|
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.
|
