| 제목 |
Predictability of AF and ILF on Language Recovery in Aphasia after Stroke |
| 소속 |
Korea University College of Medicine, Department of Physical Medicine and Rehabilitation1, Korea University, Brain convergence research center2, Korea University, Department of Biomedical Sciences3 |
| 저자 |
Jun Soo Noh1*, Yoonhye Na3, Sekwang Lee3, Minjae Cho3, Yu Mi Hwang2, Woo-Suk Tae2, Sung-Bom Pyun1,2† |
Objectives
This study aims to investigate the predictability of 6-month language function in patients with aphasia (PWA) using parameters of diffusion tensor tractography (DTT) of arcuate fasciculus (AF) and inferior longitudinal fasciculus (ILF) utilizing machine learning classification.
Methods
We collected data of PWA after stroke from the hospital. Thirty-five stroke patients were included for analysis who had 1) left hemispheric stroke, 2) performed diffusion tensor image (DTI), 3) evaluated language function after onset and 6 months poststroke using Korean version of western aphasia battery (K-WAB). We classified the language outcome at 6 months of PWA either favorable or poor recovery using three cutoff values of aphasia quotient (AQ); 1) 61.6 points (50 percentile), 2) 75.0 points (3rd quartile), 3) 80.3 points (cutoff value of normal in K-WAB). AF and ILF was reconstructed by DTIstudio and ratio of fractional anisotropy (FA) between two hemispheres (DTT index) was calculated in two tracts. Group classification was conducted with machine learning method and the input value of initial FA of AF and ILF separately or both tracts utilizing support vector machine (SVM) algorithm. And the results were compared with the clinical classification based on three cutoff values.
Results
All the results showed that classification accuracy decreased as higher cut-off value was adopted. When inputting values of AF only, classification accuracy was 74.29%, 60.00%, and 57.14% in cut-off value of 61.9, 75.0 and 80.3 points, respectively. However, the classification accuracy increased when FA value of both AF and ILF was inputted; 88.57%, 77.14% and 74.29% and it showed same result when input value was adopted from ILF only (Table 1).
Conclusion
Although AF is a well-known major language-related tract, these results indicated that ILF may have a significant role in language recovery in aphasia after stroke. |
|
Classification Accuracy by Input and Cut-off Values
|
