3Dtexture analysis
Principal Investigator
Name
Henning Muller
Degrees
PhD
Institution
HES-SO
Position Title
Professor
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-192
Initial CDAS Request Approval
Jan 25, 2016
Title
3Dtexture analysis
Summary
We would like to use the NLST data to test such 3D texture analysis algorithms for the detection of lung anomalies. We have learned combinations of Riesz wavelets and used other visual 3D features for characterising lung tissue in the past and we would like to use it on the NLST data as well to predict, for example lung cancer but also other characteristics, such as age of a person or gender that could be linked with specific texture characteristics. As it is a large data set it might also be possible to use deep learning on the data which we would like to compare with our learnt wavelet representation to characterise the lung tissue.
As the data set also contains clinical data and histopathology images there is a possibility to predict specific conditions in the clinical data based on the visual information and also use learning algorithms on multimodal data, so structure clinical data and visual characteristics. We hope that this can improve the quality of our analysis.
As the data set also contains clinical data and histopathology images there is a possibility to predict specific conditions in the clinical data based on the visual information and also use learning algorithms on multimodal data, so structure clinical data and visual characteristics. We hope that this can improve the quality of our analysis.
Aims
The main goals are to use a large data set to improve our texture analysis algorithms for lung image data.
Such a good characterisation can hopefully help us to improve predicting lung anomalies with higher quality. Also the link between visual appearance and clinical parameters can potentially help us to interpret the data and the influence, for example that age has on the lung tissue.
The objectives are thus manifold, ranging from technical objectives to goals that can improve clinical decision making.
Collaborators
Adrien Depeursinge
Yashin Dixente