1 Dec. 2022. A developer of virtual tissue staining systems aided by algorithms for medical images began public operations today, raising $15.2 million in venture funds. PictorLabs Inc. in Los Angeles is a three year-old enterprise, spun-off from biomedical engineering research at University of California in Los Angeles.
PictorLabs seeks to provide faster and safer methods for staining tissue samples for microscopic analysis and diagnostics by pathologists. Pathologists need to stain tissue samples to enhance microscope images that highlight or contrast features and properties of cells in that tissue. Tumor biopsy images, for example, are stained to highlight cancerous cells, and often indicate boundaries separating diseased from healthy tissue. In some cases, multiple staining methods are needed to highlight or contrast images of tissue samples with complex conditions.
The company says today’s image staining techniques have remained largely unchanged for decades, using chemicals to produce fluorescence that highlights or contrasts colors in pathologists’ images. PictorLabs says conventional chemical staining methods are often time-consuming and in some cases can harm the samples, limiting their effective use. In addition, says the company, today’s staining techniques can return inconsistent results and generate chemical waste needing special handling.
The PictorLabs technology is based on research by the company’s founders UCLA electrical engineering professor Aydogan Ozcan and Yair Rivenson, a research fellow in Ozcan’s lab. Rivenson, Ozcan, and colleagues describe the technology in a paper published in March 2019, where the authors adapted two types of machine learning algorithms to create virtual tissue stain images. To create the virtual stains, the UCLA team used a convolutional neural network. In convolutional neural network algorithms, image analysis is combined with machine learning to deconstruct images into layers, with features of the image layers adding to a deeper understanding of image characteristics as more images are processed.
Identified characteristic biomarkers in 12 seconds
In this case, a convolutional neural network reconstructs images with algorithm-driven layers, including stained cells. To train the image reconstruction algorithms, the researchers employ a generative adversarial network, another machine learning application, that uses two sets of algorithms to test each other, for revealing and understanding underlying relationships in the data.
Ozcan, Rivenson, and lab colleagues demonstrate their histopathology technology in a paper published in October 2022, with virtual staining to identify breast cancer tissue expressing human epidermal growth factor receptor 2, or HER2, biomarkers. The authors says the process for staining breast cancer tissue for HER2 biomarkers is usually laborious, taking a day to complete, but the team achieved the task in 12 seconds — after one-time algorithm training — which independent judges found comparable in quality to chemical staining.
Ozcan and Rivenson formed PictorLabs in 2019, with Rivenson now the company’s CEO and chief technology officer. “By virtually staining tissue images in the digital realm,” says Rivenson in a company statement, “our approach simplifies and accelerates workflows, speeding up drug discovery and diagnostic decision-making with the goal of ultimately improving patient outcomes.”
PictorLabs is raising $15.2 million in its first venture finance round, led by M Ventures in Amsterdam, the investment arm of drug and diagnostics maker Merck, health technology developer SCC Soft Computer in Clearwater, Florida, and Koç Holding, an investment company in Istanbul. According to Crunchbase, PictorLabs earlier raised nearly $4 million in seed funds. PictorLabs says it plans to use the proceeds to expand the company’s biomarker target portfolio, develop new products, and expand staff.
More from Science & Enterprise:
- AI-Aided System to Identify Diabetes Complication Risks
- AI Analytics Boost Breast Cancer Assessment Diversity
- New Funds Awarded for Rapid Infection Diagnostics
- Trial to Assess Multi-Factor Blood Test for Cancer
- AI Recreates Tumor Environment for Precision Cancer Care
We designed Science & Enterprise for busy readers including investors, researchers, entrepreneurs, and students. Except for a narrow cookies and privacy strip for first-time visitors, we have no pop-ups blocking the entire page, nor distracting animated GIF graphics. If you want to subscribe for daily email alerts, you can do that here, or find the link in the upper left-hand corner of the desktop page. The site is free, with no paywall. But, of course, donations are gratefully accepted.
[wpedon id=”42724″ align=”center”]
* * *
You must be logged in to post a comment.