| Popis: |
Understanding the role of neurotransmitter dopamine in brain function under normal or pathological states is one of the most active areas of research in neurosciences. Failures in dopamine neurotransmission affects tremendous amount of brain abilities including movement, mental, and motivation and reward systems of the brain. Ability to measure phasic release of dopamine in specific locations of the brain will lead to a powerful tool for the neuroscientists, However, the tremendous amount of image-formed data as produced from different locations of the brain makes the manual analysis of these data cumbersome. Luckily, image processing techniques will help in solving these problems effortlessly to ease and speed the analysis for neuro-physicians. In this paper, we propose a deep-learning based identification scheme to identify the release case of phasic dopamine by examining the dopamine analysis (DA) imaging attributes using a convolutional neural network (CNN). More precisely, the proposed scheme exploits the transfer learning based DarkNet-19 network to train and identify the phasic dopamine release-2019 (PDR19) dataset into two-classes; namely, “release images,” or “non-release images” The experimental outcomes demonstrated the distinction of our identification scheme, recording an identification accuracy of 99.1% with a cross entropy loss of 0.022 attained after 25 epochs each with 100 iterations (i.e., 2500 iterations) for the 2-class classifier. Besides, our identification scheme was assessed using many other assessment factors, such as the identification precision percentage (IPP), the identification sensitivity percentage (ISnP), the identification specificity percentage (ISpP), and the identification weighted average percentage (F1P). Consequently, the performance of the proposed scheme surpassed several existing dopamine identification schemes. |
