Optical Tomography System for Breast Cancer Imaging
Breast cancer causes about 40,000 deaths a year, with one in eight women being diagnosed with the disease in their lifetime. Diffuse optical tomography (DOT) has been explored over the past decade as an alternative modality for cancer diagnosis and therapy monitoring. DOT uses near-infrared light to illuminate tissue from multiple points, and detects the light that has passed through the tissue. Growing tumors require an increased number of blood vessels to provide nutrients and remove waste products. However, the new vessels are poorly formed and tend to be disorganized and leaky. Diffuse optical tomography can visualize these vascular changes by the increased levels of oxy- and deoxy-hemoglobin in the tumor region. While these tumor characteristics can be used in detecting and diagnosing breast cancer, they can also be used to predict tumor response to chemotherapy. Currently we have a dynamic optical imaging system that is capable of imaging both breasts simultaneously. We can use the data collected from the system to reconstruct three-dimensional images that show concentrations of oxy- and deoxy-hemoglobin. Also, we can track of how concentrations change inside the breast over time when the patient holds her breath. Here, we present a novel approach to breast cancer detection and monitoring.
Hainsworth Shin
Faculty
You mentioned in that DOT was a fast imaging modality. But I really didnt see or hear any information regarding how fast imaging was. In particular, how long does it take to acquire an image? Also does the rate of imaging acquisition affect resolution. Finally, what is the smallest size tumors your technology can detect?
Thanks.
Jacqueline Gunther
Hi Dr. Shin,
Thank you for your question.
The imaging system can image at 1.7 frames per second, so it takes 0.576 seconds to take one image. However, the frame rate changes depending on how many detectors we decide to use. Therefore, with all detectors the frame rate is 1.7 Hz, but with less sources and detectors we can image faster since the optical switch in the system switches to fewer positions, which takes less time.
The rate of imaging does not affect the spatial resolution at all. The resolution is more dependent on the number of voxels in the mesh that we used. The meshes are three-dimensional geometries that mimic the shape of the breast and are created to solve our reconstruction algorithms within that volume.
At the moment, we have been able to image tumors that are 1 cm or larger, but we have studies ongoing to see if we can identify smaller tumors.
Peter Pfromm
Faculty: Project Co-PI
Would you please elaborate what your contribution to the work presented is? Did you develop algorithms for data evaluation, or modify the hardware? Is the idea of hemoglobin detection the heart of your work?
Jacqueline Gunther
Hi Dr. Pfromm,
Thanks for your question.
I’ve been developing methods to reduce image artifacts, which occasionally occur in our images due patient movement or poor contact between the tissue and a detector. Also, I’ve been working on ways to quantify and analyze three-dimensional static and dynamic data.
Oxy-hemoglobin and deoxy-hemoglobin are a couple of the parameters that we can observe, but there are other parameters as well, such as water percentage, lipid, and scattering coefficient. However, for our dynamic study we focus primarily on hemoglobin, because we are trying to observe the changes in vascularity inside the tumor region.
Mary Albert
Faculty: Project Co-PI
Good work! This work is aimed to improve an existing method; can you comment on your results so far, especially toward your stated goals of reliability of the image and increased patient comfort?
Jacqueline Gunther
Hi Dr. Albert,
Thank you very much.
One of the ways that the reliability of the image and patient comfort was increased was by creating the ring design for the patient interface. First, the ring design is comfortable for the patients. There is no painful compression and they can sit the entire time. Each ring can be moved up or down and accommodate a wide range of breast sizes. Additionally, the rings are placed on camera mounts that allow us to change the angle of the rings towards the patient and provide the most comfortable settings.
Second the ring design allows us to create individualized meshes for each subject. The meshes are three-dimensional geometries created to mimic the shape of each subject’s breast and are necessary for reconstructing the images. The reconstructions are dependent on the distance between our sources and detectors, but we know this information from our patient interface. Personalized meshes for each subject increases the reliability of our images.
Additionally, I have been working on ways to reduce image artifacts with the reconstructions to increase the reliability. Occasionally there are bad source-detector readings caused by different factors such as bad tissue contact with the fibers. These bad readings appear in the image as artifacts.
Christopher Buneo
Faculty: Project Co-PI
Nice video and poster! You mentioned (and demonstrated) the rapid switching of light sources during image acquisition — how many images from each light source are required for accurate reliable reconstruction and how is this determined?
Jacqueline Gunther
Thank you Dr. Buneo,
To obtain one image, we use the data collected from all the light sources, so one frame is considered to be the data collected from all 64 detectors from each of the 32 sources for all 4 wavelengths, which would be 32 × 64 × 4 = 8192 data points in total.
We only need one frame of data to reconstruct an image and this is done for the breath hold movies seen in the video. However, since we usually collect 1000 frames of data, we average about 50 frames to create our static images (such as in the neoadjuvant chemotherapy images). Averaging the frames reduces the noise of the signal.
Karen McDonald
Faculty: Project PI
Very nice job. I particularly like the fact that this technique does not require breast compression. How do you adjust the lasers for different sized breasts and how do you calibrate the instrument? Also what if you have lots of calcium deposits – do they interfere with light propagation? You mention that the device is capable of 2 images per second but is that for one laser source and how many laser sources do you need to apply to completely image the breast (and what is the minimum tumor size you can detect)?
Jacqueline Gunther
Thank you Dr. McDonald,
Our patients often comment that the imaging system is much more comfortable than mammography and even MRI.
The system has the ability to adjust the amplification given to each source-detector pair. The amplification is adjusted to optimize the detected signal before imaging each patient. The amplification/gain settings are determined automatically using custom software in LabView. The dynamic range of the system is about 120dB.
The major chromophores in breast tissue are oxy-hemoglobin, deoxy-hemoglobin, water, and lipid, which all absorb light in the near infrared (NIR) range. Calcifications show up on x-ray mammograms because calcium deposits absorb x-rays. Absorption caused by calcifications with NIR light is much lower than hemoglobin absorption and would not be enough to interfere with light propagation to the point where it affects image reconstructions.
The 2 images per second are taking for all 32 laser sources. The optical switch can rapidly switch through all these source positions. We can actually create an entire 3D image using one light source. However, by increasing the number of sources, we increase the amount of information we can used for our reconstruction algorithms. The reconstruction algorithms are essentially trying to solve the inverse problem by using optimization techniques. By increasing the number of sources, we increase the amount of information we can used for our reconstruction algorithms, and decrease the ill-posedness of the solution.
At the moment, we have been able to image tumors that are 1 cm or larger, but we have studies ongoing to see if we can identify smaller tumors.