| Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 100 === This thesis describes a study exploiting multi-focal functional MRI(fMRI) for retinotopic mapping, or retinotopy, in the primary visual cortex. We tried to reconstruct visual image according the retinotopy and brain activities obtained by fMRI. Multi-focal method divides the visual field into several blocks and each block has its own paradigm for the visual experiment. Using this method, researchers show that they are able to distinguish the brain areas corresponding to each block simultaneously. Despite visual fMRI, this method is also applied electrophysiological analysis of visual system.
In this study, we performed a visual fMRI experiment using a specific pattern after multi-focal retinotopy. We then attempt to reconstruct the visual image by combining the results of visual fMRI and retinotopy. The study applied general linear model to analyze the fMRI signal and produced a t value to justify the existence of stimuli-related brain activities. However, judging the “existence” required selecting a threshold of the t value. We empirically found that the accuracy of the reconstructed visual image largely depended on the threshold selection. Therefore, this study proposed an approach to find the optimal t threshold according to a receiver operating characteristic analysis. The results obtained with 5 volunteers using the optimized t thresholds demonstrated an average accuracy of 80%.
In conclusion, we successfully reconstructed the visual image by the fMRI technique. Compared to previous investigations, we regard the contributions of this thesis are the optimization method for visual image reconstruction. This method leads to a completely automatic reconstruction procedure and takes visual reconstruction a step forward.
|
|---|
