Objective: To compare the accuracy of ultrasound imaging technique to that of clinical diagnosis in evaluating subcutaneous fat changes in HIV-infected subjects. Methods: HIV-uninfected control subjects (Group A), HIV-infected subjects with clinically assessed lipoatrophy (Group B), and HIV-infected subjects without clinical lipoatrophy (Group C) underwent ultrasound measurements of subcutaneous fat thickness at facial, brachial and thigh regions. ROC curve analyses were used to estimate ultrasound prediction accuracy and cut-off values of subcutaneous fat thickness. Results: 228 subjects were enrolled: 78 in Group A, 73 in Group B, and 77 in Group C. Facial lipoatrophy: ROC curve analysis identified optimal cut-off value of 13.3 mm [sensitivity, 96.0%; specificity, 76.9% AUC 0.92], 5.0 mm [sensitivity, 71.4%; specificity, 92.3%; AUC 0.90] and 11.2 mm [sensitivity, 95.8%; specificity, 89.7%; AUC 0.97] for females and 12.05 mm [sensitivity, 51.2%; specificity, 87.2%; AUC 0.74], 4.1 mm [sensitivity, 76.2%; specificity, 89.7%; AUC 0.85] and 4.35 mm [sensitivity, 60.0%; specificity, 89.7%; AUC 0.82] for males in assessing facial, brachial and crural lipoatrophy respectively. Using this cut-off values, 12/25 (48%) females and 17/49 (34.7%) males, 12/28 (42.9%) females and 23/49 (46.9%) males, 19/28 (67.9%) females and 12/49 (24.5%) males in Group C would be classified as “sub-clinical” facial, brachial and crural lipoatrophy respectively. Conclusions: The results of our study show that in the assessment of subtle bcutaneous fat changes ultrasound is more accurate than clinical evaluation and confirm the usefulness of ultrasound imaging technique in identifying lipoatrophy at an early stage.