NEW YORK (360Dx) – Armed with $8 million from a recent financing round — and with an eye toward another fundraising round — startup VoxelCloud expects to get regulatory clearance in China soon for its first three products, joining a growing number of healthcare companies leveraging artificial intelligence technologies for disease diagnostics.
In total, the Los Angeles-based company, founded in January 2016, has submitted five products for regulatory clearance in China and is in the midst of drafting documents for US Food and Drug Administration clearance. By year's end, VoxelCloud anticipates getting the nod from the Chinese Food and Drug Administration to market its AI-enabled diagnostic technologies for lung cancer, cardiac disease, and retinal disease, CoFounder and CEO Xiaowei Ding said.
Applications for a product for chest X-ray analysis and another version of a retinal technology, both also using AI technology, have been submitted to the CFDA with clearance possibly coming in about a year, he added.
VoxelCloud's technology is based on AI technology learning over a massive amount of what Ding called "ground truth" data, or data that's purely objective, resulting in disease models that could determine whether a patient had a certain disease.
"We understand that a medical diagnostic problem is a comprehensive data problem," Ding said. "Usually, we don't take the subjective diagnosis from a doctor because there are possibilities that a subjective diagnosis or opinion is different from the actual truth."
VoxelCloud's approach is multimodal, he said, and combines deep-learning methods with other traditional computational and machines learning methods. It also combines multiple sources of data. Based mainly on radiology imaging scans, blood test results, and DNA sequencing data, it also combines clinical data, such as patient's age, gender, disease history, and family history etc. The final result is a comprehensive portrait of the patient's health status from which a diagnosis can be better made.
For those using VoxelCloud's services, a patient's data is uploaded onto a cloud-based service where it will be analyzed by the company. For those healthcare providers that don't allow patient data to be shared outside of its premises, the firm will deploy a workstation inside the facilities' network, so that the data will remain inside its private network, Ding said.
Protective of its technology and know-how, VoxelCloud has not published data in support of its research. But according to Ding, as part of a national research collaboration in China involving 20 hospitals and analyzing data collected over a five-year period from more than 50,000 cardiac patients, VoxelCloud's technology achieved "pretty high correlation" with results from CT scans.
He added that the firm's cardiac event predictive accuracy "is pretty high, at least for future two-year or five-year probability" of a serious event. "We have over 80 percent accuracy compared to the actual event happening to the patient cohort we collected," Ding added.
The firm's lead product, though, is a tool for lung cancer screening, targeted especially for early stages of the disease. VoxelCloud is also building out the technology so that it can be used to diagnose other chest disorders, including emphysema and pulmonary embolism. In total, Ding said, the tool will be applicable for about 30 different ailments of the lung.
In leveraging its AI technology, VoxelCloud joins a growing list of healthcare firms using the technology for diagnostic purposes. They include Berg, which is using its AI platform on the US Department of Defense's prostate cancer biobank and preparing to commercialize a diagnostic test. Meanwhile, Finnish genetics testing shop Blueprint Genetics recently announced it raised €14 million.
In the spring, Fronteo Healthcare and the Japanese Foundation for Cancer research announced they would develop technology based on genomic analysis and AI to improve cancer diagnosis. And in June Sophia Genetics moved into the liquid biopsy space with an update to its AI technology.
According to Ding, while VoxelCloud may be riding the AI wave, he and the company's other founders were ahead of the curve in applying the technology to healthcare. "When we got started, not many people understood what we wanted to do," he said.
The company's other founders are Jianming Liang, an associate professor of biomedical informatics and computer science at Arizona State University, and currently vice president of R&D at VoxelCloud; and Demetri Terzopoulos, the chancellor's professor of computer science at the University of California, Los Angeles, and the chief scientist at VoxelCloud. Terzopoulos received an Oscar with John Platt in 2006 for technical achievement for their work in computer-generated methods used to simulate realistic cloth in movies.
Ding is a computer scientist by training, and before joining VoxelCloud he was a research assistant at Cedars-Sinai Medical Center.
He and his cofounders "kind of naturally evolved our research from traditional machine learning applied to medical problems to deep learning applied to medical problems, so it's a natural evolution."
Around 2012, he said, deep learning was still an emerging field that was used mostly for general computational tasks, such as image classification and speech recognition. "It hadn't been tried or used in medical scenarios. … [W]e started to think of this and trying to collect bigger datasets that could fit or satisfy different [computational] models."
While the AI healthcare field is becoming as crowded as a New York City subway during rush hour, not everyone is sold on the idea that the technology is ready to upend the diagnostics market. In a recent opinion piece in BMJ, Enrico Coiera, a professor of medical informatics at Macquarie University and director of the Centre for Health Informatics, called for new rules and regulations to govern the use of AI in medicine.
He wrote that in AI systems built around machine learning that extracts knowledge from large datasets, "If data are inaccurate or missing, then they may not be fit for the purpose of training an AI, or for decision-making.
"If data are biased, so too will be the AI’s knowledge. Hidden biases can discriminate against some patients, whether based on gender, ethnicity, or disease, because they are under-represented in the original data used to train the AI," he said.
And on a recent podcast, Atul Gawande, a professor at Harvard T.H. Chan School Public Health and a surgeon at Brigham and Women's Hospital, also voiced skepticism about the technology, saying that reaching a disease diagnosis is more complicated and subtle than just gathering and analyzing data.
"I think it is one of the hardest things," he said of the diagnostic process.
But others believe that AI will inevitably change medicine. According to Bertalan Meskó, a medical futurist who has written and lectured on digital health technologies, AI will become the stethoscope of the 21st century, but only if healthcare professionals and organizations are open to the technology and learn what it can do for their field.
"Artificial intelligence will redesign the whole process of providing care," he said in an email. "Digital health has made disruptive technologies available that provide a vast amount of data no physician can analyze anymore. We need deep-learning algorithms, artificial narrow intelligence, and supercomputers to get the most out of the data patients and physicians bring to the table."
For VoxelCloud, there is no debate about AI's future in medicine. Since opening shop it has raised a total of $13.5 million in private financing, including the recent $8 million Series A round, which was led by Sequoia Capital. The money will be used to expand the company's scope of medical imaging analysis services, to increase collaborations with institutions in the US and China, and for regulatory filings and marketing.
The firm, which also has three offices in China — Shanghai, Suzhou, and Beijing — is in the midst of another financing round that will target a "much bigger" amount than the two previous rounds, though Ding declined to provide details. He also declined to name who will lead the round, though he said it will be a VC firm, but not Sequoia.
The funding will go toward commercialization efforts, expansion of its products, collaborations, and regulatory work in the US, China, and elsewhere.