Remote-First-Company/NEW YORK – December 14, 2022 – VAST Data, the data platform company for the AI-powered world, announced today that Therapixel, a French company specializing in the detection of breast cancer using artificial intelligence (AI), has selected VAST’s high-capacity all-flash data platform to accelerate the training of their AI algorithms against more than a million mammograms.
With VAST, Therapixel now has a high-performance, scalable data platform to train its algorithms using a linearly-scalable all-flash system that delivers the style of random read performance required to accelerate the detection of breast cancer mutations. Therapixel’s selection of VAST has eliminated data capacity and performance bottlenecks and the solution is a platform that can now easily scale as Therapixel’s business grows.
VAST Data is a game-changer, providing the simplicity, scalability, and storage space we need to run our training algorithms. Before implementing VAST Data, we were leaning on slower HDD storage solutions that had insufficient storage space for the data needed to train the algorithms that solve breast cancer. And you cannot develop any AI solution without fast access to data,” explained Aurélien Chick, Data Manager at Therapixel. “Today, this is no longer an issue, thanks to VAST Data.
VAST’s all-flash storage gives Therapixel faster access to train and iterate better for refinement of the algorithm to interpret mammograms, enabling highly reliable breast cancer detection at first reading. VAST’s Similarity-based data reduction allowed Therapixel to double its data storage capacity and get even more space than it had imagined, enabling the company to collect more data from partners to train its models to a higher level of accuracy by exposing them to more sample data. Since deploying VAST, Therapixel now has the capacity to train on over 1 million mammograms.
Focusing on breast cancer detection with AI-based tools, Therapixel uses its knowledge in AI to enable radiologists to make better and quicker decisions. The company’s software points out only the most unusual lesions and also provides characterizations that help radiologists determine whether they should spend more time on a particular case. Until recently, many AI-based solutions for medicine imagery pointed out every single lesion, delivering a diagnosis without any additional information and creating fatigue for radiologists. Therapixel lets radiologists see in real-time what the AI-trained algorithm predicts.
In the same way that we enable and empower our radiologists to be their best and to focus on what matters, VAST Data has allowed us to focus on our research and development. It’s a good synergy,” added Chick. “We are already achieving great things with VAST Data on a system we deployed only a few months back. It’s very encouraging. We can develop ideas and test them quickly and confirm those ideas with more data to train the algorithms.