In this paper, an improved BP neural network classification algorithm based on mean fractal principal component analysis (MFPCA) for ocular vascular ultrasound images is proposed to address the problem of low classification accuracy and difficulty in real-time assurance of ocular vascular ultrasound images. The algorithm introduces the theory of fractional-order calculus into the calculation of the covariance matrix of principal component analysis (PCA), which improves the linear dimensionality reduction and the main information retention ability of PCA. The mean value algorithm is utilized to improve the fractional order PCA algorithm to calculate the fractional order parameters. A classification algorithm that uses the improved Sigmoid function as the activation function of the BP network is proposed, which can greatly reduce the classification duration and the consumption of computational resources. Comparison experiments with classical algorithms show that the MFPCA ocular vascular ultrasound image classification algorithm proposed in this paper can effectively reduce the data dimensionality by 93.5%, increase the classification accuracy to 93.75%, and greatly improve the computational speed.