Publications

The Bag of Words paradigm has been the baseline from which several successful image classification solutions were developed in the last decade. These represent images by quantizing local descriptors and summarizing their distribution. The quantization step introduces a dependency on the dataset, that even if in some contexts significantly boosts the performance, severely limits its generalization capabilities. Differently, in this paper, we propose to model the local features distribution with a multivariate Gaussian, without any quantization. The full rank covariance matrix, which lies on a Riemannian manifold, is projected on the tangent Euclidean space and concatenated to the mean vector. The resulting representation, a Gaussian of local descriptors (GOLD), allows to use the dot product to closely approximate a distance between distributions without the need for expensive kernel computations. We describe an image by an improved spatial pyramid, which avoids boundary effects with soft assignment: local descriptors contribute to neighboring Gaussians, forming a weighted spatial pyramid of GOLD descriptors. In addition, we extend the model leveraging dataset characteristics in a mixture of Gaussian formulation further improving the classification accuracy. To deal with large scale datasets and high dimensional feature spaces the Stochastic Gradient Descent solver is adopted. Experimental results on several publicly available datasets show that the proposed method obtains state-of-the-art performance.

Scene detection is a fundamental tool for allowing effective video browsing and re-using. In this paper we present a model that automatically divides videos into coherent scenes, which is based on a novel combination of local image descriptors and temporal clustering techniques. Experiments are performed to demonstrate the effectiveness of our approach, by comparing our algorithm against two recent proposals for automatic scene segmentation. We also propose improved performance measures that aim to reduce the gap between numerical evaluation and expected results.

The automatization of the analysis of Indirect Immunofluorescence (IIF) images is of paramount importance for the diagnosis of autoimmune diseases. This paper proposes a solution to one of the most challenging steps of this process, the segmentation of HEp-2 cells, through an adaptive marker-controlled watershed approach. Our algorithm automatically conforms the marker selection pipeline to the peculiar characteristics of the input image, hence it is able to cope with different fluorescent intensities and staining patterns without any a priori knowledge. Furthermore, it shows a reduced sensitivity to over-segmentation errors and uneven illumination, that are typical issues of IIF imaging.

Following the recent Internet of Everything (IoE) trend, several general-purpose devices have been proposed to acquire as much information as possible from the environment and from people interacting with it. Among the others, sensing floors are recently attracting the interest of the research community. In this paper, we propose a new model to store and process floor data. The model does not assume a regular grid distribution of the sensing elements and is based on the ground reaction force (GRF) concept, widely used in biomechanics. It allows the correct detection and tracking of people, outperforming the common background subtraction schema adopted in the past. Several tests on a real sensing floor prototype are reported and discussed

In this paper we evaluate the performance of scene detection techniques, starting from the classic precision/recall approach, moving to the better designed coverage/overflow measures, and finally proposing an improved metric, in order to solve frequently observed cases in which the numeric interpretation is different from the expected results. Numerical evaluation is performed on two recent proposals for automatic scene detection, and comparing them with a simple but effective novel approach. Experimental results are conducted to show how different measures may lead to different interpretations.

Although anaplastic large-cell lymphomas (ALCL) carrying anaplastic lymphoma kinase (ALK) have a relatively good prognosis, aggressive forms exist. We have identified a novel translocation, causing the fusion of the TRAF1 and ALK genes, in one patient who presented with a leukemic ALK+ ALCL (ALCL-11). To uncover the mechanisms leading to high-grade ALCL, we developed a human patient-derived tumorgraft (hPDT) line. Molecular characterization of primary and PDT cells demonstrated the activation of ALK and nuclear factor kB (NFkB) pathways. Genomic studies of ALCL-11 showed the TP53 loss and the in vivo subclonal expansion of lymphoma cells, lacking PRDM1/Blimp1 and carrying c-MYC gene amplification. The treatment with proteasome inhibitors of TRAF1-ALK cells led to the downregulation of p50/p52 and lymphoma growth inhibition. Moreover, a NFkB gene set classifier stratified ALCL in distinct subsets with different clinical outcome. Although a selective ALK inhibitor (CEP28122) resulted in a significant clinical response of hPDT mice, nevertheless the disease could not be eradicated. These data indicate that the activation of NFkB signaling contributes to the neoplastic phenotype of TRAF1-ALK ALCL. ALCL hPDTs are invaluable tools to validate the role of druggable molecules, predict therapeutic responses and implement patient specific therapies.

The advent of NGS dramatically changed the characterisation of multifactorial pathologies such as cancer. The high molecular variability of cancer makes essential the identification of biomarkers able to explain the differences among cancer sub-types, allowing physicians to provide patients with suitable therapies. In this context, miRNAs are considered adequate biomarkers and miRNAs profiling from miRNA-sequencing is widely used. However, state of the art tools performing miRNAs reads mapping rely on general-purpose alignment algorithms. On the other side, researches carried out in the last decade led to the identification of many miRNAs specific features that are not exploited by miRNAs aligner. Moreover, the role of miRNAs variants called ‘isomiRs’ is still an open issue. IsomiRs impact miRNA targets affinity characterization and their analysis enables a more accurate evaluation of miRNA expression profiles. In light of these considerations, there is need of algorithmic methodologies able to provide users with a complete and accurate picture of the whole miRNAs, isomiRs and interaction sites spectrum. We report the impact of the application of such methodology on 23 human miRNA-Seq datasets from GEO, for which the overall isomiRs expression level and the characteristics of the interaction sites has been evaluated. As a result, 40% of the 189M miRNAs mapped reads showed a miRNA exact sequence, whereas 50% are characterized by a sequence accounting for 3’ isomiRs and the remaining reads possess sequences compatible with 5’ and SNP isomiRs or combinations of them. Furthermore, in the 2% of the cases some interaction sites are missed. Two other samples (hESCs and NSCs), recently analysed to confirm isomiRs importance, have been also studied in terms of isomiRs and interaction sites profiles, pointing out that such characteristics require a suitable methodology for miRNA sequences analysis because they cannot be appreciated from the overall miRNAs expression profile.

JAK/STAT3 signaling pathway is often deregulated in hematopoietic disorders including peripheral T-cell lymphoma. We describe two novel mechanisms leading to the constitutive activation of STAT3 in ALK- ALCL. Oncogenic JAK1 or STAT3 mutations are associated to hyperactive pSTAT3 that regulated canonical STAT3 and ATF3 genes. Moreover, synergizing JAK1 and STAT3 mutants sustain the neoplastic growth, which can be efficiently controlled in vitro and in an ALCL patient derived tumorgraft model by JAK1/2 inhibitors. We have discovered that novel chimera, displaying concomitant transcriptional and kinase activities, are power oncogenes capable to sustain via STAT3 the ALCL phenotype and can be uniquely neutralized by a novel ROS1 inhibitor. The pharmacological inhibition of JAK/STAT3 represents a novel strategy for the treatment of molecular stratified ALCL.