هادي اسفندياري

Basic information about human brain conditions can be decoded through various non-invasive assessment techniques such as functional magnetic resonance imaging (fMRI). Brain decoding is a function that takes brain activity as input and brain states as output, and its function is eva‎luated by how it approximates the real relationship between the two natures, thus falling into the category of machine learning. It is often studied in a specific context of supervised learning. Exploring the neural activity of the human brain can reveal a person's thoughts as they perceive, think, or visualize. In recent years, deep learning methods based on deep neural networks have been successfully used in the analysis of neural imaging data. A convolutional neural network, which is a type of deep neural network, uses a convolutional core to cover a local area of the input sample and travels across the sample to provide a feature map for subsequent layers. Decoded motor modes can be used to provide neurofeedback to the individual. On the other hand, the ability to adapt the stimulus during the fMRI test provides a new level of flexibility and has potential applications in a variety of fields, including performance improvement, rehabilitation, and treatment. In this study, we propose a model based on convolutional neural networks that simultaneously uses the spatial and temporal properties of fMRI data to classify the types of motor tasks of subjects to identify the relationship between human behavior and activity in different areas of the brain. We also use class activation maps to analyze the trained model and describe the important brain regions that were involved in the classification.

تصويربرداري تشديد مغناطيسي كاركردي , رمزگشايي , يادگيري عميق , شبكه‌هاي عصبي پيچشي , نقشه‌هاي فعال‌سازي كلاسي

Functional Magnetic Resonance Imaging (fMRI) , decoding , Deep Learning , Convolutional Neural Networks , Class Activation Maps

Hadi Esfandyari

Dr. Daliri