasound pressure fields. The axial displacement provides a direct verification of the numerically computed positions of the standing wave's node, while the particle's transversal movement highlights the presence of nodes in the planar direction. Moreover, DH is used to follow the aggregation dynamics of trapped spheres in such nodes by using aggregation rate metrics.

Synaptic activity in the nervous system consumes glucose. Therefore Fluorodeoxyglucose Positron Emission Tomography (FDG-PET) may in principle provide iconographic representations of glucose utilization impairment in neurodegenerative diseases and, specifically, in Amyotrophic Lateral Sclerosis (ALS). In a previous paper, we developed a computational method that applies a modern generalization of the Hough transform (HT) to identify the spinal canal and the spinal cord in Xray Computed Tomography (CT) images of ALS patients, and combines this information with the functional data provided by FDG-PET to measure the spinal marrow metabolism in detail. In that application, ellipses were used as prototypes for the HTbased recognition of the spinal cord profile and curves with three convexities as prototypes for the HT-based recognition of the spinal canal profile. In the present work, we provide a detailed description of the theoretical and computational tools at the basis of this approach to image integration, giving specific emphasis to the image processing steps necessary to make the structural information contained in the CT data actually determined by means of the HT procedure. Information inferred from the anatomical images have been integrated with functional information from PET images in order to quantitatively evaluate the metabolic activity of the spinal marrow in 30 control subjects and 30 ALS patients

The Hough transform is a standard pattern recognition technique introduced between the 1960s and the 1970s for the detection of straight lines, circles, and ellipses with several applications including the detection of symmetries in images. Recently, based on algebraic geometry arguments, the procedure has been extended to the automated recognition of special classes of algebraic plane curves. This allows us to detect curves of symmetry present in images, that is, curves that recognize midpoints maps of various shapes extracted by an ad hoc symmetry algorithm, here proposed. Further, in the case of straight lines, the detection of lines of symmetry allows us, by a pre-processing step of the image, to improve the efficiency of the recognition algorithm on which the Hough transform technique is founded, without loss of generality and additional computational costs.

One of the current challenges in Alzheimer's Disease (AD)-related research is to achieve an early and definite diagnosis. Automatic classification of AD is typically based on the use of feature vectors of high dimensionality, containing few training patterns, which leads to the curse-of-dimensionality problem. It is indispensable to find good approaches for selecting a subset of the original set of features. In this work, a method to perform early diagnosis of AD is proposed, combining different feature reduction approaches on both brain MRI studies and expression values of blood plasma proteins. Each selected set of features is used to train a Support Vector Machine (SVM), then the set of SVM is combined by weighted sum rule. Moreover, a novel approach for considering the feature vector as an image is proposed, different texture descriptors are extracted from the image and used to train a SVM. The superior performance of the proposed system is obtained without any ad hoc parameter optimization (i.e., the same ensemble of classifiers and the same parameter settings are used in all datasets). The MATLAB code for the ensemble of classifiers will be publicly available(3) to other researchers for future comparisons. (C) 2016 Elsevier B.V. All rights reserved.

Prediction is one of the main objectives of scientific analysis and it refers to both modelling and forecasting. The determination of the limits of predictability is an important issue of both theoretical and practical relevance. In the case of modelling time series, reached a certain level in performance in either modelling or prediction, it is often important to assess whether all the information available in the data has been exploited or whether there are still margins for improvement of the tools being developed. In this paper, an information theoretic approach is proposed to address this issue and quantify the quality of the models and/or predictions. The excellent properties of the proposed indicator have been proved with the help of a systematic series of numerical tests and a concrete example of extreme relevance for nuclear fusion.

Non-invasive imaging biomarkers that abeb angiogenic response and tumor microvascular environment at an early stage of therapy could provide useful insights into therapy planning. Tibue hypoxia is related to the insufficient supply of oxygen and is abociated with tumor vasculature and perfusion. Thus, knowledge of the hypoxic areas could be of great importance. There is no golden standard for imaging tumor hypoxia yet, however Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) is among the most promising non-invasive clinically relevant imaging modalities. In this work, DCE-MRI data from neck sarcoma are analyzed through a pattern recognition technique which results in the separation of the tumor area into well-perfused, hypoxic and necrotic regions.