arXiv subjects

Giuseppe Jurman

Publications and source records attributed to Giuseppe Jurman.

At least 19 recordsLinked to original sources

arXiv2025

AI-based modular warning machine for risk identification in proximity healthcare

"DHEAL-COM - Digital Health Solutions in Community Medicine" is a research and technology project funded by the Italian Department of Health for the development of digital solutions of interest in proximity healthcare. The activity within the DHEAL-COM framework allows scientists to gather a notable amount of multi-modal data whose interpretation can be performed by means of machine learning algorithms. The present study illustrates a general automated pipeline made of numerous unsupervised and supervised methods that can ingest such data, provide predictive results, and facilitate model interpretations via feature identification.

cs.LG↗

arXiv2025

Deconver: A Deconvolutional Network for Medical Image Segmentation

While convolutional neural networks (CNNs) and vision transformers (ViTs) have advanced medical image segmentation, they face inherent limitations such as local receptive fields in CNNs and high computational complexity in ViTs. This paper introduces Deconver, a novel network that integrates traditional deconvolution techniques from image restoration as a core learnable component within a U-shaped architecture. Deconver replaces computationally expensive attention mechanisms with efficient nonnegative deconvolution (NDC) operations, enabling the restoration of high-frequency details while suppressing artifacts. Key innovations include a backpropagation-friendly NDC layer based on a provably monotonic update rule and a parameter-efficient design. Evaluated across four datasets (ISLES'22, BraTS'23, GlaS, FIVES) covering both 2D and 3D segmentation tasks, Deconver achieves state-of-the-art performance in Dice scores and Hausdorff distance while reducing computational costs (FLOPs) by up to 90% compared to leading baselines. By bridging traditional image restoration with deep learning, this work offers a practical solution for high-precision segmentation in resource-constrained clinical workflows. The project is available at https://github.com/pashtari/deconver.

eess.IV↗

arXiv2022

Mitigating Health Data Poverty: Generative Approaches versus Resampling for Time-series Clinical Data

Several approaches have been developed to mitigate algorithmic bias stemming from health data poverty, where minority groups are underrepresented in training datasets. Augmenting the minority class using resampling (such as SMOTE) is a widely used approach due to the simplicity of the algorithms. However, these algorithms decrease data variability and may introduce correlations between samples, giving rise to the use of generative approaches based on GAN. Generation of high-dimensional, time-series, authentic data that provides a wide distribution coverage of the real data, remains a challenging task for both resampling and GAN-based approaches. In this work we propose CA-GAN architecture that addresses some of the shortcomings of the current approaches, where we provide a detailed comparison with both SMOTE and WGAN-GP*, using a high-dimensional, time-series, real dataset of 3343 hypotensive Caucasian and Black patients. We show that our approach is better at both generating authentic data of the minority class and remaining within the original distribution of the real data.

cs.LG↗

arXiv2020

AI slipping on tiles: data leakage in digital pathology

Reproducibility of AI models on biomedical data still stays as a major concern for their acceptance into the clinical practice. Initiatives for reproducibility in the development of predictive biomarkers as the MAQC Consortium already underlined the importance of appropriate Data Analysis Plans (DAPs) to control for different types of bias, including data leakage from the training to the test set. In the context of digital pathology, the leakage typically lurks in weakly designed experiments not accounting for the subjects in their data partitioning schemes. This issue is then exacerbated when fractions or subregions of slides (i.e. "tiles") are considered. Despite this aspect is largely recognized by the community, we argue that it is often overlooked. In this study, we assess the impact of data leakage on the performance of machine learning models trained and validated on multiple histology data collection. We prove that, even with a properly designed DAP (10x5 repeated cross-validation), predictive scores can be inflated up to 41% when tiles from the same subject are used both in training and validation sets by deep learning models. We replicate the experiments for $4$ classification tasks on 3 histopathological datasets, for a total of 374 subjects, 556 slides and more than 27,000 tiles. Also, we discuss the effects of data leakage on transfer learning strategies with models pre-trained on general-purpose datasets or off-task digital pathology collections. Finally, we propose a solution that automates the creation of leakage-free deep learning pipelines for digital pathology based on histolab, a novel Python package for histology data preprocessing. We validate the solution on two public datasets (TCGA and GTEx).

q-bio.QM↗

arXiv2019

MASS-UMAP: Fast and accurate analog ensemble search in weather radar archive

The use of analogs - similar weather patterns - for weather forecasting and analysis is an established method in meteorology. The most challenging aspect of using this approach in the context of operational radar applications is to be able to perform a fast and accurate search for similar spatiotemporal precipitation patterns in a large archive of historical records. In this context, sequential pairwise search is too slow and computationally expensive. Here we propose an architecture to significantly speed-up spatiotemporal analog retrieval by combining nonlinear geometric dimensionality reduction (UMAP) with the fastest known Euclidean search algorithm for time series (MASS) to find radar analogs in constant time, independently of the desired temporal length to match and the number of extracted analogs. We compare UMAP with Principal component analysis (PCA) and show that UMAP outperforms PCA for spatial MSE analog search with proper settings. Moreover, we show that MASS is 20 times faster than brute force search on the UMAP embeddings space. We test the architecture on a real dataset and show that it enables precise and fast operational analog ensemble search through more than 2 years of radar archive in less than 5 seconds on a single workstation.

cs.CY↗

arXiv2019

High Resolution Forecasting of Heat Waves impacts on Leaf Area Index by Multiscale Multitemporal Deep Learning

Climate change impacts could cause progressive decrease of crop quality and yield, up to harvest failures. In particular, heat waves and other climate extremes can lead to localized food shortages and even threaten food security of communities worldwide. In this study, we apply a deep learning architecture for high resolution forecasting (300 m, 10 days) of the Leaf Area Index (LAI), whose dynamics has been widely used to model the growth phase of crops and impact of heat waves. LAI models can be computed at 0.1 degree spatial resolution with an auto regressive component adjusted with weather conditions, validated with remote sensing measurements. However model actionability is poor in regions of varying terrain morphology at this scale (about 8 km at the Alps latitude). Our deep learning model aims instead at forecasting LAI by training multiscale multitemporal (MSMT) data from the Copernicus Global Land Service (CGLS) project for all Europe at 300m resolution and medium-resolution historical weather data. Further, the deep learning model inputs integrate high-resolution land surface features, known to improve forecasts of agricultural productivity. The historical weather data are then replaced with forecast values to predict LAI values at 10 day horizon on Europe. We propose the MSMT model to develop a high resolution crop-specific warning system for mitigating damage due to heat waves and other extreme events.

q-bio.PE↗

arXiv2018

A multiobjective deep learning approach for predictive classification in Neuroblastoma

Neuroblastoma is a strongly heterogeneous cancer with very diverse clinical courses that may vary from spontaneous regression to fatal progression; an accurate patient's risk estimation at diagnosis is essential to design appropriate tumor treatment strategies. Neuroblastoma is a paradigm disease where different diagnostic and prognostic endpoints should be predicted from common molecular and clinical information, with increasing complexity, as shown in the FDA MAQC-II study. Here we introduce the novel multiobjective deep learning architecture CDRP (Concatenated Diagnostic Relapse Prognostic) composed by 8 layers to obtain a combined diagnostic and prognostic prediction from high-throughput transcriptomics data. Two distinct loss functions are optimized for the Event Free Survival (EFS) and Overall Survival (OS) prognosis, respectively. We use the High-Risk (HR) diagnostic information as an additional input generated by an autoencoder embedding. The latter is used as network regulariser, based on a clinical algorithm commonly adopted for stratifying patients from cancer stage, age at insurgence of disease, and MYCN, the specific molecular marker. The architecture was applied to Illumina HiSeq2000 RNA-Seq for 498 neuroblastoma patients (176 at high risk) from the Sequencing Quality Control (SEQC) study, obtaining state-of-art on the diagnostic endpoint and improving prediction of prognosis over the HR cohort.

q-bio.QM↗

arXiv2018

Multilayer flows in molecular networks identify biological modules in the human proteome

A variety of complex systems exhibit different types of relationships simultaneously that can be modeled by multiplex networks. A typical problem is to determine the community structure of such systems that, in general, depend on one or more parameters to be tuned. In this study we propose one measure, grounded on information theory, to find the optimal value of the relax rate characterizing Multiplex Infomap, the generalization of the Infomap algorithm to the realm of multilayer networks. We evaluate our methodology on synthetic networks, to show that the most representative community structure can be reliably identified when the most appropriate relax rate is used. Capitalizing on these results, we use this measure to identify the most reliable meso-scale functional organization in the human protein-protein interaction multiplex network and compare the observed clusters against a collection of independently annotated gene sets from the Molecular Signatures Database (MSigDB). Our analysis reveals that modules obtained with the optimal value of the relax rate are biologically significant and, remarkably, with higher functional content than the ones obtained from the aggregate representation of the human proteome. Our framework allows us to characterize the meso-scale structure of those multilayer systems whose layers are not explicitly interconnected each other -- as in the case of edge-colored models -- the ones describing most biological networks, from proteomes to connectomes.

q-bio.MN↗

arXiv2017

Convolutional neural networks for structured omics: OmicsCNN and the OmicsConv layer

Convolutional Neural Networks (CNNs) are a popular deep learning architecture widely applied in different domains, in particular in classifying over images, for which the concept of convolution with a filter comes naturally. Unfortunately, the requirement of a distance (or, at least, of a neighbourhood function) in the input feature space has so far prevented its direct use on data types such as omics data. However, a number of omics data are metrizable, i.e., they can be endowed with a metric structure, enabling to adopt a convolutional based deep learning framework, e.g., for prediction. We propose a generalized solution for CNNs on omics data, implemented through a dedicated Keras layer. In particular, for metagenomics data, a metric can be derived from the patristic distance on the phylogenetic tree. For transcriptomics data, we combine Gene Ontology semantic similarity and gene co-expression to define a distance; the function is defined through a multilayer network where 3 layers are defined by the GO mutual semantic similarity while the fourth one by gene co-expression. As a general tool, feature distance on omics data is enabled by OmicsConv, a novel Keras layer, obtaining OmicsCNN, a dedicated deep learning framework. Here we demonstrate OmicsCNN on gut microbiota sequencing data, for Inflammatory Bowel Disease (IBD) 16S data, first on synthetic data and then a metagenomics collection of gut microbiota of 222 IBD patients.

q-bio.QM↗

arXiv2017

Deep Learning for Automatic Stereotypical Motor Movement Detection using Wearable Sensors in Autism Spectrum Disorders

Autism Spectrum Disorders are associated with atypical movements, of which stereotypical motor movements (SMMs) interfere with learning and social interaction. The automatic SMM detection using inertial measurement units (IMU) remains complex due to the strong intra and inter-subject variability, especially when handcrafted features are extracted from the signal. We propose a new application of the deep learning to facilitate automatic SMM detection using multi-axis IMUs. We use a convolutional neural network (CNN) to learn a discriminative feature space from raw data. We show how the CNN can be used for parameter transfer learning to enhance the detection rate on longitudinal data. We also combine the long short-term memory (LSTM) with CNN to model the temporal patterns in a sequence of multi-axis signals. Further, we employ ensemble learning to combine multiple LSTM learners into a more robust SMM detector. Our results show that: 1) feature learning outperforms handcrafted features; 2) parameter transfer learning is beneficial in longitudinal settings; 3) using LSTM to learn the temporal dynamic of signals enhances the detection rate especially for skewed training data; 4) an ensemble of LSTMs provides more accurate and stable detectors. These findings provide a significant step toward accurate SMM detection in real-time scenarios.

cs.CV↗

arXiv2017

Phylogenetic Convolutional Neural Networks in Metagenomics

Background: Convolutional Neural Networks can be effectively used only when data are endowed with an intrinsic concept of neighbourhood in the input space, as is the case of pixels in images. We introduce here Ph-CNN, a novel deep learning architecture for the classification of metagenomics data based on the Convolutional Neural Networks, with the patristic distance defined on the phylogenetic tree being used as the proximity measure. The patristic distance between variables is used together with a sparsified version of MultiDimensional Scaling to embed the phylogenetic tree in a Euclidean space. Results: Ph-CNN is tested with a domain adaptation approach on synthetic data and on a metagenomics collection of gut microbiota of 38 healthy subjects and 222 Inflammatory Bowel Disease patients, divided in 6 subclasses. Classification performance is promising when compared to classical algorithms like Support Vector Machines and Random Forest and a baseline fully connected neural network, e.g. the Multi-Layer Perceptron. Conclusion: Ph-CNN represents a novel deep learning approach for the classification of metagenomics data. Operatively, the algorithm has been implemented as a custom Keras layer taking care of passing to the following convolutional layer not only the data but also the ranked list of neighbourhood of each sample, thus mimicking the case of image data, transparently to the user. Keywords: Metagenomics; Deep learning; Convolutional Neural Networks; Phylogenetic trees

q-bio.QM↗

arXiv2017

Towards meaningful physics from generative models

In several physical systems, important properties characterizing the system itself are theoretically related with specific degrees of freedom. Although standard Monte Carlo simulations provide an effective tool to accurately reconstruct the physical configurations of the system, they are unable to isolate the different contributions corresponding to different degrees of freedom. Here we show that unsupervised deep learning can become a valid support to MC simulation, coupling useful insights in the phases detection task with good reconstruction performance. As a testbed we consider the 2D XY model, showing that a deep neural network based on variational autoencoders can detect the continuous Kosterlitz-Thouless (KT) transitions, and that, if endowed with the appropriate constrains, they generate configurations with meaningful physical content.

hep-lat↗

arXiv2016

Convolutional Neural Network for Stereotypical Motor Movement Detection in Autism

Autism Spectrum Disorders (ASDs) are often associated with specific atypical postural or motor behaviors, of which Stereotypical Motor Movements (SMMs) have a specific visibility. While the identification and the quantification of SMM patterns remain complex, its automation would provide support to accurate tuning of the intervention in the therapy of autism. Therefore, it is essential to develop automatic SMM detection systems in a real world setting, taking care of strong inter-subject and intra-subject variability. Wireless accelerometer sensing technology can provide a valid infrastructure for real-time SMM detection, however such variability remains a problem also for machine learning methods, in particular whenever handcrafted features extracted from accelerometer signal are considered. Here, we propose to employ the deep learning paradigm in order to learn discriminating features from multi-sensor accelerometer signals. Our results provide preliminary evidence that feature learning and transfer learning embedded in the deep architecture achieve higher accurate SMM detectors in longitudinal scenarios.

cs.NE↗

arXiv2016

Differential network analysis and graph classification: a glocal approach

Based on the glocal HIM metric and its induced graph kernel, we propose a novel solution in differential network analysis that integrates network comparison and classification tasks. The HIM distance is defined as the one-parameter family of product metrics linearly combining the normalised Hamming distance H and the normalised Ipsen-Mikhailov spectral distance IM. The combination of the two components within a single metric allows overcoming their drawbacks and obtaining a measure that is simultaneously global and local. Furthermore, plugging the HIM kernel into a Support Vector Machine gives us a classification algorithm based on the HIM distance. First, we outline the theory underlying the metric construction. We introduce two diverse applications of the HIM distance and the HIM kernel to biological datasets. This versatility supports the adoption of the HIM family as a general tool for information extraction, quantifying difference among diverse in- stances of a complex system. An Open Source implementation of the HIM metrics is provided by the R package nettols and in its web interface ReNette.

q-bio.MN↗

arXiv2016

Metric projection for dynamic multiplex networks

Evolving multiplex networks are a powerful model for representing the dynamics along time of different phenomena, such as social networks, power grids, biological pathways. However, exploring the structure of the multiplex network time series is still an open problem. Here we propose a two-steps strategy to tackle this problem based on the concept of distance (metric) between networks. Given a multiplex graph, first a network of networks is built for each time steps, and then a real valued time series is obtained by the sequence of (simple) networks by evaluating the distance from the first element of the series. The effectiveness of this approach in detecting the occurring changes along the original time series is shown on a synthetic example first, and then on the Gulf dataset of political events.

physics.soc-ph↗

arXiv2015

Seasonal Linear Predictivity in National Football Championships

Predicting the results of sport matches and competitions is an arising research field, benefiting from the growing amount of available data and the novel data analytics techniques. Excellent forecasts can be achieved by advanced machine learning methods applied to detailed historical data, especially in very popular sports such as football (soccer). Here we show that, despite the large number of confounding factors, the results of a football team in longer competitions (e.g., a national league) follow a basically linear trend useful for predictive purposes, too. In support of this claim, we present a set of experiments of linear regression on a database collecting the yearly results of 707 teams playing in 22 divisions from 11 countries, in 20 football seasons.

stat.AP↗

arXiv2015

Community dynamics in connected time-dependent multilayer networks

Different strategies have been considered to extract information from social media about how similarly people react to the same news or event. In this context, a powerful method is offered by the application of graph techniques to the contents produced by social network users. In particular, large events typically attract enough content traffic along time to enable an analysis that explicitly models a dependence from the time dimension. Here we demonstrate how it is possible to extend the application of community detection strategies in complex networks to the case of time-dependent multilayer networks, whenever the connection between consecutive time layers is non-trivial. We apply the method to 400K Twitter post related to the Expo event held in Milan (Italy) between May and October 2015.

cs.SI↗

arXiv2014

Entropy Dynamics of Community Alignment in the Italian Parliament Time-Dependent Network

Complex institutions are typically characterized by meso-scale structures which are fundamental for the successful coordination of multiple agents. Here we introduce a framework to study the temporal dynamics of the node-community relationship based on the concept of community alignment, a measure derived from the modularity matrix that defines the alignment of a node with respect to the core of its community. The framework is applied to the 16th legislature of the Italian Parliament to study the dynamic relationship in voting behavior between Members of the Parliament (MPs) and their political parties. As a novel contribution, we introduce two entropy-based measures that capture politically interesting dynamics: the group alignment entropy (over a single snapshot), and the node alignment entropy (over multiple snapshots). We show that significant meso-scale changes in the time-dependent network structures can be detected by a combination of the two measures. We observe a steady growth of the group alignment entropy after a major internal conflict in the ruling majority and a different distribution of nodes alignment entropy after the government transition.

cs.SI↗