The work was carried out in two steps. First, a retrospective analysis of all chest CT performed in our radiology department on osteosarcoma and Ewing sarcoma paediatric patients who underwent the first staging CT in the period 2013-2017 was carried out. All the exams were performed using a GE LightSpeed Pro32 scanner equipped with tube current modulation but without any iterative reconstruction method. A total number of 39 patients and 190 acquisitions were selected. The exams, resident on the PACS (Fuji Synapse), were retrieved and anonymized. All patient and acquisition parameters such as age, sex, date of the exam, protocol used, slice thickness, kV, mA, rotation time, pitch, collimation width, filter type, scan length, noise index, CTDIvol and DLP etc... were recorded. A dose tracking system was available to deeply analyse dosimetric and examination data (Physico, Emme Esse, Italy).
Despite the existence of a reference low-dose protocol, important differences in the choice of acquisition parameters were found, so an analysis of image quality was then carried out to identify the best acquisition parameters and to draw up a protocol more tailored to our patients.
In this second step of the work the analysis was carried out on 76 chest CT scans performed on 9 patients having repeated CT scans (min 3, max 17) in 4 years, selected from the sample of 39 patients previously cited. Subjective image quality (IQ) was independently evaluated by 7 experienced paediatric radiologists, with a scoring scale from 0 to 4: 0 indicates a not evaluable exam, 1 unacceptable quality, 2 very poor quality, 3 adequate quality, 4 quality higher than needed.
The results obtained by the analysis of the 190 CT scans are summarized in Table 1 and in the graphs of Fig. 1 and Fig. 2, where CTDIvol and DLP are plotted as a function of the age of the patients. Data are depicted distinguishing the acquisition protocols. Observing the graphs it’s clear that acquisition CT protocols other than the reference ones are very frequently employed and dosimetric data referring to patients of the same age are often different.
The Dose Reference Levels (DRLs) suggested by the European Guidelines on Diagnostic Reference Revels for Paediatric Imaging RP 185  is reported for ages under 18. The DRL for adults as indicated in the Italian National Guideline on diagnostic and interventional DRL  is reported for ages ≥ 18 years. Even though the DRL value must never be referred to a single exam, it is extremely useful in giving an indication of the "dose level" below which the median of the sample should be.
Fig. 3 shows the frequency of Image Quality scores for all radiologists for all the exams. Since the image quality of almost one half of the exams was retrospectively judged unacceptable or very poor, images and scan parameters of exams scored 0, 1, 2 were then deeply analysed.
In Fig. 4 is reported an example of IQ mean score vs CTDIvol in CT scan repeated during a 4 years period on a single patient aged between 15-18 years, demonstrating the expected increase in IQ with the increase in CTDIvol, mainly because of noise reduction. The graph shows that a CTDIvol around 5 mGy corresponds to an IQ of about 3, i.e. acceptable by radiologists, confirming the CTDIvol value reported by the European Guideline for patients weighing 50-80 kg (i.e. 14-18 years) is reasonable.
Fig. 5, where CTDIvol vs Noise Index (NI) for all exams is reported, shows, as expected, that the noise is not the only parameter affecting the CTDIvol value, although a decreasing trend of the CTDI when NI increases can be appreciated.