A Deep Learning Localization Method for Measuring Abdominal Muscle Dimensions in Ultrasound Images

Health professionals use 2D ultrasound images to measure abdominal muscles for diagnosing and treating low back pain, but interpretation is difficult and requires specialized training. This paper presents a deep learning approach using a modified Fully Convolutional Network architecture to automate abdominal muscle thickness measurements by localizing measurement endpoints. The method achieves a Mean Absolute Error of 0.3125 on the TrA400 dataset, comparable to skilled technicians, reducing observer variability and improving clinical outcomes.