{"id":254,"date":"2022-03-10T18:17:03","date_gmt":"2022-03-10T18:17:03","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/startups\/blog\/?p=254"},"modified":"2024-10-15T01:46:34","modified_gmt":"2024-10-15T09:46:34","slug":"pangaea","status":"publish","type":"post","link":"https:\/\/www.microsoft.com\/en-us\/startups\/blog\/pangaea\/","title":{"rendered":"How Pangaea Data uses Azure for a medical AI tool that improves patient outcomes"},"content":{"rendered":"\n

Starting a company is never easy. But it can be especially painful in the healthcare industry. There are numerous regulations, HIPAA compliance requirements, and strict rules regarding patient confidentiality that you need to navigate. These factors, while important, make it extremely difficult to introduce new technology into hospitals and healthcare companies.<\/p>\n\n\n\n

However, these challenges don\u2019t mean that entrepreneurs shouldn\u2019t\u2014or can\u2019t\u2014pursue this path. Many startups have succeeded in changing healthcare through innovation, and Pangaea Data<\/span><\/a><\/span> (Pangaea) is one of them.<\/span><\/p>\n\n\n\n

Pangaea Data, founded in 2018, is a life sciences technology business that provides PIES (Pangaea\u2019s Intelligence Extraction and Summarization), which is driven by novel unsupervised artificial intelligence (AI) to extract and summarize intelligence from patient records in a federated privacy preserving manner. It then uses this intelligence to characterize hard to diagnose conditions and make predictions that can help catch undiagnosed or miscoded patients.<\/p>\n\n\n\n

This is no small task, though, and it\u2019s compounded by the fact that there\u2019s an underlying distrust of AI in the healthcare field due to earlier AI technologies that failed to deliver on their promises.  <\/span>Pangaea Data has taken approach to using AI in healthcare: Instead of trying to replace human doctors and nurses with computers, they\u2019re focusing on helping them do their jobs better by providing insights derived from big data sets. By combining the best aspects of clinical medicine and cutting-edge computer science, they\u2019ve developed an AI platform that can identify and diagnose diseases faster and less expensively than traditional methods. And they\u2019re doing this with Microsoft Azure<\/span><\/a><\/span>.<\/span><\/p>\n\n\n\n

To learn more, we spoke with Pangaea Data co-founder and CEO Vibhor Gupta. Let\u2019s take a closer look at what Pangaea Data is doing, and find out how they\u2019re using Azure.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

<\/p>\n\n\n\n

Proof of concept: Detecting Cachexia<\/h2>\n\n\n\n

One challenge of using AI in the medical industry comes when trying to build a system that works well even when data is complex and unstructured. To prove that applying unsupervised learning to extract actionable intelligence and insights is viable, Pangaea Data needed to work with a real-world data set to try to accurately detect a frequently undiagnosed or misdiagnosed condition. Cachexia is a perfect example of a condition that is significantly underdiagnosed.<\/p>\n\n\n\n

Cachexia is characterized by weight loss and muscle wasting. It\u2019s one of the most common symptoms of cancer and has been linked to high mortality rates in cancer patients. The problem is that patient data is messy, and it\u2019s difficult for doctors to identify the subtle indicators that a patient has begun to be affected by cachexia.<\/p>\n\n\n\n

To solve this problem, Pangaea Data decided to use cachexia detection as a proof of concept to validate their AI approach to early diagnosis. They analyzed the electronic health records of thousands of oncology patients to see if they were able to spot any hints (like specific clinical features) through their AI driven product (PIES), which might indicate cachexia earlier than conventional means.<\/p>\n\n\n\n

In a small dataset of 100 patients which Pangaea used, 19 cancer patients had previously been identified as having cachexia based on ICD codes in their patient records. Pangaea\u2019s AI driven product, PIES, correctly identified the initial 19 patients, and then found an additional 51 (268% more) cancer patients who were suffering from cachexia but were undiagnosed. Following this study, PIES was deployed on a larger dataset of circa 29,000 patients where it found 1052% more cancer patients with cachexia, who were hidden (or missed) in plain sight. These results were validated with help from clinicians.<\/p>\n\n\n\n

It\u2019s incredible \u2014 and important! Early detection of cachexia would have allowed those 51 patients to begin undergoing treatment sooner, which translates directly into saved lives and reduced healthcare costs.<\/p>\n\n\n\n

And detecting cachexia was just the beginning. With proof that its technology works, Pangaea is now able to partner with healthcare providers and pharmaceutical companies, allowing them to begin delivering real-world improvements to patient outcomes.<\/p>\n\n\n\n

Scaling with Azure<\/h2>\n\n\n\n

For startups, the ability to scale quickly and cost-effectively is critical to success. Azure offers secure, reliable, and cost-effective computing power. Building an AI system that works and improves patient outcomes is great, but it\u2019s only useful if you can reliably run and scale it.<\/p>\n\n\n\n

Pangaea Data relies on Azure virtual machines to host their servers and AI models. They use both CPU and GPU instances to run their models, using Azure Storage Accounts to store and manage data. They also rely on Azure networking tools such as VPNs and network security groups. Azure cloud management features automate the provisioning and maintenance of their servers, allowing the Pangaea team to scale up easily without worrying about running out of capacity.<\/p>\n\n\n\n

This highlights a key part of the value Azure provides younger businesses: it enables them to focus on building their product instead of spending time managing infrastructure. Such younger businesses can also use Azure\u2019s pay-as-you-go pricing model to minimize capital expenditures.<\/p>\n\n\n\n

Microsoft and Azure: Trusted by industry<\/h2>\n\n\n\n

Azure is also trusted by enterprises and healthcare providers around the world. By working with Microsoft and Azure, Pangaea Data can provide a level of built-in trust: Potential clients are reassured when they hear that Pangaea Data runs on Azure.<\/p>\n\n\n\n

Pangaea Data\u2019s clients who already use Azure know they can deploy their AI driven product (PIES) on Azure and it will work as expected and assure compliance with privacy regulations. And, clients who haven\u2019t yet moved to the cloud may already be interested in doing so because they know that moving their data to the cloud will make it easier for doctors, researchers, and scientists to collaborate in compliance with privacy regulations.<\/p>\n\n\n\n

Running on Azure has an additional benefit: It makes it easy for Pangaea to deploy its product (PIES) for end users across multiple pharmaceutical companies and healthcare providers for various diseases and purposes, including early diagnosis, patient stratification, precision medicine, drug discovery and clinical trials in a federated privacy preserving manner. This means that patients everywhere achieve better outcomes, while Pangaea Data can ensure that every organization\u2019s data remains secure and is never touched, copied or shared. In the privacy-conscious healthcare industry, this is critical to building trust and growing a client base.<\/p>\n\n\n\n

Pangaea\u2019s advice for Startups<\/h2>\n\n\n\n

Vibhor Gupta, the co-founder and CEO of Pangaea Data, has a few pieces of advice for other founders:<\/p>\n\n\n\n