Beyond adsorption: alternative approaches to surface functionalization with biomolecules
Active surfaces functionalized with proteins and enzymes are a prerequisite for the development of diagnostic devices and biocompatible materials. Often, the immobilization of biomolecules on a surface is carried out by approaches that cannot be easily controlled such as adsorption and chemical crosslinking. Although effective, these approaches often lead to loss of functionality of the biomolecule. Alternative approaches based on enzymatic processes or affinity and aiming at the site-specific and controllable immobilization of proteins on surfaces will be presented.
Presentation by Prof. Théo Lasser, EPFL
Seeing is believing
This talk is an invitation for a promenade into tissue structure and function and cell and subcellular organelles (see image aside). Relying on coherent imaging techniques we will try to see “diabetes”, to look into the brain for Alzheimer disease and we will finish our walk with novel insights on the cellular level based on SOFI, providing 3D and even 4D superresolved images of living cells.
We will try to present the underlying optical concepts, and conclude with an outlook for imaging with applications in medicine and lifesciences.
Presentation by PD Dr. Stavroula Mougiakakou
Machine Learning and Personalized Diabetes Management
The recent advances in the area of machine learning allowed the introduction of systems and methods for enhanced self-management of chronic diseases. Within this framework the on-going research activities of the Diabetes Technology Research Group at the University of Bern towards personalized insulin treatment and diet assessment, will be presented. How can adaptive learning, real-time data-driven modelling, computer vision, control theory and mHealth be synergistically used to tackle open issues related to diabetes self-management? Our real life experience will be shared.
Presentation by Agamennon Krasoulis
Reconstruction of finger movement from sEMG and accelerometry
In this work, we investigate finger kinematics reconstruction during two distinct sets of movements, including isometric and isotonic hand configurations, as well as functional movements and grasping of common household objects (NINAPRO database). By using sEMG and accelerometry signals, we demonstrate that continuous decoding of the position of a large number of joints is feasible. In addition, we provide evidence that finger joint kinematics decoders can generalise reasonably well on novel movements, a highly-desirable feature, currently unavailable with traditional classification methods.
Privacy protection techniques for clinical profiles
Prof. Dr. Josep Pegueroles, UPC - Barcelona Tech
New perspectives for mHealth at the point-of-care
Presensentation by Prof. Walter Karlen ETHZ
The Mobile Health Systems Lab (MHSL) at ETH Zurich is engaged in various initiatives to improve reliability of mHealth tools. In this talk, I will discuss how signal processing, algorithms and machine-user interaction can drive quality assurance in mHealth systems and facilitate complex vital sign measurements at the point-of-care.
Leveraging Medical Big Data for Improved Clinical Decision Making
In the era of Big Data, with the explosion in medical knowledge, staying abreast of the latest developments is challenging and potentially untenable. Failure of all practitioners to possess similar knowledge is evidenced by variation in practice—major impediment to quality and cost-effective practice. Decision making in oncology is particularly challenging; only approximately 5% of patients seen in practice have disease characteristics for which established clinical trial evidence would prescribe a recommended practice. An emerging paradigm of “rapid learning” is emerging, in which evidence based on medical practice, which better reflects the broad spectrum of varying patient characteristics, can potentially be used to guide care. The concept is to leverage historical medical record to find similar patient cohorts matching the characteristics of a particular patient, and review the treatments and outcomes to determine those that produced the best outcomes. However, effectively leveraging image data is difficult because images are unstructured data. In this presentation we will discuss our approaches of extracting and computing with quantitative information in images and leveraging it for enabling better clinical decision making in the data-driven, rapid learning scenario.
Biomedical data fusion for cancer data: applications linking imaging with molecular data
Presensentation by Prof. Olivier Gevaert (University of Stanford , USA)
Vast amounts of molecular data characterizing the genome, epi-genome and transcriptome are becoming available for a wide range of cancers. In addition, new computational tools for quantitatively analyzing medical and pathological images are creating new types of phenotypic data. Now we have the opportunity to integrate the data at molecular, cellular and tissue scale to create a more comprehensive view of key biological processes underlying cancer. Moreover, this integration can have profound contributions toward predicting diagnosis and treatment. I will discuss current work in progress to tackle these challenges in biomedical data fusion and show examples for non-small cell lung cancer and glioblastoma.
Notre expérience de tech transfert dans le domaine Medtech
Presentation by Prof. Dr. Philippe Renaud (EPFL)
Computational neuroimaging to better understand the restless brain
Presentation by Dimitri Van de Ville
Functional connectivity (FC) as extracted from fMRI data has become a key measure to study the brain's functional integration. FC is obtained as the correlation between timecourses of different voxels or brain regions, conventionally computed for the full duration of the resting-state scanning session. Recently, temporal variations in FC have been observed using sliding-window correlation, for which the term dynamic functional connectivity (dynFC) has been coined. Despite the risk of spurious fluctuations due to short window length, we reported meaningful dynFC that reflected complementary information w.r.t. standard FC. We recently proposed a multivariate framework to facilitate the analysis and interpretation of dynFC in terms of FC building blocks. We also showed clinical relevance of dynFC; e.g., resting-state fluctuations of FC allow to discriminate healthy controls from minimally-disabled MS patients. We also have showed that large-scale fMRI networks can be related to EEG microstate sequences that are 3 orders of magnitude faster due to the latter’s scale-free organization, which provides another motivation to further look into non-stationary features of brain activity.
Ultra-Low Power Wireless Body Sensor Nodes Design for Smart Bio-Signals Monitoring
Présentation du Prof. David Atienza Alonso (EPFL)