EuroSciPy logo


Cambridge, UK - 27-30 August 2014

Analysis of multi-modality medical image data for improving radiotherapy of head and neck cancers

Dualta McQuaid


Background: In the european union about 100,000 people are diagnosed with head and neck cancer every year and about 40,000 die from the disease. They are commonly treated with a combination of chemotherapy and radiotherapy. Individual tumours are spatially very diverse but current radiotherapy treatment is directed uniformly at the entire tumour. A major hypothesis is that a biologically inhomogeneous tumour would benefit from a carefully planned inhomogeneous radiotherapy treatment. This treatment could then be guided by imaging of the tumour function by various means and the best method is currently not clear. Clinical trials are underway to help determine the best types of imaging and at the best time points to achieve the best type of treatment. To do this many different types of medical images at many time points and for many patients are acquired.

Computational tools: The medical images are stored in the XNAT research PACS system (, which is accessed using the pyxnat python module ( Pydicom ( is used to read the data and metadata and apply necessary filters to allow all the data to be imported into the propriety radiotherapy computational platforms: Pinnacle (Philips, WI, USA) and RayStation (RaySearch, Stockholm, Sweden) these systems employ internal python and ironpython scripting respectively. These internal python scripting tools are then used to automate many data processing tasks. The image data is registered and reviewed by clinicians before a second set of python scripts prepare sets of datafiles listing image values at equivalent spatial positions for successive images and different time points and for different image modalities at particular timepoints. This data is then analysed with python using pandas ( and scikit-learn (; and with R using RPy (