Friday, September 11, 2015

Contrast Agent Diffusion Inspired Breast Cancer Research

I've recently had an article published in the Journal of Digital Imaging focused on a biomarker that may be able to assist in the characterization of breast cancer from magnetic resonance imaging (MRI) examinations. 

The paper involves taking the second order spatial derivative of an easily extractable proxy for contrast agent concentration (which is closely related to contrast agent diffusion). Theoretically this might be able to assist in characterizing and detecting those early stage malignant tumours known as ductal carcinoma in situ (DCIS) that are prone to threaten a woman's life. The reason such an approach might theoretically be selective for those cases of DCIS that are dangerous is because a tumor would typically consume considerable nutrients from the blood stream in order to grow towards being a life threatening illness. Since the contrast agent used in a breast MRI examination is injected into the blood stream, if it accumulates in DCIS then it is highly likely that the tumor is consuming an ample quantity of nutrients. Access to nutrients is a necessary step for tumor growth and eventual spreading to neighbouring tissues (ie. becoming an invasive cancer). This is used as an argument in favour of breast MRI based cancer screening over mammography, however, radiologists exhibit substantial variability in their ability to detect these DCIS lesions from MRI examinations as they are one of the hardest types of tumors to identify. I am hopeful that approaches such as the one in my accompanying paper will one day be able to assist in the detection and characterization of dangerous forms of DCIS, however, I must emphasize that this is a theoretical long term hope for the future of this work.

At present the technique that I created was demonstrated to greatly reduce the rate of false positives when incorporated into a computer-aided detection system. The technique has not been shown to be able to assist in the identification/detection of those dangerous DCIS lesions. The technique is dependent on the spatial resolution of the imaging protocol (ie. how small the dots/pixels are in the MRI data collected). This work was performed on MRI data from one of the world's first high risk screening trials, as such the spatial resolution available in the future is liable to be greatly improved relative to what was available for this study. That said, it should be pointed out that one of the biggest problems with existing computer-aided detection technologies is that they produce far too many false positives which I have demonstrated that this technique is capable of assisting with. This biomarker also is capable of assisting in characterizing local variations in vascular heterogeneity.

Four example malignant lesions with colormaps illustrating the biomarker created for this paper is provided below.

For further reading you can access the full length paper here - though like most science publications it is behind a pay wall. These are often accessible for free through University libraries.