Industry leading news about artificial intelligence in healthcare and diagnostics.

Tech Advances Help AI in Oncology, but Dirty Data Hampers Progress

As artificial intelligence (AI) in healthcare accelerates, so does the understanding of precision medicine. This will lead to strong potential for AI in oncology.

“Cancer is an extremely complex ecosystem,” said Bikash Sabata, PhD, Vice President of Software, Instrument Platform at Roche Sequencing Solutions based in Pleasanton, Calif. “And within that ecosystem, trying to go down to every part of the biology and model every part of it is going to be extremely hard. But at a system level, being able to modulate and being able to predict is something that AI plays a critical role in for all our solutions these days.”

Sabata will explain how AI is influencing next-generation sequencing (NGS) at the Artificial Intelligence in Healthcare and Diagnostics (AIHD) Conference, which occurs May 10-11 in San Jose, Calif.

Sabata_BikashSabata, along with former Food and Drug Administration Commissioner Scott Gottlieb, MD, Artiman Ventures Managing Director Ajit Singh, PhD, and other industry experts will share how healthcare leaders can best collaborate with AI developers to improve patient outcomes.

Newer AI tools are helping clinicians derive precision medicine insight from NGS, a technology that allows for the rapid sequencing of the human genome. Just prior to the last decade, earlier AI approaches might not have been up to the task.

“Even though we had AI, you would call it a previous generation of AI,” said Sabata, whose background includes innovating in clinical laboratories, startups, and Fortune 500 companies. “It involved trying to understand the biological basis of the signal and trying to model that through some statistical techniques, but that was not sufficient and just did not scale.”

For more than 15 years, Sabata’s focus has been on different aspects of diagnostics, especially as it relates to personalization and oncology. In that regard, newer technologies for AI in oncology that have only recently become available offer far greater benefits.

“Once we started getting into the newer AI technologies, like deep learning and newer networks to train [algorithms] with extremely large volumes of data, our capabilities significantly expanded,” Sabata explained. “It's just not possible for us to come up with these kinds of insights without being able to use these kinds of AI technologies.”

AI in Oncology Can Differentiate Cancer Types

Sabata detailed some of the key clinical applications that he sees the greatest potential for with new AI technologies.

“We are looking at some of these really complex diseases in oncology, teasing out the differences between the different conditions,” he said. “We can determine, for example, what a particular type of breast cancer is and can therefore find the right therapeutic pathway to help the patient. These are critical areas to be in, and AI is making a lot of difference.”

AI plays an increasingly important role in not only distinguishing between complex diseases, but also in providing earlier detection of these illnesses. Further, AI will eventually provide a better way of tracking changes in chronic, gradual diseases that develop slowly.

“How do you detect changes that are happening at a really tiny magnitude?” Sabata asked. “That is one very hard problem that I think AI is going to be able to address well.”

Dirty Data Hamstrings AI at Scale

While AI technology has seen many improvements, dirty data in healthcare remains a related problem that needs to be addressed, Sabata warned.

“Data is the biggest challenge and will continue to be for the next decade or so,” Sabata said. “We are always stumbling into data that is not clean or data where the quality is extremely questionable. The more you collect, the less quality you typically have.”

Sabata sees this as ultimately being a challenge with the way data is collected and curated.

“It is the whole process from where the sensor collects the information to the point where the AI algorithm can actually make sense of the data,” he observed. “There are so many places we stumble. And that becomes a problem because we have to do it at this massive scale.”

—Caleb Williams

Related Resources:

Artificial Intelligence in Healthcare and Diagnostics Conference

Former FDA Head Scott Gottlieb to Headline Upcoming AIHD Conference

American Cancer Society: Precision Medicine

Researchers in Five Countries Use AI, Deep Learning to Analyze and Monitor the Quality of Donated Red Blood Cells Stored for Transfusions

Pregnant men and smoking babies? 4 reasons dirty data destroys healthcare analytics