PARALLEL COMPUTING ILLUMINATING A PATH TO EARLY CANCER DETECTION
Every year about six million new cases of cancer are discovered worldwide. Catching signs of the disease early enough to treat it, is a key factor in success, and increasingly researchers are using parallel computing to help perform crucial diagnostics. Marking World Cancer Day on Feb. 4, our team in Russia shared news about promising developments by scientists using NVIDIA technology to improve cancer detection.
Abnormally proliferating cells common to cancer are known as neoplasms. Finding them in living tissues is a needle-in-a-haystack game, requiring cutting edge biomedical work. One of the more advanced diagnostic tests for neoplasms is diffuse fluorescent tomography (DFT), which looks at the absorption and scattering of light in tissues to spot dangerous growths.
It works like this: special fluorescent markers that stick to malignant cells are injected into the body. When the tissues are illuminated by light at a certain wavelength, the markers fluoresce, indicating the locations of abnormal cell proliferation. A hurdle with this test, is that light disperses on its way through the body, making it difficult to see the markers if malignant cells are located deep within the body.
To overcome this issue, researchers at the Applied Physics Institute at the Russian Academy of Science (RAS) began simulating light and radiation propagation through tissue. They developed algorithms to reconstruct the 3D position of the fluorescent markers used in DFT. The result was that they could pinpoint neoplasms with greater accuracy.
The researchers used the Monte Carlo method (which uses repeated random sampling) to achieve their simulations. The calculations required to run these simulations are intense: a typical situation requires the calculation of roughly one billion random paths. While this process can be painfully slow on a CPU, it is perfectly suited to parallel processing. When the scientists switched to a GPU-based system, the average runtime for the tests went from about two and a half hours to 1.5 minutes – a hundred-fold speedup – and as a result, the researchers could add more paths to the calculations, which ultimately increases accuracy.
The problem of light dispersion through human tissue has implications for more than just DFT diagnostics. Oncologists who treat cancer patients with radiation also need to understand how radiation propagates through tissue and internal structures. If they have more accurate models on radiation’s path through the body, doctors can more precisely target radiation therapy to malignant cells, leaving healthy ones unharmed.
We’re excited by these developments from the team at RAS and are looking to hear from others about their GPU-powered research.