Cancer diagnosis plays a crucial role in determining a patient’s path forward. However, the traditional methods used by pathologists, such as examining tissue samples under a microscope, have not evolved significantly in the last 150 years. This lack of innovation can lead to missed diagnoses and dire consequences for patients. To address this issue, Paige, a technology disruptor in healthcare, has partnered with Microsoft to develop the world’s largest image-based artificial intelligence (AI) model for digital pathology and oncology. This collaboration aims to revolutionize cancer diagnosis and patient care.
The Need for Innovation in Pathology
Pathologists are the key players in diagnosing cancer. As Thomas Fuchs, co-founder and chief scientist at Paige, states, “You don’t have cancer until the pathologist says so.” Despite their essential role in medicine, the workflow of pathologists has remained largely unchanged for decades. They rely on examining tissue samples on glass slides under a microscope, a method that is tried and true but prone to human error. The consequences of missed diagnoses can be severe, making it crucial to improve accuracy and efficiency within the specialty.
Digitizing Pathologists’ Workflow with AI
Paige specializes in developing digital and AI-powered solutions for pathologists. Their aim is to transform the pathologists’ workflow and improve diagnostic accuracy. One of their innovations is the FullFocus viewing tool, which allows pathologists to examine scanned digital slides on a screen instead of relying solely on a microscope. This tool has received approval from the Food and Drug Administration (FDA), marking a significant milestone in the adoption of digital pathology.
Additionally, Paige has built an AI model that can help pathologists identify breast cancer, colon cancer, and prostate cancer when they appear on the screen. This AI model is trained on an unprecedented amount of data, including billions of images, enabling it to identify both common and rare cancers that are notoriously difficult to diagnose. The goal is to assist doctors who are struggling with staffing shortages and increasing caseloads, ultimately improving patient outcomes.
Overcoming Challenges in Data Storage and Collection
Digitizing pathology slides requires substantial storage capacity. Storing a single slide can require over a gigabyte of storage, making large-scale data collection a costly endeavor. This has historically limited the adoption of digital pathology to wealthy academic centers that can afford the infrastructure and associated expenses. However, Paige, with its wealth of data from its origins at the Memorial Sloan Kettering Cancer Center, has been able to build its own AI-powered solutions.
Despite these advantages, the storage costs remain a barrier for smaller health systems. To address this challenge and expand its operations, Paige partnered with Microsoft. Microsoft’s cloud storage and supercomputing infrastructure provide Paige with the resources needed to build an advanced new AI model. This partnership allows Paige to overcome the storage and computational challenges associated with large-scale data collection.
Building the World’s Largest Image-Based AI Model
Paige’s original AI model, developed using more than 1 billion images from 500,000 pathology slides, was already groundbreaking. However, the collaboration with Microsoft takes it to the next level. The new AI model being built with Microsoft is “orders of magnitude larger than anything out there,” according to Fuchs. It is training on a staggering 4 million slides, encompassing both common and rare cancers.
The scale of this project is unprecedented. In fact, the data used by Paige and Microsoft for this model is 10 times larger than all the content on Netflix. This massive AI model will not only be the largest oncology model to date but also the largest publicly announced computer vision model. The goal is to capture the subtle complexities of cancer and provide a foundation for the development of clinical applications and computational biomarkers.
Advancing Cancer Diagnosis and Patient Care
The collaboration between Paige and Microsoft holds significant promise for improving cancer diagnosis and patient care. By leveraging Microsoft’s expertise and advanced computing infrastructure, Paige can train its AI model at an unprecedented scale. The large volume of digitized microscopy slides from the petabyte-scale clinical data archive will enable the model to capture the intricacies of cancer morphology with extraordinary fidelity.
The benefits of this AI model extend beyond accurate cancer detection. It has the potential to enhance our understanding of the molecular and genetic aspects of cancer, leading to more precise diagnosis and treatment plans. The model can also serve as a foundation for rare cancer models, mutation prediction, and response prediction. Ultimately, this technology aims to reduce misdiagnosis rates and improve patient outcomes.
Democratizing Access to Cancer Care
The impact of this AI model will be particularly significant in community clinics and smaller health systems, where pathologists are overwhelmed with numerous cases across all cancer types. By automating and streamlining the pathologists’ workflow, the AI model can alleviate the burden on healthcare professionals and democratize access to quality cancer care.
A Collaborative Journey Towards Transforming Cancer Care
Paige and Microsoft’s collaboration marks a groundbreaking step in the field of oncology. The companies intend to publish a paper on the model through Cornell University’s preprint server arXiv, quantifying its impact and outperforming existing academic models. However, this is just the beginning of a longer journey. The model will undergo thorough testing and collaboration with regulators to ensure its safety and accuracy before it can be deployed as a product.
The future of cancer care is being shaped by innovative partnerships like the one between Paige and Microsoft. With the development of the world’s largest image-based AI model, the potential for improving cancer diagnosis and patient care is immense. By harnessing the power of AI and advanced computing, we are moving closer to a future where cancer can be detected and treated with greater precision and efficiency.
and then 