|
INDEXED BY
 
CONFERENCE TO BE HELD IN
Tokyo, Japan
SPONSORED BY
SUPPORTED BY
|
KEYNOTE
SPEAKER
Prof. Irene Yu-Hua Gu
Chalmers University of Technology, Sweden
Dr. Irene Yu-Hua Gu received Ph.D. degree
in electrical engineering from Eindhoven University of
Technology, Eindhoven, The Netherlands, in 1992. From 1992
to 1996, she was Research Fellow at Philips Research
Institute IPO, Eindhoven, The Netherlands, post dr. at
Staffordshire University, Staffordshire, U.K., and Lecturer
at the University of Birmingham, Birmingham, U.K. Since
1996, she has been with the Department of Electrical
Engineering (previous name: Department of Signals and
Systems), Chalmers University of Technology, Gothenburg,
Sweden, where she has been a Professor since 2004. Her
research interests include statistical image and video
processing, video object tracking and recognition, machine
learning and deep learning, and signal processing with
applications. During the last several years her main
research has been focused on biomedical image analysis and
deep learning. Dr. Gu was an Associate Editor for IEEE
Transactions on Systems, Man, and Cybernetics, Part A:
Systems and Humans, and Part B: Cybernetics from 2000 to
2005, and Associate Editor for EURASIP Journal on Advances
in Signal Processing from 2005 to 2016, Editorial Board of
the Journal of Ambient Intelligence and Smart Environments
from 2011 to 2019. She is a Senior Area Editor for IEEE
Signal Processing Letters since 2021. She was elected as the
Chair of the IEEE Swedish Signal Processing Chapter from
2001 to 2004. She is a senior member of IEEE. She has
coauthored over 200 papers, and has been ranked as the top
50 scientists in the field of Computer Science and
Electronics in Sweden in the 6th Edition of 2020 ranking by
Guide2Research team.
Speech Title: "Brain Tumor
Characteristics in MRIs: Can We Train DL Networks Using
Tumor Bounding Box Images without Requiring Tumor
Annotation?"
Abstract: Deep learning (DL) has drawn
much attention lately for medical image analysis. MR
image-based brain tumor analysis is important for
determining treatment strategies. Such analysis includes
MRI-based tumor molecular subtype classification, tumor
segmentation, and many more. Most existing DL methods
require training by annotated medical data, which puts high
demands on medical experts. It is desirable to seek
alternative approaches for training brain tumors on MRIs
without using (or using less) tumor annotation, while still
maintain high DL performance on the test sets. In this talk,
we present an alternative paradigm for training DL networks,
by using tumor bounding boxes instead of annotated tumors on
MRIs. Despite bounding boxes are widely used for visual
image analysis in computer vision, its usage in medical
images remains an open issue. Two examples are presented,
one is for brain tumor subtype classification, the other for
brain tumor segmentation. Their performance is then compared
with the same network trained on annotated MRIs. These
studies have demonstrated that such paradigm is feasible for
MRI-based classification and segmentation, and is a tradeoff
between significantly reduced (or without) tumor annotation
and a slightly decreased test performance.
INVITED
SPEAKER
Assoc. Prof. Md. Altaf-Ul-Amin
Nara Institute of Science and Technology, Japan
Md. Altaf-Ul-Amin received B.Sc. degree
in Electrical and Electronic Engineering from Bangladesh
University of Engineering and Technology (BUET), Dhaka,
M.Sc. degree in Electrical, Electronic and Systems
Engineering from Universiti Kebangsaan Malaysia (UKM) and
PhD degree from Nara Institute of Science and Technology
(NAIST), Japan. He received the best student paper award in
the IEEE 10th Asian Test Symposium. Also, he received two
other best paper awards as a co-author of journal articles.
He previously worked in several universities in Bangladesh,
Malaysia and Japan. Currently he is working as an associate
professor in Computational Systems Biology Lab of NAIST. He
is conducting research on Network Biology, Systems Biology,
Cheminformatics and Biological Databases. He published
around 60 peer reviewed papers in international journals and
conference proceedings. Google scholar citation index of his
publications is currently more than 6700.
Speech Title: "Applications of KNApSAcK
Database and DPClus Algorithm: Plants to Metabolites to
Target Proteins in the Context of Jamu Medicines and Disease
Pathway Prediction"
Abstract: Initially, we developed
KNApSAcK as a species-metabolite relational database and
subsequently, inspired by its popularity we extended it to
KNApSAcK family databases by adding different types of omics
data together with data regarding edible plants and
traditional medicines mainly focusing human health care and
ecology. Previously we also developed graph clustering
algorithms DPClus and DPClusO, which we and many other
researchers applied to analysis of versatile omics data. In
the present talk, first, I will briefly focus on the
KNApSAcK database and the DPClus algorithm. Then I will
discuss a new method to predict the relation between plant
and disease using network analysis and supervised clustering
based on Jamu formulas. Jamu is the common name of
Indonesian traditional medicines. Next, I will extend the
talk on the analysis for predicting Jamu efficacy based on
metabolite composition and identifying important metabolites
by applying various machine learning techniques and
algorithms including support vector machine (SVM) and random
forest. I will then focus on prediction of target proteins
by Jamu metabolites. Finally, I will discuss application of
DPClusO algorithm in finding inflammatory bowel
disease(IBD), schizophrenia and bipolar disorder related
genes and pathways.
|
