A5031161800- AI in cancer detection
- Radiomics and Machine Learning in Medical Imaging
- Cell Image Analysis Techniques
- COVID-19 diagnosis using AI
- Generative Adversarial Networks and Image Synthesis
- Anomaly Detection Techniques and Applications
- Glioma Diagnosis and Treatment
- Colorectal Cancer Screening and Detection
- Gaze Tracking and Assistive Technology
- MRI in cancer diagnosis
- Image Processing Techniques and Applications
- Additive Manufacturing Materials and Processes
- Additive Manufacturing and 3D Printing Technologies
- Fault Detection and Control Systems
- Manufacturing Process and Optimization
- Industrial Vision Systems and Defect Detection
- Explainable Artificial Intelligence (XAI)
- Machine Learning and Data Classification
- Time Series Analysis and Forecasting
- EEG and Brain-Computer Interfaces
- Digital Transformation in Industry
- Neuroscience and Neural Engineering
- Advanced X-ray and CT Imaging
- Hemodynamic Monitoring and Therapy
- Digital Media Forensic Detection
Polytechnic University of Turin
2017-2025
IRD Fuel Cells (Denmark)
2024
Brescia University
2023
University of Brescia
2023
Geneva College
2023
WinnMed
2023
Turin Polytechnic University
2023
University of Turin
2022
The analysis of histological samples is paramount importance for the early diagnosis colorectal cancer (CRC).The traditional visual assessment time-consuming and highly unreliable because subjectivity evaluation.On other hand, automated extremely challenging due to variability architectural colouring characteristics images.In this work, we propose a deep learning technique based on Convolutional Neural Networks (CNNs) differentiate adenocarcinomas from healthy tissues benign lesions.Fully...
Computer-aided analysis of biological images typically requires extensive training on large-scale annotated datasets, which is not viable in many situations. In this paper, we present Generative Adversarial Network Discriminator Learner (GAN-DL), a novel self-supervised learning paradigm based the StyleGAN2 architecture, employ for image representation case fluorescent images.We show that Wasserstein Networks enable high-throughput compound screening raw images. We demonstrate by classifying...
Thanks to their capability learn generalizable descriptors directly from images, deep Convolutional Neural Networks (CNNs) seem the ideal solution most pattern recognition problems. On other hand, image representation, CNNs need huge sets of annotated samples that are unfeasible in many every-day scenarios. This is case, for example, Computer-Aided Diagnosis (CAD) systems digital pathology, where additional challenges posed by high variability cancerous tissue characteristics. In our...
Background: Multiple radiomics models have been proposed for grading glioma using different algorithms, features, and sequences of magnetic resonance imaging. The research seeks to assess the present overall performance glioma. Methods: A systematic literature review databases Ovid MEDLINE PubMed, EMBASE publications published on between 2012 2023 was performed. carried out following criteria Preferred Reporting Items Systematic Reviews Meta-Analysis. Results: In meta-analysis, a total 7654...
With the advent of digital pathology, there has been an increasing interest in providing pathologists with machine learning tools, often based on deep learning, to obtain faster and more robust image assessment. Nonetheless, accuracy these tools relies generation large training sets pre-labeled images. This is typically a challenging cumbersome process, requiring extensive pre-processing remove spurious samples that may lead failure. Unlike their plain counterparts, which tend provide...
Ultrasound (US) scans of inferior vena cava (IVC) are widely adopted by healthcare providers to assess patients’ volume status. Unfortunately, this technique is extremely operator dependent. Recently, new techniques have been introduced extract stable and objective information from US images automatic IVC edge tracking. However, these methods require prior interaction with the operator, which leads a waste time still makes partially subjective. In paper, two deep learning methods, YOLO (You...
So far, very little attention has been paid to the role of autonomic nervous system in augmentative alternative communication solutions. In this regard, pupil near reflex, one component triadic accommodative response a visual plane shift in-depth, may play key role. Such reflex does not necessitate any requirement skeletal muscles, and thus be preserved diseases affecting somatic motoneurons, such as amyotrophic lateral sclerosis. On basis, pupillary response, i.e. constriction far-to-near...
Abstract Despite the many advantages and increasing adoption of Electron Beam Powder Bed Fusion (PBF-EB) additive manufacturing by industry, current PBF-EB systems remain largely unstable prone to unpredictable anomalous behaviours. Additionally, although featuring in-situ process monitoring, show limited capabilities in terms timely identification failures, which may result into considerable wastage production time materials. These aspects are commonly recognized as barriers for industrial...
The Industry 4.0 paradigm has deeply changed classical manufacturing by introducing data-based analytics and decision-support strategies. At the state of art, data used for monitoring is mostly originated sensors, that undergo a fusion step to align different sources. However, this only relative monitored process, it does not include corresponding operating conditions parameters, are known Manufacturing Execution System (MES). Such information currently either included or labeled hand, thus...
Detecting complex anomalies on massive amounts of data is a crucial task in Industry 4.0, best addressed by deep learning. However, available solutions are computationally demanding, requiring cloud architectures prone to latency and bandwidth issues. This work presents VARADE, novel solution implementing light autoregressive framework based variational inference, which suited for real-time execution the edge. The proposed approach was validated robotic arm, part pilot production line,...
Renal Cell Carcinoma is typically asymptomatic at the early stages for many patients. This leads to a late diagnosis of tumor, where curability likelihood lower, and makes mortality rate high, with respect its incidence rate. To increase survival chance, fast correct categorization tumor subtype paramount. Nowadays, computerized methods, based on artificial intelligence, represent an interesting opportunity improve productivity objectivity microscopy-based diagnosis. Nonetheless, much their...
The prognosis of renal cell carcinoma (RCC) malignant neoplasms deeply relies on an accurate determination the histological subtype, which currently involves light microscopy visual analysis slides, considering notably tumor architecture and cytology. RCC subtyping is therefore a time-consuming tedious process, sometimes requiring expert review, with great impact diagnosis, treatment neoplasms. In this study, we investigate automatic classification 91 patients, diagnosed clear RCC, papillary...
Industry 4.0 involves the integration of digital technologies, such as IoT, Big Data, and AI, into manufacturing industrial processes to increase efficiency productivity. As these technologies become more interconnected interdependent, systems complex, which brings difficulty identifying stopping anomalies that may cause disturbances in process. This paper aims propose a diffusion-based model for real-time anomaly prediction processes. Using neuro-symbolic approach, we integrate ontologies...
The early detection of anomalous behaviors from a production line is fundamental aspect Industry 4.0, facilitated by the collection massive amounts data enabled Industrial Internet Things. Nonetheless, design and validation anomaly algorithms, mostly based on sophisticated Machine Learning models, heavily rely availability annotated datasets realistic anomalies, which very difficult to obtain in real line. To address this problem, we introduce Robotic Arm Dataset (RoAD), specifically...
To better define the overall performance of current radiomics-based models for discrimination pediatric posterior fossa tumors. A comprehensive literature search databases PubMed, Ovid MEDLINE, EMBASE, Web Science, and Scopus was designed conducted by an experienced librarian. We estimated sensitivity (SEN) specificity (SPE). Event rates were pooled across studies using a random-effects meta-analysis, χ2 test performed to assess heterogeneity. Overall SEN SPE differentiation between MB, PA,...
US-guided neuronavigation exploits the simplicity of use and minimal invasiveness Ultrasound (US) imaging high tissue resolution signal-to-noise ratio Magnetic Resonance Imaging (MRI) to guide brain surgeries.More specifically, intra-operative 3D US images are combined with pre-operative MR accurately localise course instruments in operative field invasiveness.Multimodal image registration is an essential part such system.In this paper, we present a complete software framework that enables...
Ideally, Convolutional Neural Networks (CNNs) should be trained with high quality images minimum noise and correct ground truth labels. Nonetheless, in many real-world scenarios, such is very hard to obtain, datasets may affected by any sort of image degradation mislabelling issues. This negatively impacts the performance standard CNNs, both during training inference phase. To address this issue we propose Wise2WipedNet (W2WNet), a new two-module Network, where Wise module exploits Bayesian...