Imaging at diagnosis in oncology patients is crucial as it provides strong arguments for differential diagnosis, anatomical location, biopsy guidance, and local and distant metastases assessment. Most paediatric cancers are treated according to international protocols and trials that allow pooling of results of multiple centres, hence providing sufficient data to obtain significant differences between various treatment arms. This strategy allowed national and international groups (notably the International Society of Paediatric Oncology [SIOP] and the Children’s Oncology Group [COG]) to significantly increase the overall survival of affected children. Nevertheless, this is only possible if the protocol stratification is correctly followed, i.e., if the appropriate inclusion criteria are precisely used. Tumour stage is not the only inclusion/stratification criterion used, since treatments are also decided on the basis of biology (tumour markers), cyto/histological data (histological risk group, cytology or histology of bone marrow, lymph nodes and CSF) and, increasingly, on tumour genetics (chromosomal abnormalities, gene deletion, translocation, amplification). However, imaging is still one of the most important issues. Therefore, depending on tumour type, radiologists should be aware of the anatomical regions to analyze, the best imaging techniques to choose, and the appropriate diagnostic criteria. This information may be obtained from the literature when based on well-designed studies with pathological correlations. However, many staging criteria are also protocol specific and available only in the dedicated protocol imaging guidelines. In an ideal world, these guidelines should always be written by paediatric radiologists (after discussion with paediatricians, surgeons and radiotherapists) and should be available for any new treatment protocol. Actually, this is the only condition to unify the staging, and, therefore the stratification and treatment results. Prospective or retrospective central image review is increasingly used in oncology trials. As the quality of the review process is linked to the quality of the images, all imaging data of paediatric cancers should always be stored in DICOM format (on CD-ROM or on PACS, when available). Haematological malignancies (leukaemia, lymphoma) account for about 40% of all paediatric neoplasms and CNS tumours for about a further 25%. Among the remaining 35%, the most frequent paediatric solid neoplasms are neuroblastic tumours (neuroblastoma, ganglioneuroblastoma; representing about 9% of all paediatric cancers), renal tumours (about 6%; mostly Wilms tumour), soft-tissue sarcomas (about 5%; more than half being rhabdomyosarcoma), bone tumours (5%; mainly osteosarcoma and Ewing sarcoma), retinoblastoma (2%) and malignant gonadal and germ-cell tumours (3.5%). Carcinomas (adrenocortical, thyroid, nasopharyngeal, etc) and melanomas are rare (2.5%) compared to adults. Malignant hepatic tumours (hepatoblastoma, hepatocellular carcinoma, embryonal sarcoma, etc) only account for about 1% of all paediatric cancers. The aim of this review is to underline the imaging key points for disease staging of the most common solid tumours typically observed in children, i.e., Wilms tumour, neuroblastoma and rhabdomyosarcoma. Pediatr Radiol (2009) 39 (Suppl 3):S482–S490 DOI 10.1007/s00247-009-1193-x