Armed with more than 20 issued patents and over 60 additional pending applications worldwide, our groundbreaking technology anchors on the notion that a picture is indeed worth a thousand words: Our system collects over 1,000 microscope images of each blood sample, in order to count cells and identify anomalies.
Blood is dense with cells, whereas microscopy works best when cells are neatly arranged in a flat layer. At Sight, we solved this problem uniquely using our patented Live Monolayer Imaging (LMI™) method: our single-use cartridge permits elegantly simple sample preparation, which preserves cell morphology, ensuring that our algorithms can achieve precise results. And because cell shape information is preserved, we will be able to provide even more data in the future—including valuable insight into blood cell morphology.
How do we clearly tell one cell type from another? We stain the cells with a patented combination of dyes that reveals normally-unseen chemical features. We then use a combination of brightfield and fluorescence microscopy to collect over 1,000 “multispectral” images of each sample. Our analyzers identify different cell populations through the combination of their optical and chemical signatures.
Merging a robust hardware sporting the latest optical instruments with advanced software for our Failsafe system, Sight’s analyzers provide uncompromised lab-grade accuracy without cumbersome QC requirements. Our Failsafe system routinely checks more than 30 parameters to ensure sample integrity, instrument calibration, test kit quality, and an error-free sample preparation process.
In order to accurately classify and enumerate cells, we’ve harnessed the latest AI technology and even improved it. Our machine vision is driven by a combination of convolutional neural networks, physics-based models and feature-based computer vision. These models are trained using our own learning schemes, which we have specifically developed and validated to derive accurate diagnostic results.
We have already accumulated image data from close to a million blood samples, and we continue to expand our database. This database serves as an invaluable tool in training Sight’s AI algorithm to continue to improve our diagnostic performance, as well as to develop new capabilities. One example of such R&D effort is detecting subtle changes in leukocyte morphology—indicating an underlying infection—and subtle changes in cell population—indicating certain cancer types.
With close to one petabyte of blood imagery data, our ambition is to go far beyond what we’ve already accomplished with complete blood count. We are confident that our technology and growing database of blood imagery will enable us to better understand the optical and chemical signatures of disease in the blood, and in turn help unlock the potential to improve the early detection and treatment of serious conditions like sepsis, cancer, and infections.