Current recommended methods for managing carious lesions involve sealing the carious biomass within the tooth. At present, it is only possible to monitor sealed lesions over time using radiography, and digital subtraction radiography (DSR) has been demonstrated to be more accurate and reproducible compared to pairwise comparison of radiographs. The aim of this research project was to investigate in-vitro, whether alteration of X-ray source subject projection geometry affected the accuracy and reproducibility of DSR for detecting demineralisation in artificially created occlusal cavities. The discriminatory ability and reproducibility of a grading system for assessing changes in the proximal relationships of teeth, on paired digital radiographs taken with horizontal variations in X-ray source subject projection geometry, was also investigated. Digital radiographs were obtained of 40 extracted molar teeth with occlusal cavities following 7 and 15 degree horizontal, and 10 and 15 degree vertical angulation variations in X-ray source subject projection geometry in addition to 0 degrees. Following placement of demineralising solution in 19 cavities, 0 degree radiographs were taken after 12, 18 and 24 hours. Digital subtraction images were produced for each variation in angulation and length of demineralisation and scored by examiners as to their level of certainty that demineralisation of the occlusal cavity had occurred. The proximal relationships between teeth on the paired digital radiographs taken following 7 and 15 degree horizontal angulation variations in X-ray source subject projection geometry compared to 0 degrees were also scored by examiners using a grading system. The highest accuracy for detecting demineralisation in occlusal cavities using DSR was obtained when a reproducible 0 degree X-ray projection geometry was used. However, no statistically significant reduction in accuracy was identified compared to this when digital subtraction images were produced following a 7 degree horizontal angulation variation in X-ray source subject projection geometry after 12 and 24 hours demineralisation. Intra- and inter-examiner reproducibility was moderate. When a 7 degree horizontal angulation variation in X-ray source subject projection geometry existed between paired digital radiographs, the majority of differences observed regarding the size of inter-proximal spacing, or proximal overlapping between teeth was less than half the width of enamel. The intra- and inter-examiner reproducibility of the grading system was almost perfect.