Page 18
Notes:
allied
academies
Journal of Dermatology Research and Skin Care | Volume 2
May 14-15, 2018 | Montreal, Canada
Spring Dermatology &
Skin Care Expo Conference
A
pproximately 5.4 million new skin cancer cases are
diagnosed in United States every year. It was reported
that the survival rate increases from 14% to 97% with
early treatment hence placing high importance in the early
detection of this disease. We will, therefore, introduce a
computer-aided-diagnosis (CAD) algorithm for differential
diagnosis of pigmented lesions in HGM images by which the
pathological structures described by the dermatologist in
National Taiwan University’s Hospital are used to quantify
healthiness, age, distinguish gender, etc. of the subject in
addition to assisting the enhancement of the efficiency
and inter/intra consistencies during HGM image reading.
These corresponding knowledges of human experts were
also transferred to the Artificial Intelligence (AI) machine in
characterizing the feature layers of the Deep Convolution
Neural Network (CNN). With limited data of approximately
2000 HGM images based on transfer learning, detection of
basal cell carcinoma (BCC) was achieved with a recognition
rate of 97.3% accuracy, 98.7% sensitivity and 95.9%
specificity.
Speaker Biography
Chris Gwo Giun Lee is an investigator in the field of data science including
bioinformatics and multimedia. He has previously held leading and managerial
positions in the industry such as System Architect in former Philips Semiconductor in
Silicon Valley. He has joined NCKU in 2003 where he found and is currently directing the
Bioinfotronics Research Center (BITS). He conducts highly multidisciplinary research
having collaborations with IBM TJ Watson Research Center on cloud computing; Inform
Genomics Inc., found by experts from MD Anderson Cancer Research Center, Harvard
Medical School, and MIT, on analytics architecture for precision medicine; Banner
Health Research on intelligent health cloud for Alzheimer disease; National Taiwan
University on harmonically generated microscopy medical image processing, etc.
e:
clee@mail.ncku.edu.twSkin cancer detection via deep analytics of harmonically generated microscopy (HGM) images
Chris Gwo Giun Lee
Bioinfotronics Research Center, Taiwan