Evaluation using mouse subcutaneous tumour models is a key process in cancer drug development. Tumour material is implanted subcutaneously and tumour growth measured using callipers. However this methodology can have poor reproducibility and accuracy due to observer variation. Furthermore the physical pressure of using callipers can distress the mouse and lead to tumour damage. Non-invasive digital tumour imaging would reduce handling stresses and allow volume determination without physical contact. This thesis focusses on capturing 2D digital images of subcutaneous tumours, then using image processing and machine learning methods to determine 3D volume. The biggest challenge faced was lack of differentiation between tumour and surrounding skin, rendering tumour boundary identification difficult. Whilst image processing methods such as colour segmentation and edge detection were unsuccessful, machine learning proved more successful. Three convolutional neural networks, VGG-Face, VGG-19 and VGG-16 models were evaluated, with VGG-Face producing the best results. Using the layer FC7 before RELU activation for extraction in the VGC-Face model, a tumour recognition rate of 98.86% was achieved. This was increased to 100% through a semi-automatic step with detection repeated on cropped versions of negatively classified images. Finally, volume was determined through extracting image features using the VGG-Face model and conducting partial least squares regression (error of 0.1). This work has successfully demonstrated that with computational methods the volume of subcutaneous tumours can be evaluated through non-invasive digital imaging without need to have contact with the tumour itself, thus offering refinement benefits to the mice as well as eliminating observer bias.