The Effect of Ultrasound-Guided Treatment of Symptomatic Partial Rotator Cuff Injury Under the Background of Deep Learning

Abstract

Deep learning is one of the most powerful machine learning algorithms at present. The convolutional neural network model has the ability to automatically learn features, and has a greater performance advantage than other deep learning models in the field of image processing. Based on deep learning, the effects of platelet-rich plasma injection of betamethasone on shoulder pain and function in patients with partial rotator cuff injury were investigated. Eighty patients with tendinosis or a partial tear (<1.0 cm) were recruited between 2021 and 2022. Mixed injection of 1 mL betamethasone and 5 mL lidocaine, following 5 mL PRP injected in 7-day intervals, were infiltrated directly into the lesion of supraspinatus interstitial RCT under ultrasound control. The patients were assessed by the SPADI, CMS and ASES before and 1, 3, 6, 9 months after injection. MRI was also performed before and 6 months after injection. All patients had significant better ASES, SST, CMS score and improvement of motion and pain relief after PRP injection, compared with the baseline before injection (p < 0.05). In the MRI data, statistical pre- and post-comparisons revealed significant differences between location of partial rupture and grade of tendinopathy. There were no mortality and major complications observed during follow-up. Under the background of deep learning, ultrasound-guided injection of PRP with betamethasone can progressively relieve shoulder joint pain and improve motor function in patients with partial rotator cuff injury. Ultrasound-guided injection of PRP with betamethasone is safe and useful for rotator cuff disease.