The purpose of this study was to assess the diagnostic value of imaging-guided core needle biopsy for the diagnosis of musculoskeletal lesions. Between 2004 and 2007, 309 biopsies (ultrasound 151, computed tomography 89, and fluoroscopy 69) were included. There were 142 soft tissue and 167 bony lesions. Diagnostic yields and accuracies were assessed using the chi-square test or Fisher's exact test with Bonferroni's correction when necessary. Overall diagnostic yield was 90.6% for all 309 lesions (bone 91.6% vs. soft tissue 89.3%, p = 0.5125). The diagnostic accuracy of the 185 core needle biopsies, which were confirmed by definitive surgical biopsies, was 84.3% (bone 88.9% vs. soft tissue 79.1%, p = 0.0669). The yields of homogenous bone tumours (96.8%) were not significantly higher than those of bone tumours with a heterogenic architecture (86.4%, p = 0.0794). The difference between accuracies for homogenous bone tumours (89.1%) and heterogenous bone tumours (85.0%) was not significant (p = 0.6930). However, for soft tissue tumours, homogenous tumours had a significantly higher diagnostic yield than heterogenous tumours (97.5% vs. 81.4%, p = 0.0036). Diagnostic accuracy for homogenous tumours was also significantly higher than that for heterogenous soft tissue tumours (94.4% vs. 60.6%, p < 0.0001). The image-guided percutaneous needle biopsy of musculoskeletal lesions is a safe and effective procedure if it is performed selectively in soft tissue tumours with homogenous architectures.