Predicting Tumour Response

Abstract

Every year, thousands of women are diagnosed with breast cancer. To increase their survival chance and improve the treatment plans available, it is important to accurately predict the patient's response to treatment. For HER2+ breast cancer patients, normally a set number of chemotherapy rounds are administered. However, perhaps some patients would benefit from more chemotherapy or conversely, fewer rounds are sufficient. A reaction-diffusion model as well as a mechanically coupled reaction-diffusion model are used to predict the tumour cellularity. Using two magnetic resonance imaging (MRI) scans, a patient-specific model is calibrated to calculate the response for each individual patient. A third MRI scan is used to validate the results. Several preprocessing techniques are used to obtain the necessary information from the MRI scans such as tumour cellularity, segmentation of tissues and breast region.

The model calculations are performed in 2-dimensions and the slice thickness chosen determines how much effect the incorporation of surrounding tissue has. Using thicker slices means that the surrounding tissue has a larger effect on the tumour growth. For patients that respond well to treatment, the models are both able to calculate the downward trend in tumour cellularity. For thin slices (1.6 mm), both models perform similarly. Using thicker slices (6.5 mm), the mechanically coupled model shows better performance. The most important additions to the model are implementation in 3-dimensions so the slice thickness no longer has an effect. Most importantly, the chemotherapy should be incorporated to better obtain a patient-specific prediction of the response to treatment.