Google took advantage of this year's Annual Meeting of the American Association for Cancer Research - held April 16 - to introduce a new prototype that could make the automated analysis of the pathology of cancers a much more widespread practice. That's according to Google's Research Blog, which sheds some light on the project. The company refers to its new platform as an Augmented Reality Microscope (ARM). The technology is built on a modified light microscope, which enables real-time image analysis of results compiled by machine learning algorithms in the microscope's field of view. Visual feedback is shown on the screen in a variety of ways, depending on the problem the algorithms are trying to solve. For example, it can include arrows, contours or heatmaps, animations, or text to highlight whichever aspects of a tissue sample the system wants to point out.
Obviously, the platform wouldn't be much use to pathologists if it required new and expensive equipment. To that end, Google has focused on designing a system that can be added to the light microscopes that are already in use. Better still, the low-cost components required are already available and don't require whole slide digital versions of the tissues being analyzed. While the prototype and its accompanying paper are still very much under review, the promise of the technology goes beyond examining the pathology of any single type of cancer. Currently, it's being tested to detect breast cancer metastases in lymph node specimens or alongside prostatectomy specimens for prostate cancer. However, thanks to the fact that it ties directly into a light microscope, it could effectively be programmed to search for known indicators that any pathologist might look for.
The technical details which surround that analysis capability are equally impressive. The platform can operate and provide changing visual results in real time at up to 4-40x magnification at 10 frames per second. That could greatly accelerate the turnaround times where pathology-based analysis is concerned. Since it's based on machine learning, it will also continue to learn and improve with further use, if the platform or similar systems become widely adopted in the medical field. That won't necessarily happen at any point in the near future but it does seem to represent significant progress in Google's ongoing health efforts.