Typical treatment for locally advanced breast cancer includes pre-surgical chemotherapy treatment to enable breast conserving surgery. Some patients, however, do not respond to chemotherapy, and endure months of treatment and unpleasant side-effects with no benefit. Near infrared spectroscopy is emerging as a promising candidate to differentiate patients responding to treatment from non-responders. Frequency-domain diffuse optical spectroscopy (FD-DOS) is capable of measuring concentrations of oxy-, and deoxyhemoglobin, lipid, and water in biological tissue. Changes in these concentrations over the course of chemotherapy can help predict a patient’s chemotherapy response. Most FD-DOS systems either use manually positioned, handheld probes, or complex arrays of source and detector fibers to acquire data from many tissue locations, allowing for the generation of 2D or 3D maps of tissue. Here, we present a new method to rapidly acquire a wide range of source-detector separations via high-speed mechanical scanning of a single source-detector pair. The scanning pattern chosen allows for the generation of axial line images of chromophore concentrations while the probe is stationary. Linear translation of the probe results in B-mode images that are capable of measuring the size and location of an absorbing inhomogeneity such as a tumor. Recently developed high speed acquisition electronics allow DOS data to be collected at nearly 100 Hz resulting in an a-line rate of 2-4 Hz. By utilizing a Deep Neural Network to estimate chromophore concentrations from the raw data, clinically relevant, depth resolved diffuse optical images of human tissue can be presented in real-time.
|