My main research are
predicting the disease (biomarkers associated with the disease) with medical images
using deep learning / machine learning.
statistical analysis for genome-wide association studies with medical images.
Embedded Files
Research Experience
Deep learning / Machine learning with medical images
2022.12 ~ Present Multi-stream deep learning model for predicting early stage of Alzheimer's Disease with Rey Complex Figure Test images
Predict MCI from CN subjects with RCFT images
Our model consists of dual networks (spatial stream / scoring stream).
scoring network is the automated scoring system previously develped.
External validation (AUC : 0.82 / accuracy : 74%)
2021.5 ~ 2022.8 Deep learning model development for automating Rey Complex Figure Test scoring.
obtained 20,040 scanned images and scores rated by 32 experienced psychologists.
trained the automated scoring system using the DenseNet architecture.
r : 0.99 / MAE : 0.95 (point)
2020.2 ~ 2022.5 Deep learning model for predicting amyloid-beta accumulation with MRI and genetics for mild cognitive impairment.
T1-MRI, Flair and Apoe4 genotype were utilzed as the input.
Cross Vision Transformer (ViT) model was adopted.
AUC 0.84 / sensitivity 73% / specificity 83%
approved Clinical Test Phase III from Ministry of Food and Drug Safety (NCT05383053).
2021.10 ~ 2022.12 Machine learning model for predicting conversion with brain age prediction
constructed a brain age prediction model using the brain volume and cortical thickness features from CN (r = 0.803, MAE = 3.25 years.)
With this model (brain age delta), conversions from CN / MCI were predicted
MCI conversion AUC : 0.65 / AD conversion AUC : 0.76
2019.1 ~ 2019.12 Deep learning model development for predicting ovarian cancer regions with whole slide images
generated patches from ovarian cancer whole slide images.
predicted path-based cancer regions with ResNet classification model.
AUC : 0.98 / ACC : 94%
Genetic association studies with medical images
2023.1 ~ Present Genome-wide association studies of visual memory function with longitudinal Rey Complex Figures Tests
conducted Linear mixed models to identify genetic variation associated with the decline of memory function across time with longitudinal data
validated significant SNPs with RNA gene count analysis and survival analysis.
found new candiated SNPs associated with Alzheimer's Disease.
2019.10 ~ 2023.1 Machine learning-based statistical methods for Alzheimer's Disease GWAS with T1-MRI using weightd method
assumed that the disease status for MCI and phenotype-unknown subjectsis missing.
developed the AD prediction model with AD and CN, and classified missing group into AD or CN group.
performed case/control GWAS including missing group using weighted generalized linear equation.
2019.3 ~ 2021.8 Genome-wide association studies of brain imaging phenotypes with T1-MRI
conducted genome-wide association studies for brain magnetic resonance imaging measures of hippocampal volume and entorhinal cortical thickness.
a missense variant, rs77359862 in the SHARPIN gene was associated with entorhinal cortical thickness and hippocampal volume.
the variant significantly reduced the binding of linear ubiquitination assembly complex proteins, SHPARIN and HOIP, altering the downstream NF-κB signaling pathway.
Page updated
Google Sites
Report abuse