We developed a fast Raman spectroscopic discrimination system based on a slit-scanning confocal microscope and machine learning. The speed of discrimination was improved by reducing the number of measurements, without measuring all points in the field of view. During discrimination, the system continues to evaluate the spectra already obtained, which guarantees the accuracy of the discrimination and enables early detection of anomalies by optimizing the measurement positions. We performed discrimination using a mixture of polystyrene (PS) and polymethyl methacrylate (PMMA) microbeads as a sample to mimic cancer tissue and that of fatty liver tissue using mouse liver tissue samples. The results showed that the discrimination was about 2-11 times faster than that by slit scanning confocal microscopy.
KEYWORDS: Raman spectroscopy, Light sources and illumination, Machine learning, Medical research, Random forests, Microscopy, Microscopes, Engineering, Diagnostics, Decision trees
We propose a method that combines high-speed Raman imaging with a machine learning technique, multi-armed bandit, to achieve rapid and accurate identification of samples under observation. First, our method dvides the field of view of the sample into small sections, and it returns either ’positive’ or ’negative’ based on whether the sections with high anomaly indices exceed a certain proportion. Moreover, the points to be measured are determined dynamically and automatically generating a series of optimal illumination patterns.
We developed spontaneous Raman microscopy using Bandit algorithm to realize fast diagnosis of the existence of anomalies or not with guaranteeing accuracy. The algorithm evaluates obtained Raman spectra during measurement to judge if the diagnosis is completed with ensuring an allowance error rate that users decided and also to generate optimal illumination patterns for the next irradiation which are optimized to accelerate the detection of anomaly. We present our simulation and experimental studies to show that our system can accelerate more than a few tens times faster than line-scanning Raman microscopy which requires full scanning over all pixels.
We present our recent study combined multi-armed Bandits algorithm in reinforcement learning with spontaneous Raman microscope for the acceleration of the measurements by designing and generating optimal illumination pattern “on the fly” during the measurements while keeping the accuracy of diagnosis. We present our simulation and experimental studies using Raman images in the diagnosis of follicular thyroid carcinoma and non-alcoholic fatty liver disease, and show that this protocol can accelerate more than a few tens times in speedy and accurate diagnoses faster than line-scanning Raman microscope that requires the full detailed scanning over all pixels.
The on-the-fly Raman image microscopy designs to accelerate measurements by combining one of reinforcement machine learning techniques, bandit algorithm utilized in the Monte Carlo tree search in alpha-GO, and a programmable illumination system. Given a descriptor based on Raman signals to quantify the likelihood of the predefined quantity to be evaluated, e.g., the degree of cancers, the on-the-fly Raman image microscopy evaluates the upper and lower confidence bounds in addition to the sample average of that quantity based on finite point/line illuminations, and then the bandit algorithm feedbacks the desired illumination pattern to accelerate the detection of the anomaly, during the measurement to the microscope.
Most conventional bandit algorithms assume that infinite number of measurements or samples provides us with 100% accuracy. However, in Raman measurements we should develop both a Raman descriptor to quantify the degree of anomaly, and a new algorithm to take into account the finite accuracy lower than 100%. This microscope can also be applied to other problems, besides detection of cancer cells, such as anomaly or defects of materials. The algorithm itself is also beneficial and transferrable to the other microscopes such as infrared image microscope.
We investigated the axial resolution and signal-to-noise ratio (SNR) characteristics in deep-tissue imaging by 1.7-μm optical coherence tomography (OCT) with the axial resolution of 4.3 μm in tissue. Because 1.7-μm OCT requires a light source with a spectral width of more than 300 nm full-width at half maximum to achieve such high resolution, the axial resolution in the tissue might be degraded by spectral distortion and chromatic dispersion mismatching between the sample and reference arms. In addition, degradation of the axial resolution would also lead to reduced SNR. Here, we quantitatively evaluated the degradation of the axial resolution and the resulting decrease in SNR by measuring interference signals through a lipid mixture serving as a turbid tissue phantom with large scattering and absorption coefficients. Although the axial resolution was reduced by a factor of ∼6 after passing through a 2-mm-thick tissue phantom, our result clearly showed that compensation of the dispersion mismatching allowed us to achieve an axial resolution of 4.3 μm in tissue and improve the SNR by ∼5 dB compared with the case where dispersion mismatching was not compensated. This improvement was also confirmed in the observation of a hamster’s cheek pouch in a buffer solution.
Optical coherence microscopy (OCM) is a high-resolution imaging technique based on optical coherence tomography and confocal microscopy. The recent studies on OCM operating at 800-1300 nm spectral region have shown that OCM enables to visualize micrometer- or sub-micrometer-scale structures of animal tissues. Although OCMs offers such high-resolution label-free imaging capability of animal tissues, the imaging depth was restricted by multiple light scattering and light absorption of water in samples. Here, for high-resolution deep-tissue imaging, we developed an OCM in the 1700-nm spectral band by using a supercontinuum (SC) source with a Gaussian-like spectral shape in the wavelength region. Recently, it has been reported that the 1700-nm spectral band is a promising choice for enhancing the imaging depth in the observation of turbid scattering tissues because of the low attenuation coefficient of light. In this study, to clarify that the 1700-nm OCM has a potential to realize the enhanced imaging depth, we compared the attenuation of the signal-to-noise ratio between the 1700-nm and 1300-nm OCM imaging of a mouse brain under the same signal detection sensitivity condition. The result shows that the 1700-nm OCM enables us to achieve the enhanced imaging depth. In this 1700-nm OCM, we also confirmed that the lateral resolution of 1.3 µm and axial resolution of 2.8 µm in tissue were achieved.
We developed full-range, ultrahigh-resolution (UHR) spectral-domain optical coherence tomography (SD-OCT) in 1.7 um wavelength region for high-resolution and deep-penetration OCT imaging of turbid tissues. To realize an ultrahigh axial resolution, the ultra-broadband supercontinuum source at 1.7 um wavelength with a spectral width of 0.4 um at FWHM and home-built spectrometer with a detection range from 1.4 to 2.0 um were employed. Consequently, we achieved the axial resolution of 3.6 um in tissue (a refractive index n = 1.38). To observe deep regions of turbid tissues while keeping the ultrahigh axial resolution, a full-range OCT method to eliminate a coherent ghost image was utilized for our UHR-SD-OCT. Because the full-range method allows us to avoid the formation of a coherent ghost image when the zero delay position is in the inside of specimens, we set the zero delay position to the laser focus position in this study, and then, a region of interest in specimens was moved to the laser focus position where the highest signal intensity is achieved, resulting in the improvement of the observation depth. Thanks to the deep-penetration property of the 1.7 um light and elimination of a ghost image, we successfully demonstrated the visualization of the mouse brain structures at a depth over 1.5 mm from the surface with the 1.7 um UHR-SD-OCT. In this experiment, we confirmed that the brain specific structures, such as corpus callosum, pyramidal cell layer, and hippocampus, were clearly observed.
Optical coherence tomography (OCT) is a non-invasive optical imaging technology for micron-scale cross-sectional imaging of biological tissue and materials. We have been investigating ultrahigh resolution optical coherence tomography (UHR-OCT) using fiber based supercontinuum (SC) source. Although UHR-OCT has many advantages in medical equipments, low penetration depth is a serious limitation for wider applications. Recently, we have demonstrated high penetration depth UHR-OCT by use of fiber based Gaussian shaped SC source at 1.7 μm center wavelength. However, the penetration depth has been limited by the low power of SC source. In this paper, to realize deeper penetration imaging, we have developed the high power Gaussian shaped SC source at 1.7 μm wavelength region based on the custom-made Er-doped ultrashort pulse fiber laser with single-wall carbon nanotube and nonlinear phenomena in fibers. This SC source has 43.3 mW output power, 242 nm full-width at half maximum bandwidth, and 109 MHz repetition rate. The repetition rate and average power were almost twice as large as those of previous SC source. Using this light source, 105 dB sensitivity and ultrahigh resolution of 4.3 μm in tissue were achieved simultaneously. We have demonstrated the UHR-OCT imaging of pig thyroid gland and hamster’s cheek pouch with this developed SC source and compared the images with those measured by the previous SC source. We have observed the fine structures such as round or oval follicles, epithelium, connective tissue band, and muscular layer. From the comparison of the UHR-OCT images and signals, we confirmed the improvement of imaging contrast and penetration depth with the developed SC source.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.