AI Trends Weekly Brief


AI in Healthcare Advancing on Multiple Medical Fronts

Healthcare will be the lead industrial area of the Fourth Industrial Revolution, characterized by a range of new technologies fusing the physical, digital and biological worlds, impacting all disciplines, economies and industries. One of the major catalysts for change is artificial intelligence, says Dr. Bertalan Mesko, the Medical Futurist, with a PhD in genomics and an Amazon Top 100 author.

Healthcare is contributing to the growth of digital data, which in 2013 encompassed 4.4 zettabytes. By 2020, the digital universe of data we create and copy annually, will reach 44 zettabytes, or 44 trillion gigabytes, Dr. Mesko reports. “The world of Big Data is so huge that we need artificial intelligence to be able to keep track of it,” he states in a post entitled, AI Will Redesign Healthcare.

The number of startups in the healthcare AI market is increasing. Deals to healthcare-focused AI startups went from less than 20 in 2012 to nearly 70 in 2016, according to CB Insights, the data-focused research firm. The firm tracks 106 AI startups, as of September 2017, a list not meant to be exhaustive.

The major categories are: imaging and diagnostics, patient data and risk analytics, remote patient monitoring, drug discovery, oncology and lifestyle management and monitoring.

Hospitals are rapidly adopting AI. In a survey by Healthcare IT News and HIMSS Analytics answered by 85 respondents, 35 percent of healthcare organizations reported they will leverage AI within two years. The greatest areas of initial impact were seen to be: clinical decision support, population health, patient diagnosis, precision medicine and hospital/physician workflow.

“The problem we are trying to solve is one of productivity,” stated Andy Slavitt, the former acting administrator of the Centers for Medicare and Medicaid Services, stated during the Light Forum conference attended by CB Insights. The event brought together CEOs, healthcare IT experts, policymakers and physicians at Stanford University. “We need to be taking care of more people with less resources, but if we chase too many problems and business models or try to invent new gadgets, that’s not going to change productivity. That’s where data and machine learning capabilities will come in.”

In Boston, Partners HealthCare has announced a 10-year collaboration with GE Healthcare to integrate deep learning technology across their network.

Google, Microsoft, Apple and IBM are among the big companies with efforts underway to exploit the market for AI in healthcare. It’s a gold rush.

Healthcare has consistently been the top industry for AI investments, according to CB Insights. Deals to AI in healthcare startups increased 29 percent year-over-year to hit 88 deals in 2016 (startups plus later rounds), and was on track through two quarters in 2017 to surpass that and reach a six-year-high. In Q2 2017, 29 deals were struck, a record number for the sector, three more than the previous record.

A race is on to assemble bigger and higher quality healthcare-related databases. Some say IBM Watson Health has the advantage, with Watson launched in 2011 and trained on healthcare. Others are skeptical of Watson. Social Capital CEO Chamath Palihapitiya at LightForum called Watson “a joke.” Still, IBM continues to sign up partners, including MAP Health Management, to bring their machine learning capabilities to substance abuse disorder treatment, as reported in mobihealthnews. IBM is also working with Sutter Health to develop methods to predict heart failure based on electronic health record (EHR) data.

IBM took umbrage and issued a statement the day after Palihapitiya’s comment was aired. It said in part, “Watson is not a consumer gadget but the AI platform for real business. It has been trained on six types of cancers with plans to add eight more this year. Does any serious person consider saving lives, enhancing customer service and driving business innovation a joke?”

AI in healthcare is in an early stage. “Where we are today is ground zero basically,” stated Roam Analytics CEO and cofounder Alex Turkeltaub at Light Forum. “We’re more or less figuring out the commercial pathway.”


Data Science Advances Boost Machine Learning in Healthcare

A new crop of data science breakthroughs are advancing the use of machine learning in healthcare, with new tools becoming available that use natural language processing, pattern recognition, and deep learning to support better care.

Top initiatives and projects were outlined in a recent report in HealthIT Analytics. At Indiana University-Purdue University Indianapolis, researchers are turning machine learning algorithms loose on pathology slides to predict relapse rates for acute myelogenous leukemia. In a small study published earlier this year, one algorithm was able to identify patients who would relapse with 100 percent accuracy.

And at Stanford University, machine learning tools performed better than human pathologists when distinguishing between two types of lung cancer.  The computer also bested its human counterparts at predicting patient survival times.

The ability to extract meaning from large volumes of free text is also critical for clinical decision support and predictive analytics — another area where machine learning is starting to shine.

Identifying and addressing risks quickly can significantly improve outcomes for patients with any number of serious conditions, both clinical and behavioral.

At Beacon Health Options, a behavioral health management company, machine learning can clarify a fuzzy diagnosis process and help forestall mounting complications in complex patients.

“We sit on an awful lot of data, which is organized in a very traditional claims-based model,” said Dr. Emma Stanton, Associate Chief Medical Officer for Beacon Health Options. “We can see whether someone has had an outpatient appointment or an inpatient admission, but the data doesn’t tell us a great deal about whether the patient has actually gotten better as a result of accessing that care.”

“So while we are a data-driven company and rely on that information for everything we do, we are keenly aware that there are limitations in our insights due to the way that data is organized and analyzed.  There is a tremendously exciting opportunity to use machine learning to improve those processes and dig deeper into that data and all the other variables that impact an individual’s life.”

Beacon Health Options is using machine learning to fine-tune its risk stratification capabilities, allowing case managers to reach out to high-risk patients more proactively and better coordinate their care.

Here we look at selection of startups in AI and healthcare to gain some insights.


Atomwise Applying Deep Learning to Drug Discovery

Developing pharmaceuticals through clinical trials sometimes take more than a decade and costs billions of dollars. Speeding this up and making more cost-effective would have an enormous effect on today’s healthcare and how innovations reach everyday medicine, as reported by the Medical Futurist.

Atomwise uses supercomputers that root out therapies from a database of molecular structures. Last year, Atomwise launched a virtual search for safe, existing medicines that could be redesigned to treat the Ebola virus. They found two drugs predicted by the company’s AI technology which may significantly reduce Ebola infectivity. This analysis, which typically would have taken months or years, was completed in less than one day. “If we can fight back deadly viruses months or years faster that represents tens of thousands of lives,” said Alexander Levy, COO of Atomwise.“Imagine how many people might survive the next pandemic because a technology like Atomwise exists.”

Atomwise recently announced a program to spur the discovery of new drugs by university scientists. The firm is seeking proposals from scientists to receive 72 potential medicines each, generated specifically for their research by Atomwise using AI. It is called the Artificial Intelligence Molecular Screen (AIMS) program.

“It’s this easy: researchers tell us the disease and protein to target, we screen millions of molecules for them, and then they receive 72 custom-chosen compounds ready for testing,” said Dr. Han Lim, MD, PhD, Atomwise’s Academic Partnerships Executive. “As a former UC Berkeley principal investigator, I helped design the kind of program I wish existed for my own work.”

Recipients will be announced three months from the submission deadline in this first of its kind program. Atomwise raised $6 million in a seed round in 2015.

Learn more at models cell biology to speed cancer treatment

A dedicated team of AI developers, medical professionals and bioinformaticians at has spent six years researching and building an AI solution to design personalized treatments for any cancer type or patient faster than any traditional healthcare service, as summarized by The Medical Futurist. The technology models cell biology on the molecular level and can run millions of simulated experiments each day. It can identify the best drug to target a specific tumor, and identify complex biomarkers and design combination therapies.

The key to Turbine’s uniqueness is its molecular model of cancer biology, which it uses to run scores of simulated experiments, guided by AI to identify the biomarkers that signal sensitivity to treatment. As a result, the technology is already used in collaborations with Bayer, the University of Cambridge and top Hungarian research groups to find new cancer cures, speed up the time to market, and save the lives of patients suffering from currently incurable forms of the lethal disease.

The firm’s website states: “Discovery of new cancer drugs is urgently needed, yet going too slow. It’s time to make it as efficient as it should be. The current model of cancer drug discovery is based on decades-old methods. Due to the lack of a reliable tool that predicts a compound’s mechanism of action, researchers have to run multiple rounds of laboratory experiments to make predictions about a drug’s effects. Turbine can add valuable insights into how a drug works from the preclinical phase through phases I and II to reach phase III trials faster, and with a reduced failure rate.”

Learn more at


CareSkore Tries to Reduce Hospital Readmissions

CareSkore provides personalized population health management, leveraging machine learning to generate real-time predictive and prescriptive analytics to provide and end-to-end patient management platform. The AI-enhance post-discharge engagement reduces risk of patient behaviors that can lead to poorer outcomes.

Through its Zeus algorithm, CareSkore predicts based on a combination of clinical, labs, demographic and behavioral data, how likely it is that a patient will be readmitted to a hospital. Patients should be a more clear picture about their health. CareSkore raised $4.3 million in a seed round in August 2016.

CareSkore released its Population Health Management as a Service tool in February 2017, providing access to its Zeus analytics engine to third party applications through an API.

“We have too many proprietary application interfaces in healthcare. PHMaaS allows us to provide insights within existing applications that are already part of user workflows,” stated Jaspinder Grewal, CEO of CareSkore, in a press release.

Learn more at CareSkore.


Oncora Medical Seeks to Improve Radiation Therapy

Oncora Medical is a software startup in Philadelphia that has developed a predictive analytics platform for oncology, capable of analyzing diverse healthcare data and applying advanced machine learning techniques to product clinical insights.

Oncora is working to improve the radiation therapy experience for physicians and patients, with software to help oncologists design more personalized radiation treatment plans.

Cofounder David Lindsay was doing clinical work as an MD/PhD student at the University of Pennsylvania, when he recognized that radiation oncologists had no integrated digital database that collected and organized medical records.

Oncora has raised $2.6 million in four rounds from investors. In 2017, it plans to roll out its Precision Radiation Oncology platform to three major medical centers to help 10,000 patients receive personalized treatments.

Oncora announced in April 2017 a strategic alliance with the University of Texas MD Anderson Cancer Center to focus on building the next generation of precision medicine software for radiation oncology.

The Oncora website states, “We are a motivated group of data scientists, clinicians, machine learning experts and healthcare software developers intensely focused on improving the quality of radiation therapy treatments.”

Learn more at Oncora Medical.


Freenome Detecting Cancer Using DNA

Freenome is a two-year-old liquid biopsy diagnosis platform that detects the cell-free DNA sequencing of cancer. The firm raised $65 million in a Series A round in Q1 2017, led by Andreessen Horowitz.

Liquid biopsy companies are competing to identify cancer by relying on a patient’s DNA, rather than having to extract tissue. The medical test companies, many emerged from universities, have not yet been able to pinpoint exactly where a cancer is growing, how serious a threat it is, and whether it will respond to treatment, according to an account in TechCrunch.

AI including machine learning is being used to help analyze the DNA data. The 25-person Freenome has now tested thousands of blood samples and says its tests outperform current screening tests on the market for four types of cancer: prostate, breast, colorectal and lung.

Now with the money raised Freenome plans to expand into clinical trials, with the help of 25 research partners including the University of California, San Diego and UCSF. The company wants to be able to answer questions about each cancer.

With lung cancer, for example, it wants to be able to resolve whether a test shows a specific type of carcinoma that is more responsive to chemotherapy. The machine learning component “can create a screening test in software and learn from its mistakes and answer not just a single question, but multiple questions,” CEO Gabe Otte told TechCrunch.

The goal is to test up to 10,000 blood samples in its own and partner labs within 12 months. After that, it will move on to talks with regulators for needed approvals.


Flatiron Health Measuring Data on What Doctors Prescribe

Flatiron Health is a technology and services company focused on accelerating cancer research and improving patient care. The company’s platform enables cancer researchers and care providers to learn from the experience of patients. Flatiron partners with over 265 community care clinics, three major academic research centers and 12 of the 13 top therapeutic oncology companies.

Flatiron has raised $313 million in three rounds since its founding in June 2012.

Flatiron founders Zach Weinberg and Nat Turner began in the advertising technology market, and have no prior experience in the world of medicine, stated an account in Fast Company. “We felt very early on that providing a clinical context to the engineering and product teams as they were building things was important,” Turner stated.

The company currently employs 25 people with medical degrees of some kind, and 104 engineers and technology specialists, who collaborate closely in the development of the company’s products.

According to an account in a June 2017 edition of The Cancer Letter, Flatiron made available data gleaned data from electronic health records of nearly 35,000 cancer patients, making it possible to correlate the evolution of a drug’s approval with real-world prescribing patterns.

The data showed that physicians rely on clinical trial results immediately after they are presented or published, well before FDA approval. No one had previously aggregated this information in real-time, according to the Cancer Letter.

Maria Koehler, vice president of oncology strategy, innovation and collaborations at Pfizer Oncology, said the data from the bioinformatics company was valuable. “It shows how the market changes and the speed of the change. This is tremendously important,” she stated. “It shows which other drugs are replaced in the therapeutic setting by the introduction of the new drugs.”

She added, “I think that this will revolutionize not only how regulators look at the world, but also how payers look at the world, because we never had this type of data.”

The Flatiron datasets are a result of the company’s collaboration with FDA to develop a framework for using real-world evidence to inform the agency’s regulatory decisions. The mandate was folded into the 21st Century Cures Act, a wide-ranging health care reform bill passed by Congress in December 2016. Flatiron is among the first to provide the agency with rich data.

Learn more at Flatiron Health.