Background: Paediatric tumour cytological diagnosis by image-guided fine needle aspiration biopsy (FNAB) with rapid on-site evaluation (ROSE) has not gained wide acceptance despite increasing publications advocating the procedure.Objective: The primary aim was an audit of the diagnostic yield and accuracy of paediatric image-guided FNAB with ROSE at a single institution. Evaluation of safety was a secondary aim.Method: Details of consecutive cases of paediatric image-guided FNAB with ROSE for suspected non-benign deep-seated lesions performed from 01 January 2014 to 30 April 2020 were retrieved from the institutional radiology and laboratory databases. Diagnostic yield and accuracy were evaluated using clinico-pathological-radiological correlation and/or subsequent histological specimen diagnosis correlation. Complications and the frequency of key radiological features potentially affecting yield and accuracy were described.Results: Of 65 cases retrieved, cytology showed malignancy in 52, benign features in five and one indeterminate diagnosis; seven samples were insufficient for cytological assessment. Of the 65 cases, 58 had subsequent formal histological diagnosis. The overall diagnostic yield was 98.5%, with 94.5% sensitivity, 100.0% specificity, 100.0% positive predictive value, 75.0% negative predictive value and 95.3% diagnostic accuracy. All cases (n = 26) demonstrating restricted diffusion on MRI yielded adequate samples and cyto-histopathological correlation.Conclusion: Paediatric image-guided FNAB with ROSE has a relatively high diagnostic yield, sensitivity, specificity, positive predictive value and accuracy in the diagnosis of deep-seated tumours. The relatively low negative predictive value may reflect insufficient samples obtained from cystic and/or benign lesions. Sampling from areas of restricted MRI diffusion may enhance diagnostic yield.