Prof. Daniel Topgaard
Lund University, Sweden

Keynote Speech 1
Prof. Daniel Topgaard
Title: Enhancing diffusion MRI using AI

Abstract:
Diffusion MRI provides information about micrometer-scale structures in the living human brain in apparent contradiction of the millimeter-scale pixel dimensions of the actually recorded images. Despite a large body of literature demonstrating its utility in both research and clinical practice, the method faces numerous challenges that in principle could be overcome with AI. These challenges include excessive time requirements for data acquistion or processing, difficulties in correlating imaging metrics with disease states, and the ill-posedness of translating the imaging results to more fundamental and understandable microstructural properties such as cell densities, sizes, shapes, orientations, and membrane permeabilities. Building on many years of applying basic knowledge from physics and chemistry to increase the information content of the acquired data, this talk will give an overview of our recent and on-going work in using AI to enhance diffusion MRI in the aspects of signal processing, disease state classification, and microstructural interpretation.

Biography:
Daniel Topgaard is a Professor of Physical Chemistry at Lund University, Sweden. Following a Ph.D. degree in Lund 2003 he was a postdoctoral fellow with Alex Pines at UC Berkeley developing NMR spectroscopy for inhomogeneous fields. His current research is focused on solid-state NMR and diffusion MRI methods for investigating structure and dynamics of soft matter systems from lipid membranes to the living human brain. He is a member of the Scientific Council for Natural and Engineering Sciences at the Swedish Research Council, the Board of Trustees of the European Magnetic Resonance (EUROMAR) conference, the Executive Committee of the Experimental NMR Conference (ENC), and the Executive Editorial Board of the Magnetic Resonance journal. He is an Honorary Professor at the Zhongnan Hospital of Wuhan University and Visiting Professor at Beijing University of Posts and Telecommunications.

Prof. Jinsub Park
Hanyang University, Korea

Keynote Speech 2
Jinsub Park*
Department of electronic Engineering, Hanyang University
*Corresponding Author: jinsubpark@hanyang.ac.kr

Title: Artificial Neural Chips and Synaptic Devices for Real-time Unsupervised Learning

Abstract:
Spiking neurons are essential tools for building energy-efficient biomimetic spatiotemporal systems because they communicate with other neurons using sparse and binary signals. However, artificial neurons having a capacitor for emulating integrate function of biological neurons have a limit to achieving high neuronal density. In addition, a low-voltage operation (<1.0 V) is essentially necessary to connect with the modern complementary metal-oixde-semiconductor-field-effect-transistor (C-MOSFET)-based integrated circuits. Here, for the first time, the capacitorless memristive-neural integrated-chip was designed by a conventional 28 nm C-MOSFET process in a foundry with extremely low-voltage operation (<0.7 V) via the rupture dynamics of Ag metallic filaments formed in GeS2 chalcogenide material. For the synaptic devices, we fabricated highly reliable forming-free Cu-doped Oxygenated amorphous carbon(α-C:Ox) resistive memories (CCRMs) with multi-level properties, where resistive switching occurs via a hybrid conducting path consisting of conductive sp² covalent bonds and Cu filaments. Our neuron presented the nonlinear increase in the firing of action potential from 1 to 72 when the amplitude of input-voltage-pulse was linearly increased from 0.5 to 0.7 V like a human sensory system via the memristive-neural integrated-chip. Moreover, a spiking neural network was co-designed by connecting the memristive-neural integrated-chip and a software program via serial communications, and real-time unsupervised learning was performed by a simplified spike-timing-dependent-plasticity learning rule. Finally, the actually hand-written digit image of a smartphone taken from the live webcam was successfully classified through the trained spiking neural network in real-time. In this conference, the details for the resistive switching mechanisms of emerging memory devices and various applications will be discussed.

Key Words: AI-Semiconductor, Neural Chip, Memristive Memory, Synapse device.

Biography:
Professor Jinsub Park is a full professor at Hanyang University in Department of Electronic Engineering and a head of department of Semiconductor Engineering. He received Ph. D. from Tohoku University, Sendai, Japan in Applied Physics and Engineering. Prof. Park has authored and co-authored more than 150 high-impact journal papers including Adv. Func. Matter, Nano letters, Nanoscale Horizons, Composite B: Engineering, Nano research. Currently, he is serving as the senior vice president of the Korean Semiconductor Display Technology Association. He is a recipient of deputy prime minister award and industry minister award in 2020 and 2024, respectively. His group is developing advancement of semiconductor technologies including Emerging memory (SOM, STT MRAM, ReRAM) devices, Energy harvesting, Thermal Management and optoelectronics.