Method for automatically tracking muscle feather angle by combining convolutional neural network and Kalman filtering

Abstract

The invention discloses a method for automatically tracking a muscle feather angle by combining a convolutional neural network and Kalman filtering. The method specifically comprises the following steps: (1) preprocessing an ultrasonic image; (2) carrying out muscular membrane detection; (3) obtaining a muscle fiber direction observation value; (4) carrying out muscle fiber direction correction; and (5) calculating a feather angle. According to the method, the direction of the current muscle fiber is measured by using the deep convolutional neural network, and tracking of the feather angle isachieved by combining with the Kalman filter. According to the method, the robustness of a feather angle calculation algorithm is improved, the application field of an automatic feather angle labelingalgorithm is expanded, and an automatic feather angle tracking method is provided for an ultrasonic image sequence with poor quality.

Description

technical field [0001] The invention relates to the technical field of muscle plume angle detection, in particular to a method for automatically tracking muscle plume angle by combining a convolutional neural network and a Kalman filter. Background technique [0002] Fascicle plumage angle is an important parameter related to musculoskeletal function, which changes when the muscle stretches or contracts. Detection of muscle feathering angles allows for early detection of muscle lesions. The ultrasound image used to quantitatively measure the morphological parameters of muscle tissue is defined as sonomyography (SMG), from which muscle structural parameters such as muscle cross-sectional area, muscle thickness, feather angle, and muscle fiber length can be obtained. [0003] As a hot spot in the field of machine learning in recent years, deep learning has gained great attention in the field of image processing by virtue of its high prediction accuracy in recognition applicat...

AI Technical Summary

Problems solved by technology

[0009] In the prior art, ①Use the voting Hough transform to calculate the angle of the sarcolemma line and the muscle fiber bundle on the ultrasonic image. Although the Hough transform has a better effect in locating the sarcomere, it cannot locate the muscle fiber well; ②Using the ultrasonic image The local Radon transform is used to calculate the angle of the sarcolemma and muscle fiber bundles. When the muscle fibers are unclear and the noise is large, the performance of the Radon transform drops significantly; The method also relies on the measurement value of the traditional image processing method, that is, the detection of local Radon transformation, so the limitation is still large

Muscle fibers have the characteristics of intermittent and intermittent, and the method based on image processing is not robust

Images