采用深度学习法检测6-OHDA帕金森病大鼠模型黑质多巴胺能神经元

Abstract

目的:采用基于深度学习的目标检测算法对6-羟基多巴(6-hydroxydopamine,6-OHDA)单侧定点注射导致的帕金森病(Parkinson's disease,PD)大鼠模型黑质多巴胺能神经元进行计数。方法:采用6-OHDA右侧黑质部位定点注射导致的PD大鼠模型,选择阿扑吗啡诱导旋转的方法进行行为学检测;用酪氨酸羟化酶(TH)免疫组化、胶质纤维酸性蛋白(GFAP)免疫组化、α-synuclein免疫组化分别进行染色,观察大鼠黑质部位的病理学改变;根据TH免疫组化染色结果,用深度学习的方法对TH阳性的多巴胺能神经元计数算法进行初探。结果:PD大鼠模型建立后,行为学检测筛选出成功的模型,并对黑质损伤情况有初步的判断; 3种免疫组化的结果从病理学方面证明了模型的成功。采用深度学习的方法对TH免疫组化中的阳性神经元进行计数,得到关于神经元计数的初步算法,其准确率和召回率均处于较高水平。结论:在右侧黑质部位定点注射6-OHDA的大鼠模型上,用深度学习的方法对黑质多巴胺能神经元计数进行探究,得到了具有实用意义的初步算法。

Objective: To count the dopaminergic neurons in the substantia nigra of Parkinson's disease(PD) rat model induced by unilateral injection of 6-hydroxydopamine(6-OHDA) with detection algorithm based on deep learning. Methods: The PD rat model was induced by unilateral injection of 6-OHDA in right substantia nigra,and atropin-induced rotation was used for behavioral test; tyrosine hydroxylase(TH) immunohistochemistry,glial fibrillary acidic protein(GFAP) immunohistochemistry and α-synuclein immunohistochemistry were performed to observe the pathological changes in the substantia nigra of rats. According to the results of TH immunohistochemical staining,the algorithm of TH positive dopaminergic neuron counting was studied by deep learning. Results: After the establishment of PD rat model,successful model was screened out by behavioral test and preliminary judgment was made on the status of substantia nigra damage. The results of three kinds of immunohistochemistry tests proved the success of the model from the perspective of pathology. The method of deep learning was used to count the positive neurons in TH immunohistochemistry,and a preliminary algorithm on the count of neurons was obtained.The accuracy and recall of the neurons were both at high level. Conclusion: A preliminary algorithm based on deep learning with practical significance was obtained to investigate the dopaminergic neurons in the substantia nigra of rat model of Parkinson's disease.