Fig 1 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Automated lateralization of hippocampal sclerosis. (a). In the training phase, an optimal region of interest is defined for each modality to systematically sample features (T1-derived volume, T2-weighted intensity, and FLAIR/T1 intensity) across individuals. To this purpose, in each patient paired t-tests compare corresponding vertices of the left and right subfields, z-scored with respect to healthy controls. The resulting group-level asymmetry t-map is then thresholded from 0 to the highest value and binarized; for each threshold, the binarized t-map is overlaid on the asymmetry map of each individual to compute the average across subfields. Then a linear discriminant classifier is trained for each threshold, and the model yielding the highest lateralization accuracy (in this example LDA model 3) is used to test the classifier. (b). Lateralization prediction in a patient with MRI-negative left TLE. Coronal sections are shown together with the automatically generated asymmetry maps for columnar volume, T2-weighted, and FLAIR/T1 intensities. On each map, dotted line corresponds to the level of the coronal MRI section and the optimal ROI obtained during training is outlined in black
Source publication
Epilepsy is a prevalent chronic condition affecting about 50 million people worldwide. A third of patients suffer from seizures unresponsive to medication. Uncontrolled seizures damage the brain, are associated with cognitive decline, and have negative impact on well-being. For these patients, the surgical resection of the brain region that gives r...
Context in source publication
Context 1
... verification or long-term measures of seizure outcome after surgery; moreover, absence of validation in independent datasets has precluded assessment of generalizability. To tackle these shortcomings, our group recently designed an automated surface-based linear discriminant classifier trained on T1-and FLAIR-derived laminar features of HS ( Fig. 1) [44]. As HS is typically characterized by T1-weighted hypointensity and T2-weighted hyperintensity, the synthetic contrast FLAIR/T1 maximized their combined contributions to detect the full pathology spectrum. The classifier accurately lateralized the focus in 85% of patients with MRI-negative but histologically verified HS. Notably, ...
Citations
... We attempted to minimize the number of diseases and the difference in average age and male-female ratios among the remaining subject groups. Their file numbers and detailed medical information can be found in the Appendix of [28]. ...
... This was done for the purpose of acquiring resting-state EEG data. An index of which 50 s was used for which subject can be found in the Appendix of [28]. The preprocessing was kept at a minimum because we observed (by performing some preliminary tests) that Granger causality connectivity derivations are very sensitive to EEG filter operations. ...
... Both EEG preprocessing and Granger causality calculation were performed in Brainstorm version 3.200124 [29]. A detailed description of these steps can be found in Appendix G of the master thesis written based on this study [28]. This preprocessing method was applied to the EEG data of all 60 subjects, and was used for all deep learning experiments in this study. ...
Multisensory integration refers to the integration of multiple senses by the nervous system. Auditory andtactile features are closely related senses as can be understood from the fact that adjectives such as soft,rough, and warm are used commonly for auditory and tactile features. Previous studies show that auditorycues play an important role to assess the roughness of a surface. Different characteristics of auditory cuessuch as amplitude and frequency may cause perceiving surface rougher or smoother. In this study, weinvestigate the effects of harmonic and inharmonic sounds on roughness perception to examine whetherauditory roughness will affect the tactile roughness perception while they are presented simultaneously.We expected the participants to perceive surfaces rougher while they listen to inharmonic sounds due toauditory roughness. We presented simultaneous and sequential harmonic and inharmonic sounds withthree sandpapers with different roughness levels (P100, P120, P 150 grit numbers) to the participants. Wefound that participants perceive sandpaper with the P120 grit number rougher while they listen tosimultaneous inharmonic sounds than simultaneous harmonic sounds. However, any effect of harmonicityon the sandpapers with P100 and P150 grit numbers was not observed. We suggest that auditoryroughness may enhance tactile roughness perception for surfaces with particular roughness levels,possibly when the roughness estimation from the tactile sense remains ambiguous.
... This was done for the purpose of acquiring resting-state EEG data. An index of which 50 seconds were used for which subject can be found in the appendix of (Rijnders, 2021). ...
... For each of the 60 subjects, and for each electrode combination, four GC matrices were calculated, one for each frequency band (delta = 1-4 Hz, theta = 5-7 Hz, alpha = 8-13 Hz, beta = 14-29 Hz, gamma = 30-55 Hz). A detailed description of how these operations were performed and configured in Brainstorm is presented in (Rijnders, 2021). ...
... ; https://doi.org/10.1101/2021.09.24.21264004 doi: medRxiv preprint be utilized effectively as the neuromarkers for epilepsy. In the original study (Rijnders, 2021) the CNN architecture consisted of not one but two convolution layers, which employed smaller kernel sizes and stride. This was however computationally more complex, while resulting in similar performance as the simple architecture of the current study, therefore the architecture was simplified. ...
Objective: This study investigates the performance of a CNN algorithm on epilepsy diagnosis. Without pathology, diagnosis involves long and costly electroencephalographic (EEG) monitoring. Novel approaches may overcome this by comparing brain connectivity using graph metrics. This study, however, uses deep learning to learn connectivity patterns directly from easily acquired EEG data.
Approach: A convolutional neural network (CNN) algorithm was applied on directed Granger causality (GC) connectivity measures, derived from 50 seconds of resting-state surface EEG recordings from 30 subjects with epilepsy and a 30 subject control group.
Main results: The learned CNN filters reflected reduced delta band connectivity in frontal regions and increased left lateralized frontal-posterior gamma band connectivity. A diagnosis accuracy of 85% (F1-score 85%) was achieved by an ensemble of CNN models, each trained on differently prepared data from different electrode combinations.
Conclusions: Appropriate preparation of connectivity data enables generic CNN algorithms to be used for detection of multiple discriminative epileptic features. Differential patterns revealed in this study may help to shed light on underlying altered cognitive abilities in epilepsy patients.
Significance: The accuracy achieved in this study shows that, in combination with other methods, this approach could prove a valuable clinical decision support system for epilepsy diagnosis.
