Better Imaging Helps Treatment of Colorectal Cancer
Colorectal cancer (CRC) is a worldwide scourge, with more than 650,000 new cases and 400,000 deaths occurring each year. In light of these statistics, accurate diagnosis and effective treatment planning for CRC are of paramount importance, and medical imaging techniques, including magnetic resonance imaging (MRI), play a critical role in detecting colorectal tumors, determining the stage of the disease, and monitoring patients’ responses to treatment. However, these CRC images are generally noisy, complex, and highly textured, which makes their analysis challenging.
Clinicians must address three crucial questions in CRC treatment planning:
- How big is the tumor, and how far has it infiltrated the surrounding fat?
- What is the margin of clearance between the mesorectal fascia (MF) and the tumor? This margin is referred to as circumferential resection margin (CRM). If the minimum CRM is below a certain threshold, surgical treatment is less likely to be successful.
- Has the cancer spread to any lymph nodes? A metastasis of the disease is quite likely if the lymph nodes are malignant.
Recognizing the inherent difficulties in analyzing CRC images, researchers at the University of Oxford's Department of Engineering Science and e-Research Centre have developed automated and semi-automated image analysis methods to help clinicians extract quantitative data from MRI images. These data provide useful insights into the three key clinical questions listed above, in a way that is not possible with simple visual assessment of the images.
To address the first two key clinical questions, we have developed non-parametric statistics-based contour evolution methods to segment the colorectal MRI images, to delineate the MF, and to estimate the CRM. The maximum of the average difference between an expert’s delineation of the MF and that segmented by the new algorithm is just over 2 millimeters. With these methods, we can also produce a 3-D map of the resection margin, which can be used in surgical planning.
But perhaps the most exciting aspect of our work has been our development of the first algorithms that can reliably assess lymph node status in colorectal cancer. The "holy grail" of colorectal cancer image analysis, the status of lymph nodes is the determinant of whether a patient should proceed to surgery or should be assessed for metastatic spread. In the latter case, systemic chemotherapy may be called for, or, equally likely, palliative care.