People often automatically connect "optical imaging" with beautiful and colorful microscopic images of cells. It is true that traditional and emerging microscopy techniques (con-focal, 2-photon, photoacoustic microscopy, OCT etc) are important members of optical imaging, but there are more.
The COTI lab focuses on a particular type of optical imaging, called diffuse optical tomography (DOT). A unique capability of DOT is that it can "see through" several or even over a dozen centimetres of opaque tissues without needing to cut open the tissue (also known as non-invasiveness), kind of like x-ray CT or MRI. However, compared to x-ray CT, DOT only uses safe low-energy near-infrared photons, similar to the sunlight shining on your face when you take an outdoor walk. In comparison, x-ray photons are much more energetic and may damage cells if the radiation is prolonged. Using DOT, we can computationally reconstruct 3-D maps of blood concentrations (how much blood per volume) and oxygen saturation (the percentage of red blood cells that carry oxygen) in a thick tissue sample. These physiological markers are excellent indicators for diseases such as malignant tumors, inflammation or vascular diseases. In comparison, most microscopy techniques can only see through tissues no deeper than 1 mm.
In our research, we primarily use red and infrared wavelengths in the light spectrum to perform DOT measurements. This is because the major light absorbers in the tissue, known as chromophores, are primarily oxygenated hemoglobin (HbO2), deoxygenated hemoglobin (HbR), water (H2O), lipids, melanin (the pigment made up our skin and hair colors) etc, have a relatively low absorption between 600 nm (red-colored) and 1000 nm (infrared and invisible) in light wavelengths. This window is referred to as the (1st) optical window. The provides researchers an opportunity to image deeply embedded tissue structures using red and near-infrared light.
Although it sounds fantastic if one can use light to create tomographic (i.e. 3-D) images of tissues, DOT is actually pretty difficult to do in general. There are a lot of challenges that have to be solved first in order to make this work. First, near-infrared photons are much much less energetic than x-ray photons. So, when it enters the tissue, it does not follow a straight-line path like x-ray does (after collimation of course), instead, it is bounced around, a lot, by tissue cells like a ping-pong ball. This is called scattering. Tissue scattering is the biggest "enemy" for creating clear images (for any modality, not just DOT), because people can not precisely know where such photon had traveled.
Photons not only scatter a lot in the tissue, they also attenuate quickly by hitting absorbers like blood. The high absorption of tissues makes light signals extremely weak (10-7~10-8) after propagating centers of tissues and arriving at a detector. So, in order to get enough photons to give meaningful measurements, DOT requires to use extremely sensitive light detectors, such as PMT (photon multiplier tubes) and APD (avalanche photodiode), EM-amplified CCD cameras etc.
The presence of both scattering and absorption makes light spread-out in the tissue like a drop of food coloring in the water - they spread out in all directions and attenuate exponentially as they expand. The light intensity through complex tissues can be approximated by a partial differential equation known as the diffusion equation (DE). If tissues have complex structures (such as human brain with multiple layers, bones, or skins), one has to solve this DE numerically using computer algorithms. This often requires specialized software and we have to write our own solvers.
Modeling how light travels is only a part of DOT. In real-world applications, the tissue structure is not even known (if it had been known ahead of time, then there was no need to do any imaging). So, how can one figure out the internal tissue structure by shining light on the surface of an unknown tissue structure (like human head or breast), and then measuring light intensity reemerging on the tissue surface is the core question in DOT (and many tomographic imaging modalities in general). This is the so called "inverse problem". In the case of x-ray CT, the inverse problem can be solved by a linear equation, but this no longer works in DOT because of the presence of scattering and complex relationship between absorption/scattering to light distribution. To solve the DOT image reconstruction problem, we have to develop sophisticated reconstruction algorithms, and this is still a major research area for innovation, including how to improve image quality using prior information.
In short, our lab uses near-infrared light to create 3-D maps of tissue blood concentrations and other useful markers to help doctors to find tumors and diseases many centimetres below the tissue surface. We develop both optical imaging instruments as well as sophisticated computer programs to make DOT a promising technology for the future of healthcare.
In our project of multi-modal breast imaging system, we want to add functional information using DOT to the morphological (tissue shapes) information provided by conventional modalities such as x-ray mammography or tomosynthesis. As you know, early diagnosis of breast cancer is critically important. However, the current clinical approach, x-ray mammography, is poor in specificity – over 85% of the recalls yielded benign findings. Furthermore, mammography misses 40% of early stage cancers. Over the past decade, we have been developing a novel imaging technique to find breast cancers by combining safe, non-invasive near-infrared diffuse optical imaging with high-resolution x-ray mammography. Since 2009, we've been leading this research and conducted a clinical study of hundreds of patients, in collaboration with world's leading-expert in breast imaging, Dr. Daniel Kopans from MGH Avon Center. With our innovative image reconstruction algorithms, we have demonstrated the potentials in differentiating malignant from benign lesions using the functional and structural information together. Our findings were highlighted as a front-cover article in Radiology. In 2011, our lab started a collaboration with Philips Healthcare to accelerate the clinical translation of this technique. Our breast imaging research project was highlighted by the former Massachusetts Governor Deval Patrick in his speech during the “Friends of Cancer Research” forum in 2014.
Currently, we are developing our next-generation breast imaging system, OMCI (optical mammography co-imager), funded by an NIH R01 grant from the National Cancer Institute (NCI).
Supported by an NIH NIGMS R01 grant, our team have been continuously improving MCX/MMC, and made it one of the fastest and feature-rich light transport simulation platform available today. As of 2020, our papers have attracted over 1100 citations, 2,400 registered users and 30,000 downloads worldwide. In the meantime, we are continue pushing the state-of-the-art of Monte Carlo light modeling and developed dual-grid MMC (DMMC), split-voxel MC (SPMC), implicit MMC (iMMC), photon-replay, photon-sharing etc in the past years.
If you have a strong background in GPU programming, computational physics, and open-source software development, please reach out to Dr. Fang. We are constantly expanding our team to develop top-of-the-line software to our community.