To save millions of lives from cancer, a team Google researchers innovated an AR (Augmented Reality) and ML (Machine Learning) powered microscope to help in real-time detection of cancer disease. On Monday, at the annual meeting of AACR – American Association for Cancer Research held in Chicago, Illinois, the Tech-giant (Google) illustrated a prototype ARM (Augmented Reality Microscope) platform that is designed to aid accelerate and democratize the implementation of deep learning devices for pathologists on the Earth.
The platform comprises a modified light microscope that activates real-time picture analysis and presentation of the outcome of machine learning algorithms right straight into the field of view. The Augmented Reality Microscope can be retrofitted into present light microscopes utilizing readily-availed components, cost-effect and with no need of entire slide digital forms of the tissues being examined.
The technical lead, Martin Stumpe and the Product Manager, Craig Mermel wrote in a blog that in a principle, the augmented reality microscope can offer a wide range of visual feedback that comprises text, heatmaps, arrows, contours. It is capable of running various types of ML algorithms to solve different problems like quantification, object detection.
Apps of deep learning for medical disciplines contains dermatology, ophthalmology, radiology, and pathology have revealed great potential.
Though, since the direct tissue visualization utilizing a compound light microscope rests the main means by which a pathologist identifies the disease, a serious barricade to the extensive acceptance of deep learning in pathology is the need of a digital demonstration of the microscopic tissue.
Deep learning models and modern computational components like those built upon source software called TensorFlow will permit a variety of pre-trained models to function on this platform. The team at Google configured ARM (Augmented Reality Microscope) to run 2 different cancer detection algorithms. One of these detects breast cancer metastases in the specimens of lymph node and other one detects prostate cancer in specimens of prostatectomy.
While both of these cancer models were initially trained on the pictures from a whole slide scanner with suggestively dissimilar optical configuration, the mockups performed extraordinarily well on the Augmented Reality Microscope with no added re-training, noted the Google Brain Team.