Infertility is a growing global concern, affecting countless millions of would-be parents. Yet the success rate of today’s in vitro fertilization (IVF) technology is only about 20% per IVF cycle – a process that is traumatic, painful and expensive. To improve those outcomes and help couples successfully conceive, it is essential to accurately determine the viability of each embryo – the higher the accuracy of doing so, the more likely the embryo will result in pregnancy.
This process of embryo identification is being enhanced by AI fertility startup Presagen, an independent software vendor (ISV) and member of the Intel® AI Builders program, through the use of deep learning and computer vision – technologies that are perfectly placed to address this challenge. Traditionally, embryologists in the IVF laboratory make their own visual judgment of the best embryo to implant; in contrast, using state-of-the-art deep learning techniques for the assessment of embryos have shown improvements of over 25%1.
Presagen’s Life Whisperer application brings the benefits of artificial intelligence (AI) to fertility-challenged couples in the easiest possible way, while addressing a growing problem in the medical industry. Using deep learning and computer vision, Life Whisperer identifies the morphological features that constitute a healthy embryo, which are often invisible to the human eye. Clinical studies in multiple IVF facilities bear witness to Life Whisperer’s effectiveness in delivering a marked improvement beyond traditional manual methods alone.
IVF clinicians and embryologists in labs and hospitals can access Life Whisperer via a simple, intuitive web interface that is accessible from anywhere and features drag-and-drop capability for loading microscope images and generating results. With little training required, Presagen hopes to make Life Whisperer widely available using a scalable and on-demand cloud-based application.
An AI Solution for Diverse IVF Environments
Life Whisperer’s approach is unique in that it looks beyond improving accuracies on internal or similar medical image data sets – it addresses the problem of translatability, or how the model is able to handle new, unseen data sets of embryo images. Developers at Presagen sourced many highly diverse images across different camera configurations, settings, and clinical practices. They were able to construct an AI model that could generalize for unfamiliar clinics with different IVF work practices, located in different countries and demographics. In addition, Life Whisperer can be fine-tuned to work across multiple camera setups. As a result, the healthcare industry can embrace AI technologies with greater confidence, knowing that careful attention during model training, validation and testing can produce robust, highly accurate imaging – often determining viability in seconds, including the time to upload the image file using the simple drag-and-drop interface.
How Life Whisperer Applies Intel® AI Technologies
Intel customized a deep learning framework that was used in the development of Life Whisperer’s AI model. The Intel® Optimization of PyTorch*, a python-based deep learning framework for AI models, was shown to work optimally on Intel® Xeon® Scalable processors versus the generic PyTorch distribution of this library. The Intel AI Builders engineering team helped install and correctly run the library, and assisted in adapting files for Docker* containers, which helped make AI model server management more modular and efficient.
For the Life Whisperer application, the primary benefit from the software optimizations lies in the speed improvement in inference time. The AI algorithms that form the core of this healthy embryo identification solution run optimally on Intel machines and when using Intel® AI technology.
Presagen’s affiliation with Intel has enabled the startup to broker new contacts with other technology allies through events such as Intel® Partner Connect for such services as public relations, branding, and financial relationships with Intel Capital.
Plans for the Future
Life Whisperer’s developers have completed most of their AI refinement for their initial product rollout, completing and field-testing its web application. Starting in Australia, they’re already making the technology available via web browser and secure login. The goal is to push Life Whisperer worldwide, and to that end developers are actively seeking regulatory approval in all four targeted countries. Going forward, Presagen’s challenge will be to make their AI technology-driven product for the healthcare industry available as a scalable business, serving patients in the easiest way possible.
Presagen’s AI platform brings together deep learning and computer vision to analyze large data-sets of medical images to rapidly create web-based medical diagnostic tools that can be used on-demand by medical institutes anywhere in the world.
Starting with medical images, Presagen’s platform automates the manual tasks of AI training, with access to a large burst computing infrastructure to reduce the time and cost of AI software development.
Presagen’s Life Whisperer product uses AI to better identify viable embryos during IVF to improve pregnancy success rates, assisting fertility-challenged couples.
To learn more, visit: https://www.lifewhisperer.co/
Listen to the Intel on AI podcast featuring Dr. Jonathan Hall, scientific lead at Presagen: https://builders.intel.com/ai/social-hub/podcast/ai-enhanced-fertility-presagen-intel
1According to Presagen’s 12-clinic study conducted across four countries: the U.S., Australia, New Zealand and Malaysia.