A Study of Lumbar Vertebral Body and Lumbar Disc Herniation Disease Detection Based on YOLOv8n

Aiming at the problem of diverse and complex structures of different tissues in lumbar spine MRI images. In this paper, an improved YOLOv8n lumbar spine and lumbar disc herniation detection method is proposed. The ODConv module is introduced in the backbone part to replace the traditional static convolution module to improve the feature extraction ability of the model; DySample is used to replace the nearest-neighbor interpolation upsampling to improve the ability of the model to convey semantic information to the feature maps; the LSK attention mechanism is used as a basic module and combined with ASFF to form LSK-ASFF, which is introduced into the multi-scale feature fusion structure of the network to realize the adaptive spatial feature fusion of different levels of features to improve the accuracy and generalization ability of the network. The experimental results show that the precision(P), F1 score and mAP of the improved YOLOv8n model in the lumbar MRI dataset reach 90.1%, 89.95% and 93.9%, which are 9.1%, 4.78% and 3.8% higher than those of the baseline model, respectively. The improved network model comprehensively increased the detection accuracy and provided an effective method for the detection of lumbar spine and lumbar disc herniation.