2021 Fiscal Year Annual Research Report
Development of Data-driven Prediction Model using 3D Multimodal Deep Neural Networks for Estimating the Evolution of White Matter Hyperintensities Associated with Small Vessel Disease in Brain MRI
| Project/Area Number |
20K23356
|
|---|
| Research Institution | Institute of Physical and Chemical Research |
|---|
Principal Investigator |
Rachmadi Muhammad 国立研究開発法人理化学研究所, 脳神経科学研究センター, 基礎科学特別研究員 (60874881)
|
|---|
| Project Period (FY) |
2020-09-11 – 2022-03-31
|
| Keywords | White matter lesions / Progression of WMHs / Disease prediction model / Deep learning / WMHs / Dementia / Human brain MRI |
|---|
| Outline of Annual Research Achievements |
White matter hyperintensities (WMHs) and their evolution over time are the focus of this research. WMHs are neuroradiological features seen in T2-FLAIR brain MRI and associated with stroke and dementia. Clinical studies indicate that the volume of WMHs on a patient may decrease (i.e., regress), stay the same, or increase (i.e., progress) over time.
In this fiscal year, we successfully developed a more accurate predictive model for WMHs evolution by performing joint prediction of WMHs evolution and stroke lesions segmentation. Furthermore, auxiliary input of stroke lesions probability maps also improved the performance of our model. These findings are important because (1) they confirmed previous clinical studies which elucidated that is as strong correlation between WMHs evolution and stroke lesions and (2) more accurate prediction of WMHs evolution can help physicians to create patient specific treatment for dementia patient.
In conclusion, this research project has successfully developed an accurate deep learning model for predicting WMHs evolution by combining (1) probabilistic model of deep learning for modeling spatial uncertainty in the prediction of WMHs evolution, (2) volume interval estimation for better interpretation of predictive model’s results, and (3) stroke lesions information in the form of joint prediction of WMHs evolution and stroke lesions segmentation and stroke lesions probability maps.
|
| Remarks |
The latest resulst are in the middle of writing for journal publication this year (2022).
|
