Keynote speakers BHI-BSN-2022

Building trustworthy AI systems with reliable components
ABSTRACT: AI medical systems, designed to support diagnosis, therapy planning and monitoring, have a long history, but recently they received a renewed strong attention due to the advancements in machine and deep learning and to the large and increasing availability of digital data. The need of protecting citizens, providing them with safeguards against misuse of AI approaches, and in particular of data-driven technologies, has pushed towards the implementation of “trustworthy” AI systems, lawful, ethical and robust. This talk will discuss how components, based on reliability principles, may provide the basis for the design and implementation of successful AI solutions. Finally, the talk will advocate that only a proper socio-technical approach will eventually provide trustworthy systems.

Current and Future of AI in Clinical Medicine: Lessons Learned this Decade and Future Paradigm for Clinical Impact

ABSTRACT: Artificial intelligence has gradually been introduced and adopted in the clinical medicine realm, but the clinical impact has not been as big as it was hoped. Among the issues include: lack of continual clinician/data scientist synergy, inadequate data and IT infrastructure, failure to maintain AI models that have generalizability, inadequate education for clinicians and administrators, and too little appreciation for the complexities of clinical medicine and decision making. The future of clinical medicine needs to have much stronger clinician involvement and direction so the level of cognition will be much higher. This future paradigm will need to involve AI technologies such as reinforcement learning and digital twins.

Near-human Sensing in Fabric Smart Space

ABSTRACT: In future network, the provisioning of ultra-low latency, non-intrusive and immersive service experience creates various challenges, among which near-human sensing is of great importance to obtain multi-modal information without disturbing user. This talk introduces the development of various functional fabrics, which have provided new thoughts for generating novel near-human services interconnected by fabric sensors, body area network, edge cloud and visualization system. In order to embrace digital intelligent world, this talk also presents the fabric smart space empowered by intelligent fabric agents, which gather multidimensional sensory data and interactive information via near-human sensing technologies. Finally, several examples with the use of fabric smart space are given in terms of sport, healthcare and medical scenarios.

Designing Digital Tools for ADRDs that Double as Assessments and Interventions

ABSTRACT: The world’s population is aging, and the increasing number of older adults with Alzheimer’s disease and related dementias (ADRDs) is a challenge our society must address. New modes of technology offer unprecedented opportunities to address some of the needs that accompany cognitive decline by providing automated health assessment and memory interventions. In this work, we create EMMA, a memory management app, that combines the two capabilities. Through participatory design with older adults and caregivers, we design an app that is accessible and effective as a compensatory aid for older adults with memory decline. By collecting data from app usage in combination with sensor data, we extract digital markers that predict multiple clinical measures. We evaluate this app using data from 14 participants with mild cognitive impairment. We observed moderate to large correlations between predicted and ground-truth assessment scores for each clinical assessment.

Indoor Radar Sensing of Elderly People : Overcoming the Barriers in Home Care Technology

ABSTRACT: The use of new technologies in monitoring of elderly people has significantly grown in the last years. Although indoor radar monitoring is still in its early stage of development, it carries great potential to be one of the leading technologies in the future. In this presentation, the basics and the advantages of indoor radar sensing, compared to other existing technologies as cameras and wearable devices, will be provided and discussed.

Harnessing ultrasound for modulation of the central and peripheral nervous system

ABSTRACT: Focused ultrasound (FUS) neuromodulation has previously been proposed as a promising technique to drive neuronal activity and has been shown throughout a breadth of applications including in mice, rats, non-human primates and humans as a novel technique for the noninvasive manipulation of neuronal activity using ultrasound. Our group and others have demonstrated excitation of both the central (CNS) and peripheral nervous system (PNS). In the CNS, motor- and cognitive-related brain regions of mice were induced by targeting specific brain structures. Higher acoustic pressures increased the success rate. Pupil dilation was observed when neuromodulating regions in the brain covering the superior colliculus and other anxiety-related structures such as hippocampus and locus coeruleus. In the PNS, we showed for the first time stimulation of the sciatic nerve with FUS eliciting a physiological motor response was recorded in vivo. Clipping the sciatic nerve downstream of stimulation eliminated EMG activity during FUS stimulation. Peak-to-peak EMG responses and latencies were comparable to conventional electrical stimulation methods. Histology along with behavioral and thermal testing did not indicate damage to the nerve or surrounding regions. Finally, underlying mechanisms on the Piezo2 channel and clinical studies on pain mitigation will be shown. Our studies demonstrate the capability of FUS to modulate target specific regions in both the brain and the periphery with several potential clinical applications.

Bioelectronic tools to study the gut-brain axis.

ABSTRACT: Polymeric electroactive materials and devices can bridge the gap between hard inflexible materials used for physical transducers and soft, compliant biological tissues. An additional advantage of these electronic materials is their flexibility for processing and fabrication in a wide range of formats. In this presentation, I will discuss our recent progress generating 3D conducting polymer devices, to simultaneously host and monitor complex multi-cellular models of tissues and organs. Electrophysiological recording of parameters such as tissue impedance, epithelial and endothelial barrier tissue integrity and neuronal activity, are all made possible thanks to the conducting polymer devices and are validated with traditional biological readouts such as immunofluorescence or cytokine analysis. Building on our previous work that showcased a bioelectronic model of the human intestine, we are now incorporating elements of the microbiome and the immune system as well as the enteric nervous system. Coupling this model with our model of the neuro-vascular unit (including blood brain barrier) currently in progress, will bring us to our goal of a physiologically representative in vitro model of the gut-brain-microbiome axis. Alongside our in vitro work, I will show how our recent work on developing electronic probes to study the enteric nervous system. Transitioning from in vitro human and rat to in vivo rat models allows us to integrate electrophysiological recordings of neuronal activity with tissue impedance to really begin to unravel gut-brain axis signaling.

AI-enabled Sensing and Interventions for Global Health.

ABSTRACT: The challenges and opportunities to improve the global health cycle are at critical inflection points under the strain of a world-wide pandemic, international conflict, and large-scale environmental disasters. AI-enabled sensing, decision support, and actions can leverage the enormous data generated and consumed through the global health steps of monitoring, diagnosis, intervention, training, prevention, and informing the public. This presentation will discuss how body sensor networks and health informatics platforms can work in concert with population-level and environmental sensing to assess health threat phenomenology, exposure dosimetry, medical intervention efficacy. These advances can be used to scale interventions, guide health and emergency response policy, enhance training of healthcare providers and first responders, and more effectively engage the public.

Deep Learning based Medical Image Reconstruction

ABSTRACT: This talk will introduce various deep learning methods we developed for fast MR acquisition, low-dose CT reconstruction, and low-cost and low-dose PET acquisition. The implementation of these techniques in scanners for real clinical applications will be demonstrated. Also, comparisons with state-of-the-art acquisition methods will be discussed.