Category: Senza categoria (Page 2 of 4)

Artificial intelligence for early diagnosis and clinical decision making in neurodegenerative disorders

14 January 2021 /

Federica Agosta

Abstract

Magnetic resonance imaging (MRI) is playing an increasingly important role in the study of neurodegenerative diseases, delineating the structural and functional alterations determined by these conditions. Advanced MRI techniques are of special interest for their potential to characterize the signature of each neurodegenerative condition and aid both the diagnostic process and the monitoring of disease progression. This aspect will become crucial when disease-modifying (personalized) therapies will be established. In the past decade, artificial intelligence has been applied enthusiastically in the field of medicine, outperforming other established methods. In the field of neurodegenerative diseases and biomarkers, artificial intelligence algorithms applied to MRI have proven their worth in many ways including aiding the image-based prediction of different neurological diseases, the anatomical segmentation of specific brain structures, and the discovery and development of new therapies.

Short Bio

Federica Agosta was born in Milano on 01/04/1978, took her Graduation in Medicine in 2003, Post-Degree Graduation in Neurology in 2008, and PhD in Experimental Neurology in 2012. She is an Assistant Professor of Neurology at Vita-Salute San Raffaele University and Group Leader of the Neuroimaging of Neurodegenerative Diseases Unit at the Institute of Experimental Neurology, Division of Neuroscience, Ospedale San Raffaele (OSR), Milano, where she conducts research in patients with neurodegenerative conditions. She has a broad background in clinical neurology and neuroimaging, with specific training and expertise in MRI and neurodegenerative diseases. In the period 2002-2007, as junior research fellow at the Neuroimaging Research Unit, OSR, she collaborated in many research projects on the use of quantitative MR techniques in the study of normal aging, multiple sclerosis, amyotrophic lateral sclerosis (ALS), and Parkinson’s disease. During those years, she improved her skills on acquisition and post-processing techniques applied to functional MRI, diffusion tensor MRI and morphometry. This knowledge was instrumental in the success of her visiting fellowship at the Memory and Aging Center, UCSF in 2007-2008, during which she expanded her understanding of Alzheimer’s disease, frontotemporal dementia, and primary progressive aphasia. During her PhD at Vita-Salute San Raffaele University (2009-2012), she dealt with several aspects of pathophysiology of neurodegenerative diseases using MR techniques, with particular interest in young onset dementia and ALS. Dr. Agosta has been member of the Steering Committee of the Italian Society for the Study of Dementia (SINdem) from 2016 to 2018 and then she held the position of Secretary of the society from 2018 to 2020. In SINdem, she is also the coordinator of the Neuroimaging Study Group. Federica Agosta is also Chair of the Neuroimaging Society in ALS (NISALS) and of the Neuroimaging Panel of the European Academy of Neurology. She is also participating actively to the Neuroimaging Subcommittee of the Italian Neurological Society. Her research has led to the publication of over 200 Pubmed-referenced papers (H Index 59, Scopus). Since 2016, Federica Agosta is Section Editor of the NeuroImage: Clinical journal. In 2010, Dr. Agosta has been selected as young researcher participant for the 60th Interdisciplinary Meeting of Nobel Laureates for her outstanding contribution to the application of MRI techniques to the study of neurological diseases. In 2016, she has been awarded an ERC Starting Grant.

Keywords

Neurodegenerative diseases, MRI, Diagnosis, Stratification, Prediction

Unstructured data, ML and AI for healthcare and industry 4.0 applications

13 January 2021 /

Donatello Apelusion Gassi, Giuseppe Leonardo Cascella

Abstract

In this keynote, the state of the art of AWS’ tools for AI and ML application in healthcare will be considered. Thanks to their flexibility and scalability, Idea75 will also show how the same tools can be successfully applied in IoT and industry4.0 solutions. Real-world use cases of unstructured data, information extraction, custom models, and deep learning will be discussed.

Short Bio

Donatello A. Gassi is the Head of Solutions Architecture for the EMEA Telecom Vertical at AWS and is responsible for creating architectural best practices and working with Telco customers on how they use the cloud and innovation to transform their own businesses or reinvent their markets. Donatello is a business orientated technologist with nearly 20 years of experience in the technology industry in a variety of roles. Prior to joining AWS, Donatello was Senior Manager with a global consulting firm, where he contributed to the creation and growth of the consulting and technology practice serving Telecom providers across Europe. Donatello earned his master degree in Electrical and Telecommunication Engineering from Bari Polytechnic and his master of Business Administration from the Massachusetts Institute of Technology. Donatello is passionate about his customers and about technology and how it is changing how we live and work.

Giuseppe L. Cascella received his MSc degree with honours (2001), PhD in Electrical Engineering (2005) from the Politecnico di Bari (Italy), and two EU Marie Curie Fellowships in the Nottingham University (UK) and the University of Malta. He completed his PostDoctoral Research project (2008) “Adaptive Memetic Algorithms for Electric Drives”.
He published 50+ peer-reviewed scientific papers on industry4.0 and AI applications, has been the coordinator of 30+ R&D industrial projects and received the following awards:
• 2018 AWS Activate Builder, “Cloud Data Analytics for Energy Efficiency”, Amazon Web Services, 2018.
• best experiment in Open Call 2, BEinCPPS, H2020, EU funded, project 680633, 2018.
• “SmartSupervisor for Cognitive Energy Efficiency”, A&T award for the best innovative i4.0 solution, 2017.
He is member of IEEE (Institute of Electrical and Electronics Engineers) and of the Italian National Order of Engineers.

Keywords

Machine learning, Deep learning, Industry4.0, IoT, Model identification

Clinician-in-the-loop Artificial Intelligence: for a fairer model of clinical knowledge exploitation

13 January 2021 /

Fabio Massimo Zanzotto

Abstract

Artificial Intelligence is replacing clinicians in some tasks: are you ok with this?

Keywords

Artificial Intelligence, Explainability, Accoutability, Ethics

Artificial Intelligence in MRI: from raw data to analysis

12 January 2021 /

Daniel Remondini

Abstract

AI can provide novel opportunities within MRI along all the path from data generation to final analysis. In this talk we will show some recent applications that our group is performing in different steps of this path: Deep Learning methods to circumvent the curse of dimensionality in MR fingerprinting MR image enhancement through DL super-resolution Age prediction through brain MRI

Short Bio

Prof. Daniel Remondini is Full Professor at the Department of Physics and Astronomy of Bologna. His main interests of research are: Biomedical Data Analysis and modelling with approaches from Physics, Statistics, Bioinformatics and Systems Biology. In particular, he has been developing and applying methods based on the Theory of Networks to complex systems (biological, social), and for the analysis of high-throughput biomedical data (multi-omics at genetic and epigenomic level, biomedical imaging).

Keywords

MR fingerprinting, Super resolution, Predictive models

Clinical Applications of AI in Diagnostic Imaging

12 January 2021 /

Tommaso Banzato

Abstract

This presentation focuses on the Clinical Applications of AI
in Diagnostic Imaging. The main field of application are explored. Namely, the possibility to integrate AI-based systems for optimal patient scheduling, image quality improvement, automated detection of imaging findings, and for the analysis of free text reports. Some use cases are provided.

Short Bio

Dr. Tommaso Banzato is an Assistant Professor at the University of Padua. He obtained his DVM degree at the University of Padua and his PhD degree in 2013 at the same university with a thesis on the normal diagnostic imaging features of snakes, and lizards. After the PHD he had a post-doctoral position until 2018 when he became assistant professor at the same university. Dr. Banzato has authored more than 25 peer-reviewed research publications in diagnostic imaging. He performed an extensive research activity focused on the standardization of different imaging techniques in exotic pets including snakes,lizards, rats, rabbits, and birds. At present time his research is mainly focused on the possible applications of texture analysis and deep learning on diagnostic imaging of companion animals. He was recently funded by the University of Padua for a project aimed to use deep learning to predict the grading of human meningiomas using a translational application of a methodology he has developed in dogs.

Keywords

clinical, diagnostic imaging, MR, CT

Unsupervised deep learning for MR reconstruction using physics-informed cycleGAN

12 January 2021 /

Jong Chul Ye

Abstract

Recently, deep learning approaches for accelerated MRI have been extensively studied thanks to their high performance reconstruction in spite of significantly reduced run- time complexity. These neural networks are usually trained in a supervised manner, so matched pairs of subsampled and fully sampled k-space data are required. Unfortunately, it is often difficult to acquire matched fully sampled k-space data, since the acquisition of fully sampled k-space data requires long scan time and often leads to the change of the acquisition protocol. Therefore, unpaired deep learning without matched label data has become a very important research topic. In this paper, we propose an unpaired deep learning approach using a optimal transport driven cycle-consistent generative adversarial network (OT-cycleGAN) that employs a single pair of generator and discriminator. The proposed OT-cycleGAN architecture is rigorously derived from a dual formulation of the optimal transport formulation using a specially designed penalized least squares cost. The experimental results show that our method can reconstruct high resolution MR images from accelerated k- space data from both single and multiple coil acquisition, without requiring matched reference data.

Shor Bio

Jong Chul Ye is a Professor of the Dept. of Bio/Brain Engineering and Adjunct Professor at Dept. of Mathematical Sciences of Korea Advanced Institute of Science and Technology (KAIST), Korea. He received the B.Sc. and M.Sc. degrees from Seoul National University, Korea, and the Ph.D. from Purdue University, West Lafayette. Before joining KAIST, he was a postdoctoral fellow at the University of Illinois at Urbana Champaign, Senior Member of Research Staff at Philips Research, NY, and Senior Researcher at GE Global Research, Niskayuna, NY. He has served as an associate editor of IEEE Trans. on Image Processing, IEEE Trans. on Computational Imaging, and an editorial board member for Magnetic Resonance in Medicine. He is currently an associate editor for IEEE Trans. on Medical Imaging, and a Senior Editor of IEEE Signal Processing Magazine. He is an IEEE Fellow, Chair of IEEE SPS Computational Imaging TC, and IEEE EMBS Distinguished Lecturer. He was a General Co-chair for 2020 IEEE Symp. On Biomedical Imaging (ISBI) (with Mathews Jacob). His current research focus is deep learning theory and algorithms for various imaging reconstruction problems in x-ray CT, MRI, optics, ultrasound, remote sensing, etc.

Keywords

unsupervised learning, MR reconstruction, cycleGAN

Overcoming the challenges of data paucity in deep learning for neuroimaging

12 January 2021 /

Simeon Spasov

Abstract

Medical imaging datasets usually comprise multi-modal data with high dimensionality and complexity from relatively few subjects. In this talk, I address the challenges of applying deep learning models to such data formats, specifically 1) limiting overfitting and improving performance; 2) improving computational efficiency and model optimization. I propose the use of 3D separable convolutions which decompose the conventional convolution in two steps, hence reducing the number of parameters needed for implementation. I demonstrate that such parameter-efficient model architectures achieve state-of-the-art performance on classification and image reconstruction tasks. In conclusion, I also discuss transfer learning as a promising future research direction which can allow for increasing the availability of training data, as well as reducing the need for potentially expensive or dangerous data collection procedures.

Short Bio

I am a PhD student at the Computer Laboratory at the University of Cambridge. My research interests include applying deep learning on medical images, parameter-efficient deep learning models as well as probabilistic generative models. I have interned at the Montreal Institute for Learning Algorithms as well as Amazon Alexa.

Keywords

neuroimaging, deep learning, parameter-efficiency

Current challenges and future perspectives of machine learning techniques in medical imaging

12 January 2021 /

Stefano Diciotti

Abstract

The application of machine learning techniques is rapidly emerging in Neuroimaging. In this presentation, I will show some important challenges related both to data and algorithms. The quality control and the employment of large (high-quality) datasets are fundamental aspects for training effective machine learning models. Efforts should also be put in the improvement of transparent reporting and reproducibility related to data, code and papers. These are the main requirements for producing high quality machine learning research in the Neuroimaging field. Finally, I will give you a short mention of explainable AI methods that emphasize the development of more interpretable, explainable models – the ultimate goal of big data in Neuroimaging.

Short Bio

Stefano Diciotti was born in Florence (Italy), in 1975. He received the Laurea degree (with honors) in Electronic Engineering from the University of Florence, (Florence), in 2001, and the Ph.D. degree in Bioengineering from the University of Bologna, (Bologna, Italy), in 2005. In 2013, he took the position of Researcher at the University of Bologna, where, in 2019, he became an Associate Professor in Biomedical Engineering. His current research interests include artificial intelligence for health and well-being and medical imaging. Prof. Diciotti is a Senior Member of the IEEE and is affiliated at the Alma Mater Research Institute for Human-centered Artificial Intelligence (Bologna).

Keywords

Big-data, data quality, explainable AI, transparent reporting, reproducibility

Improving Advanced Imaging Workflows with AI

12 January 2021 /

Patrick Bolan

Abstract

This talk addresses the potential for using artificial intelligence methods for improving workflow for advanced MRI methods. These workflows benefit from fast computation and object segmentation inherent in many AI techniques. Applications in magnetic resonance spectroscopy and multi-channel RF transmit systems are presented, and potential areas for further exploration are discussed.

Keywords

Magnetic Resonance Spectroscopy, Parallel Transmit, High Field,

Impact of AI and deep learning on imaging of neurodegenerative diseases

12 January 2021 /

Duygu Tosun-Turgut

Abstract

Biomarkers have become increasingly important to understand the biology of neurodegenerative diseases. We now see a paradigm shift recasting the definition of neurodegenerative disease in living people from syndromal to a biological construct. Effective implementation of such biological constructs though requires widespread availability of biomarkers. This talk will address some of the challenges and AI based advances in neuroimaging-based biomarkers for faster, safer, and smarter operationalization of biomarker-based classification, risk assessment, diagnosis, prognosis, and even prediction of therapy responses in neurodegenerative diseases.

Short Bio

Duygu Tosun-Turgut, PhD, is an Associate Professor of Radiology and Biomedical Imaging at the University of California – San Francisco and Director of the Medical Imaging Informatics and Artificial Intelligence at the San Francisco Veterans Affairs Medical Center. Dr. Tosun obtained her BSc in Electrical and Electronic Engineering from Bilkent University, Turkey in 1999, and she received her MSE in Electrical and Computer Engineering from The Johns Hopkins University, Maryland in 2001. In 2003, she completed her MA in Mathematics from The Johns Hopkins University, and she earned her PhD in Electrical and Computer Engineering from The Johns Hopkins University in 2005, followed by a postdoctoral fellowship in Neurology from the University of California, Los Angeles in 2008. Her research is in the field of neuroimaging in aging, neurodegenerative diseases, and psychiatric disorders and aims to apply advanced imaging technology to identify multi-disciplinary and multi-modality biomarkers to detect the pathophysiological progression of neuropathologies before they cause irreversible damage to the brain. Dr. Tosun-Turgut aims to develop validated imaging markers, potentially providing a means of monitoring the efficacy and regional specificity of drug therapy for neurodegenerative diseases. This will have a broad use in early diagnosis, facilitating initiation of prevention strategies in those at risk, and boost the power of drug therapy trials by selecting those at greatest risk of neurodegenerative diseases. Dr. Tosun-Turgut’s research has been funded by the National Institutes of Health, California Department of Public Health, Department of Defense, Alzheimer’s Association, Michael J Fox Foundation, and through various pharmaceutical collaborations.

Keywords

neuroimaging, Alzheimer, deep learning, multimodal, biomarkers