Objective: To evaluate the diagnostic accuracy of ultrasound (US) in comparison with magnetic resonance imaging (MRI), which is the gold standard for detecting full- and partial-thickness supraspinatus (SS) tendon tears at our general practice. Methods: From January 2015 to January 2016, 107 patients suffering from shoulder pain underwent both US and MRI of the shoulder at a medical center to identify their shoulder pathology, and were included in this work. The US was performed by a physiatrist at a US laboratory, and the US findings for the SS tendon were classified as normal, tendinosis (calcific/ non-calcific), a partial-thickness tear, or a full-thickness tear. MRI results were reported by a radiologist. The sensitivity, specificity, and accuracy of the sonographic assessment were determined by comparing them with the MRI findings. Correlations between the US and MRI results were analyzed using Cohen's kappa coefficient. Results: The results showed that the kappa coefficient of association was 0.80 for SS full-thickness tendon tears. This indicated that the agreement between US and MRI for the diagnosis of SS full-thickness tendon tears was excellent. Moreover, the accuracy of US in detecting full-thickness tears of the SS tendon, if we take MRI to be the gold standard, was 0.91, with a sensitivity of 0.91, and a specificity of 0.90. However, when using US to detect partial-thickness tears of the SS tendon, the accuracy and sensitivity were low-0.73 and 0.30 respectively-although there was high specificity (0.92). Conclusions: US was highly accurate for detecting SS full-thickness tendon tears in daily general practice, but was less sensitive and accurate for detecting SS partial-thickness tendon tears. Further large-scale studies are needed.