SVISION is awarded two new patents with broad coverage for AI microscopy applications

SVISION is awarded two new patents with broad coverage for AI-enabled image segmentation and pattern recognition.

BELLEVUE, WA, US, March 4th, 2021 - SVISION LLC, the world’s leading AI microscopy software company, a pioneer in AI-enabled computer vision and the creator of Aivia has been issued two new US patents covering parameter free image pattern recognition as well as fast, guided training and applying of deep neural networks for semantic segmentation.

AI-enabled image analysis and knowledge creation solutions continue to revolutionize the life science and bio pharma fields thus accelerating scientific discoveries and the development of novel therapeutics. The latest inventions by SVISION addresses two major bottlenecks related to the deployment and adoption of AI-enabled computer vision solutions.

Automated parameterization image pattern recognition method.

10,769,432 (September 8th, 2020)

A typical image pattern recognition processing flow includes the segmentation of pattern regions from images, the detection of patterns of interest and the classification of patterns into different classes. The state-of-the-art image pattern recognition tools require a user to have a good understanding of image processing algorithms and master several user-facing parameters before one can efficiently use the tools. The new patent covers a complete parameter-free image segmentation, pattern detection and classification pipeline. In this pipeline, image segmentation, pattern detection and classification parameters are automatically learned from the image data itself. The user only needs to define the patterns of interest (e.g. a cell type or phenotype) by labeling a few representative patterns in learning images. The trained pipeline can be applied without any parameter settings to any number of images and the internal pipeline parameters can also be updated by additional learning. This approach has now been deployed for all users of Aivia’s and AiviaWeb’s Pixel Classifier.

Optimal and efficient machine l earning method for deep semantic segmentation.

10,891,523 (January 12th, 2021)

Deep neural networks are the best (i.e. yield the most accurate results) solution for image semantic segmentation. However, this type of approach often relies on access to vast amounts of labelled data (ground truth for supervised learning) which in many cases is costly or not readily available thus rendering this type of solution unviable. The new patent solves this problem by using small amounts of partial ground truth for fast learning followed by optimal learning to iteratively update the deep model. The patent further covers optimal transfer learning from pre-trained semantic segmentation deep model. This invention is currently being validated in SVISION’s R&D projects and will in the future be deployed at scale in Aivia, AiviaWeb and AiviaCloud.

Automated image analysis at scale is a key requirement for many applications in life sciences and bio pharma. A solution to this challenge will be via AI-boosted computer vision. Learning from individual users or user groups and leveraging that learning to create general purpose AI solutions for insight creation is part of our vision for the future of image analysis. The two new patents further strengthen our IP portfolio in this space and show our commitment to continuous innovation. Moreover, they are part of the required building blocks to create our vision. Since the launch of Aivia, 4 years ago, its users have been among the first to directly benefit from AI-enhanced image analysis and insight creation. While we continue to strive for greater AI-powered automation for image analysis, Aivia users can already experience parameter-free image segmentation as well as a wide range of other machine learning solutions and pre-trained models for image restoration, segmentation, and virtual staining.

Dr. Luciano Lucas, SVISION’s Executive Vice President, Director of Aivia Development and co-inventor of US Patent 10,769,432.

"In order to accelerate the adoption of AI-enabled computer vision solutions, we continue to focus on removing the hurdles of AI deployment. The 10,769,432 patent broadly covers our user-friendly interface enabled by smart parameter-free, auto-tuning machine learning invention. The 10,891,523 patent has broad claims supporting optimal learning of deep models with only small amounts of ground truth. This is a significant step forward to address users need for efficient training, applying, and validating deep models. SVision is committed to provide the best AI-enabled computer vision solutions to users through our Aivia platform. We will continue to survey the fast-evolving machine learning field and are ready to adapt proven models into Aivia through our inventions that remove barriers to adoption."

Dr. James Lee, SVISION’s Founder, President and CEO, an inventor of 91 issued US patents and co-inventor of both new patents.


SVISION works with scientists and engineers at the technological frontier, and pioneers image-based decision technologies that propel major breakthroughs in the life science, electronics and materials industries. SVISION is a technological innovator with 37 issued US patents, and commercial interests in X-ray inspection, survey, search / alignment, video inspection and life sciences. SVISION makes and markets Aivia microscopy image analysis software. Aivia development is partially funded by the National Institutes of Health (NIH) under multiple Small Business Innovative Research (SBIR) programs worth over $16 M. For more information, visit

Media contact:

Luciano Lucas, PHD

+1 425 773 1548