Thesis Project Form

Chest X-rays in premature infants require a delicate balance between reducing radiation doses and maintaining diagnostic quality. The absence of objective and reproducible metrics for assessing image quality limits the optimisation of protocols. Artificial Intelligence offers advanced tools for automatic, standardised and clinically relevant assessment, applicable in paediatric imaging and radiation protection.

The overall objective of the thesis is to design, train and validate an Artificial Intelligence model capable of automatically estimating the diagnostic quality of AP chest X-ray images acquired at bedside in premature infants, in a real clinical setting at a paediatric IRCCS (Scientific Institute for Research, Hospitalisation and Healthcare). The work involves: (i) analysing DICOM data and quality scores assigned by expert radiologists; (ii) defining image-based features and/or training supervised deep learning models using the composite quality score as a reference; (iii) evaluating the model's performance in terms of accuracy, robustness and generalisability; (iv) analysis of the impact of dose, additional filtration and clinical factors on model estimates, with the aim of supporting the optimisation of low-dose radiological protocols.

Students will acquire advanced skills in medical image analysis, artificial intelligence applied to healthcare, and clinical data management. They will also develop skills in integrating computational models, clinical assessment, and radiation protection in paediatrics.