Random Subwindows for Robust Image Classification
CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Hi-index | 0.00 |
Early and accurate prediction of tissue outcome is essential to the clinical decision-making process in acute ischemic stroke. We present a quantitative predictive model that combines tissue information available immediately after onset, measured using fluid attenuated inversion recovery (FLAIR), with multimodal perfusion features (Tmax, MTT, and TTP) to infer the likely outcome of the tissue. A key component is the use of randomly extracted, overlapping, cuboids (i.e. rectangular volumes) whose size is automatically determined during learning. The prediction problem is formalized into a nonlinear spectral regression framework where the inputs are the local, multi-modal cuboids extracted from FLAIR and perfusion images at onset, and where the output is the local FLAIR intensity of the tissue 4 days after intervention. Experiments on 7 stroke patients demonstrate the effectiveness of our approach in predicting tissue fate and its superiority to linear models that are conventionally used.