Pooling publicly-available MRI data from multiple sites allows to assemble extensive groups of subjects, increase statistical power, and promote data reuse with machine learning techniques. The harmonization of multicenter data is necessary to reduce the confounding effect associated with non-biological sources of variability in the data. However, when applied to the entire dataset before machine learning, the harmonization leads to data leakage, because information outside the training set may affect model building, and potentially falsely overestimate performance. We propose a 1) measurement of the efficacy of data harmonization; 2) harmonizer transformer, i.e., an implementation of the ComBat harmonization allowing its encapsulation among the preprocessing steps of a machine learning pipeline, avoiding data leakage by design. We tested these tools using brain T1-weighted MRI data from 1740 healthy subjects acquired at 36 sites. After harmonization, the site effect was removed or reduced, and we showed the data leakage effect in predicting individual age from MRI data, highlighting that introducing the harmonizer transformer into a machine learning pipeline allows for avoiding data leakage by design.

In this work, we explore how to classify asteroids in co-orbital motion with a given planet using Machine Learning. We consider four different kinds of motion in mean motion resonance with the planet, nominally Tadpole at L4 and L5, Horseshoe and Quasi-Satellite, building 3 data sets defined as Real (taking the ephemerides of real asteroids from the JPL Horizons system), Ideal and Perturbed (both simulated, obtained by propagating initial conditions considering two different dynamical systems) for training and testing the Machine Learning algorithms in different conditions. The time series of the variable ? (angle related to the resonance) are studied with a data analysis pipeline defined ad hoc for the problem and composed by: data creation and annotation, time series features extraction thanks to the tsfresh package (potentially followed by selection and standardization) and the application of Machine Learning algorithms for Dimensionality Reduction and Classification. Such approach, based on features extracted from the time series, allows to work with a smaller number of data with respect to Deep Learning algorithms, also allowing to define a ranking of the importance of the features. Physical Interpretability of the features is another key point of this approach. In addition, we introduce the SHapley Additive exPlanations for Explainability technique. Different training and test sets are used, in order to understand the power and the limits of our approach. The results show how the algorithms are able to identify and classify correctly the time series, with a high degree of performance.

This book explores the application of Deep Convolutional Neural Networks to the field of ancient Egyptian Hieroglyphs transliteration. Such tools belong to the broader field known as Artificial Intelligence (AI), which is probably today the most impacting and disruptive technology we are working on. In the field of ancient languages transliteration and translation, AI applications are just at the beginning, despite they are spreading faster day after day. So, who knows which will be the impact of such technology in the years coming

The objective of this work is to show the application of a Deep Learning algorithm able to operate the segmentation of ancient Egyptian hieroglyphs present in an image, with the ambition to be as versatile as possible despite the variability of the image source. The problem is quite complex, the main obstacles being the considerable amount of different classes of existing hieroglyphs, the differences related to the hand of the scribe as well as the great differences among the various supports, such as papyri, stone or wood, where they are written. Furthermore, as in all archaeological finds, damage to the supports are frequent, with the consequence that hieroglyphs can be partially corrupted. In order to face this challenging problem, we leverage on the well-known Detectron2 platform, developed by the Facebook AI Research Group, focusing on the Mask R-CNN architecture to perform segmentation of image instances. Likewise, for several machine learning studies, one of the hardest challenges is the creation of a suitable dataset. In this paper, we will describe a hieroglyph dataset that has been created for the purpose of segmentation, highlighting its pros and cons, and the impact of different hyperparameters on the final results. Tests on the segmentation of images taken from public databases will also be presented and discussed along with the limitations of our study.

Whispering Gallery Mode (WGM) hollow microcavities turn out to be the site of an extremely rich and complex phenomenological scenario when pumped with a continuous-wave laser source. The coexistence of numerous non-linear and optomechanical effects have been reviewed in this paper. In our previous works we have investigated and described non-linear emissions as the stimulated Brillouin and Raman scattering, the degenerated and non-degenerated Kerr effects, such as four wave mixing (FWM). These effects happened concomitantly to parametric optomechanical oscillations which are the consequence of the radiation pressure. We have confirmed the regenerative oscillation of acoustic eigenmodes of the cavity leading to parametric instabilities and the activation of optomechanical chaotic oscillations. Finally, we have demonstrated that the blue-side excitation of WGM resonances lead to the chaos transition with a spectral evolution depending on the cavity size.

Radiomics and artificial intelligence have the potential to become a valuable tool in clinical applications. Frequently, radiomic analyses through machine learning methods present issues caused by high dimensionality and multicollinearity, and redundant radiomic features are usually removed based on correlation analysis. We assessed the effect of preprocessing--in terms of voxel size resampling, discretization, and filtering--on correlation-based dimensionality reduction in radiomic features from cardiac T1 and T2 maps of patients with hypertrophic cardiomyopathy. For different combinations of preprocessing parameters, we performed a dimensionality reduction of radiomic features based on either Pearson's or Spearman's correlation coefficient, followed by the computation of the stability index. With varying resampling voxel size and discretization bin width, for both T1 and T2 maps, Pearson's and Spearman's dimensionality reduction produced a slightly different percentage of remaining radiomic features, with a relatively high stability index. For different filters, the remaining features' stability was instead relatively low. Overall, the percentage of eliminated radiomic features through correlation-based dimensionality reduction was more dependent on resampling voxel size and discretization bin width for textural features than for shape or first-order features. Notably, correlation-based dimensionality reduction was less sensitive to preprocessing when considering radiomic features from T2 compared with T1 maps.

Nowadays, Deep Learning is advancing in any branch of knowledge, allowing to build tools supporting the work of experts in areas apparently far from the information technology field. In this study we exploit this possibility by focusing on ancient Egyptian hieroglyphic texts and inscriptions. In particular, we explore the ability of different convolutional neural networks (CNNs) to segment glyphs and classify pictures of ancient Egyptian hieroglyphs coming from different datasets of images. Regarding classification, three well-known CNN architectures (ResNet-50, Inception-v3 and Xception) were taken into consideration and trained on the available images, using both the paradigm of transfer learning and training from scratch. Moreover, modifying the architecture of one of the previous networks, we developed a specifically dedicated CNN, named Glyphnet, tailoring its complexity to our classification task. Performances were measured using standard metrics, giving significant results for all the tested networks, with the proposed Glyphnet outperforming the others, in terms of performance as well as ease of training and computational saving. The ancient hieroglyphs segmentation was faced in parallel, using a deep neural network architecture known as Mask-RCNN. This network was trained to segment the glyphs, identifying the bounding box, which will be the input for a network for classification. Even though in this paper we focused on the single hieroglyph segmentation and classification tasks, new and profitable perspectives are opened by the application of Deep Learning techniques in the Egyptological field. In this view, the proposed work can be seen as a starting point for the implementation of much more complex goals, such as: coding, recognition and transliteration of hieroglyphic signs; toposyntax of the hieroglyphic signs combined to form words; linguistics analysis of the hieroglyphic texts; recognition of corrupt, rewritten, and erased signs, towards even the identification of the "hand" of the scribe or the school of the sculptor. This work shows how the ancient Egyptian hieroglyphs identification task can be supported by the Deep Learning paradigm, laying the foundation for developing novel information tools for automatic documents recognition, classification and, most importantly, the language translation task.

Alzheimer's disease (AD) is the most common neurodegenerative disorder in the elderly with an incidence that progressively increases worldwide [1]. One of the main neuropathological hallmarks of AD is the presence of amyloid-b protein (Ab) aggregates which forms extracellular amyloid plaques [2]. At present, clinical diagnosis of AD relies on NINCDS-ADRDA (National Institute of Neurological and Communicative Diseases and Stroke/Alzheimer's Disease and Related Disorders Association) criteria that permit to classify the disease as possible or probable, but not definite (which still requires neuropathological examinations) [3]. This is partially due to the fact that the clinical, laboratory and instrumental biomarkers investigated are not specific for AD and can be altered in other neurodegenerative conditions [4]. In this work, we present an innovative approach in which a seed amplification assay (SAA) capable to detect traces of pathological Ab species in the cerebrospinal fluid (CSF) of patients with AD [5] is combined with Surface Enhanced Raman Spectroscopy for the ultrasensitive analysis of CSF collected from extensively-characterized patients with AD or other neurological conditions. Our findings show that SERS analysis of SAA end products through an optimized low-cost silver nanowires/PTFE SERS-active substrate [6,7], supported by machine learning approach [8] and correlated with the other clinical, instrumental and laboratory findings, could reveal chemo-structural information useful to distinguish AD from other neurological diseases in living patients.

Establishing standardized methods for a consistent analysis of spectral data remains a largely underexplored aspect in surface-enhanced Raman spectroscopy (SERS), particularly applied to biological and bio-medical research. Here we propose a Machine Learning (ML) based approach for classification of protein species. Principal Component Analysis (PCA), t-distributed stochastic neighbour embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP) where used for dimensionality reduction, along with supervised and unsupervised methods to quantify how closely resembled SERS spectral profiles belonging to different species (Albumin from bovine serum, Albumin from human serum, Lysozyme, Human holo-Transferrin, Human apo-Transferrin) are. In particular, ML algorithms such as Support Vector Machine, K-Nearest Neighbours, Linear Discriminant Analysis and the unsupervised K-means were applied to original and multipeak fitting on SERS spectra respectively. This strategy simultaneously assures a fast, full and successful discrimination of proteins and a thorough characterization of the chemo-structural differences among them, ultimately opening up new routes for SERS evolution toward sensing applications and diagnostics of interest in life sciences.

Thanks to their photophysical properties and the ease of synthesis and functionalization, gold nanoparticles (AuNPs) represent an ideal tool to develop colorimetric paper-based biosensors. Colloidal suspensions of AuNPs exhibit different colors depending on their size, shape and state of aggregation and their surface is suitable for functionalization with a wide variety of biomolecules. Here, we used anisotropic gold nanorods (AuNRs) for their multiplexability and their intrinsic brightness (10-fold higher than standard gold nanospheres), to label oligonucleotides for identifying a specific target DNA, both as PCR amplified fragment and as transgene into a cloning vector, by a dot-blot assay. The recognition of pathogenetic targets indeed represents a perspective of extreme interest in the clinical and environmental fields, e.g., to identify the microorganisms involved in infections and to trace the diffusion of antibiotic resistance or genetically modified organisms. To improve the analytical sensitivity and to obtain an automated and reproducible quantification of samples, we have also assessed the perspective to analyze dot-blot membranes with a supervised machine learning approach after a dedicated methodology for the acquisition of standardized photographs. Our work demonstrated the feasibility of a synergic use of plasmonic particles and artificial intelligence paradigms to accurately realize a rapid colorimetric paper-based detection.

Owing to their unique photophysical properties and ease of synthesis and functionalization, gold nanoparticles represent an excellent choice to develop colorimetric paper-based biosensors. Here, we selected anisotropic gold nanorods for their outstanding multiplexability and intrinsic brightness with respect to usual gold nanospheres. In particular, we compared two prominent strategies of functionalization with oligonucleotides to detect a specific DNA target, i.e. direct thiolation with a mercapto-probe, or amidation with an amino-probe in association with a carboxy-cross-linker. We found that the former provides much more consistent results, at the expense of greater complexity.

Paper-based biosensors featuring immunoconjugated gold nanoparticles have gained extraordinary momentum in recent times as the platform of choice in key cases of feld applications, including the so-called rapid antigen tests for SARS-CoV-2. Here, we propose a revision of this format, one that may leverage on the most recent advances in materials science and data processing. In particular, we target an amplifable DNA rather than a protein analyte, and we replace gold nanospheres with anisotropic nanorods, which are intrinsically brighter by a factor of~ 10, and multiplexable. By comparison with a gold-standard method for dot-blot readout with digoxigenin, we show that gold nanorods entail much faster and easier processing, at the cost of a higher limit of detection (from below 1 to 10 ppm in the case of plasmid DNA containing a target transgene, in our current setup). In addition, we test a complete workfow to acquire and process photographs of dot-blot membranes with custommade hardware and regression tools, as a strategy to gain more analytical sensitivity and potential for quantifcation. A leave-one-out approach for training and validation with as few as 36 sample instances already improves the limit of detection reached by the naked eye by a factor around 2. Taken together, we conjecture that the synergistic combination of new materials and innovative tools for data processing may bring the analytical sensitivity of paper-based biosensors to approach the level of lab-grade molecular tests.

NAVIGATOR is an Italian regional project to boost precision medicine in oncology with the aim to make it more predictive, preventive, and personalised by advancing translational research based on quantitative imaging and integrative omics analyses. The project's goal is to develop an open imaging biobank for the collection and preservation of a large amount of standardised imaging multimodal datasets, including computed tomography, magnetic resonance imaging, and positron emission tomography data, together with the corresponding patient-related and omics-related relevant information extracted from regional healthcare services using an adapted privacy-preserving model. The project is based on an open-source imaging biobank and an open-science oriented virtual research environment (VRE). Available integrative omics and multi-imaging data of three use cases (prostate cancer, rectal cancer, and gastric cancer) will be collected. All data confined in NAVIGATOR (i.e. standard and novel imaging biomarkers, non-imaging data, health agency data) will be used to create a digital patient model, to support the reliable prediction of the disease phenotype and risk stratification. The VRE that relies on a well-established infrastructure, called D4Science.org, will further provide a multiset infrastructure for processing the integrative omics data, extracting specific radiomic signatures, and for identification and testing of novel imaging biomarkers through big data analytics and artificial intelligence.

Researchers nowadays may take advantage of broad collections of medical data to develop personalized medicine solutions. Imaging bio-banks play a fundamental role, in this regard, by serving as organized repositories of medical images associated with imaging biomarkers. In this context, the NAVIGATOR Project aims to advance colorectal, prostate, and gastric oncology translational research by leveraging quantitative imaging and multi-omics analyses. As Project's core, an imaging bio-bank is being designed and implemented in a web-accessible Virtual Research Environment (VRE). The VRE serves to extract the imaging biomarkers and further process them within prediction algorithms. In our work, we present the realization of the data models for the three cancer use-cases of the Project. First, we carried out an extensive requirements analysis to fulfill the necessities of the clinical partners involved in the Project. Then, we designed three separate data models utilizing entity-relationship diagrams. We found diagrams' modeling for colorectal and prostate cancers to be more straightforward, while gastric cancer required a higher level of complexity. Future developments of this work would include designing a common data model following the Observational Medical Outcomes Partnership Standards. Indeed, a common data model would standardize the logical infrastructure of data models and make the bio-bank easily interoperable with other bio-banks.

Prostate cancer (PCa) is the most frequent male malignancy and the assessment of PCa aggressiveness, for which a biopsy is required, is fundamental for patient management. Currently, multiparametric (mp) MRI is strongly recommended before biopsy. Quantitative assessment of mpMRI might provide the radiologist with an objective and noninvasive tool for supporting the decision-making in clinical practice and decreasing intra- and inter-reader variability. In this view, high dimensional radiomics features and Machine Learning (ML) techniques, along with Deep Learning (DL) methods working on raw images directly, could assist the radiologist in the clinical workflow. The aim of this study was to develop and validate ML/DL frameworks on mpMRI data to characterize PCas according to their aggressiveness. We optimized several ML/DL frameworks on T2w, ADC and T2w+ADC data, using a patient-based nested validation scheme. The dataset was composed of 112 patients (132 peripheral lesions with Prostate Imaging Reporting and Data System (PI-RADS) score >= 3) acquired following both PI-RADS 2.0 and 2.1 guidelines. Firstly, ML/DL frameworks trained and validated on PI-RADS 2.0 data were tested on both PI-RADS 2.0 and 2.1 data. Then, we trained, validated and tested ML/DL frameworks on a multi PI-RADS dataset. We reported the performances in terms of Area Under the Receiver Operating curve (AUROC), specificity and sensitivity. The ML/DL frameworks trained on T2w data achieved the overall best performance. Notably, ML and DL frameworks trained and validated on PI-RADS 2.0 data obtained median AUROC values equal to 0.750 and 0.875, respectively, on unseen PI-RADS 2.0 test set. Similarly, ML/DL frameworks trained and validated on multi PI-RADS T2w data showed median AUROC values equal to 0.795 and 0.750, respectively, on unseen multi PI-RADS test set. Conversely, all the ML/DL frameworks trained and validated on PI-RADS 2.0 data, achieved AUROC values no better than the chance level when tested on PI-RADS 2.1 data. Both ML/DL techniques applied on mpMRI seem to be a valid aid in predicting PCa aggressiveness. In particular, ML/DL frameworks fed with T2w images data (objective, fast and non-invasive) show good performances and might support decision-making in patient diagnostic and therapeutic management, reducing intra- and inter-reader variability.

We describe the implementation of hierarchical materials made of hydrophilic micro-droplets in a silicone continuous phase to mimic the interactions of bio-Tissue with multiple physical agents, such as those implied in photoacoustic imaging.

Recently a new trend towards a more systematic use of reflectance Hyperspectral Imaging (HSI) has emerged in major museums. Extensive acquisition of HSI data opens up new research topics in terms of comparative analysis, creation and population of spectral databases, linking and crossing information. However, a full exploitation of these big-size data-sets unavoidably raises new issues about data-handling and processing methods. Along with statistical and multivariate analysis, solutions can be borrowed from the Artificial Intelligence (AI) area, using Machine Learning (ML) and Deep Learning (DL) methods. In this explorative study, different algorithms based on AI methods are applied to process HSI data acquired on three Picasso' paintings from the Museu Picasso collection in Barcellona. By using a "data-mining approach" the HSI-data are examined to unveil new correlations and extract embedded information.

Nowadays, advances in Artificial Intelligence (AI), especially in machine and deep learning, present new opportunities to build tools that support the work of specialists in areas apparently far from the information technology field. One example of such areas is that of ancient Egyptian hieroglyphic writing. In this study, we explore the ability of different convolutional neural networks (CNNs) to classify pictures of ancient Egyptian hieroglyphs coming from two different datasets of images. Three well-known CNN architectures (ResNet-50, Inception-v3 and Xception) were taken into consideration and trained on the available images. The paradigm of transfer learning was tested as well. In addition, modifying the architecture of one of the previous networks, we developed a specifically dedicated CNN, named Glyphnet, tailoring its complexity to our classification task. Performance comparison tests were carried out and Glyphnet showed the best performances with respect to the other CNNs. In conclusion, this work shows how the ancient Egyptian hieroglyphs identification task can be supported by the deep learning paradigm, laying the foundation for information tools supporting automatic documents recognition, classification and, most importantly, the language translation task.

We disclose a new composite featuring noble-metal nanorods in porous chitosan/polyvinyl alcohol mats or sponges for applications in wound healing and monitoring. The plasmonic component provides synergistic opportunities for the optical activation of functions as near-infrared laser welding, and the remote assessment of parameters of prognostic relevance in wound monitoring, like the environmental level of oxidative stress. At the same time, the polymer blend is ideal to bind connective tissue upon photothermal activation, and to support fabrication processes that ensure high porosity, such as electrospinning, thus paving the way to cellular repopulation and antimicrobial protection. In particular, we address the stabilization of the electrospun mats by cross-linking in a vapor of glutaraldehyde, and their cytotoxicity to a model of relevance in wound dressing like human fibroblasts.

Purpose: To perform a systematic review on the research on the application of artificial intelligence (AI) to imaging published in Italy and identify its fields of application, methods and results. Materials and Methods: A Pubmed search was conducted using terms Artificial Intelligence, Machine Learning, Deep learning, imaging, and Italy as affiliation, excluding reviews and papers outside time interval 2015-2020. In a second phase, participants of the working group AI4MP on Artificial Intelligence of the Italian Association of Physics in Medicine (AIFM) searched for papers on AI in imaging. Results: The Pubmed search produced 794 results. 168 studies were selected, of which 122 were from Pubmed search and 46 from the working group. The most used imaging modality was MRI (44%) followed by CT(12%) ad radiography/mammography (11%). The most common clinical indication were neurological diseases (29%) and diagnosis of cancer (25%). Classification was the most common task for AI (57%) followed by segmentation (16%). 65% of studies used machine learning and 35% used deep learning. We observed a rapid increase of research in Italy on artificial intelligence in the last 5 years, peaking at 155% from 2018 to 2019