9 key thoughts on how machine learning and deep learning will affect healthcare

Artificial intelligence is becoming more important in the healthcare space. Data gathering for machine learning and deep learning capabilities have immense possibilities to improve diagnostics, care pathway creation and reproducibility in surgical procedures to ultimately achieve better clinical outcomes.

The technology can also assist physicians with generating reports and administrative responsibilities, giving them more time to spend with patients.

Here, nine clinical care and health IT company executives discuss how they expect machine learning and deep learning to improve healthcare in the future.

Waqaas Al-Siddiq. Founder and CEO of Biotricity (Redwood City, Calif.). "Deep learning can impact wearables focused on specific conditions, like remote cardiac monitoring, at an individual level by indicating how to personalize algorithms according to one's particular biometric and patient data. The incorporation of machine learning can assist in the interpretations of the analysis of the unstructured data delivered from these medical-grade wearable devices. The initial analysis is typically provided by mathematical algorithms trained to detect anomalies in this data. Machine learning, combined with artificial intelligence, would then seek to perform an interpretation of such a report, just as a physician would, in order to save physician time. Such capabilities effectively reduce physician time, enabling them to focus on the most critical patients and streamline the care process."

Roman Franklin and Mitali Maheshwari. IT Analysts at MD Buyline (Dallas). "Artificial intelligence in the short term for healthcare will bring about transparency and empowerment of decision-making. The ability to gather and decipher vast amounts of data [and] then provide a statistically driven answer will completely change the landscape for patients, physicians and hospitals.

Machine learning can work to recognize trends in data sets, such as patient examination files, which may help a physician recognize the early signs of cancer or a potential risk based on other patient data, family history, habits, etc. Awareness and early diagnosis can help save millions of lives.

In the purchasing space, AI will bring transparency of cost and spending trends to help create a tailored supply chain experience, like what Amazon has done for the general consumer. Imagine a completely automated purchasing department that receives the preferred equipment at the best price and at the most efficient time.

Thirty years from now, we could see the majority of patients never leaving their house for treatment. We could see robotic or software programs that provide remote patient monitoring and care. We could see everyone equipped with wearable technology that measures vital health parameters and delivers an instant feedback report. This device could research a database filled with millions of predictive health data points and spit back a life plan to follow: what to eat, how to dress, when to exercise. Furthermore, these devices could connect to the nervous system and transmit neuro signals to modify blood flow, reduce stress [or] even slow the aging process. Fundamentally, there is no limit to the capabilities of a technology that is constantly learning from itself and making improvements."

Manan Goel. Vice President of Products at Kinetica (San Francisco). "Cognitive systems will augment humans for efficiencies across the entire healthcare value chain. Trends such as digitization, personalization and mechanization are forcing healthcare organizations to deal with a tsunami of data from sensors, machines, fitness devices, mobile apps and EHRs. The speed, size, shape of data far outpaces the ability of current systems and humans to comprehend, draw insights and act on data. Cognitive techniques such as machine learning and deep learning are needed to process structured and unstructured data at scale and automatically discover patterns and anomalies to augment human intelligence with machine intelligence and deliver value across the healthcare value chain. AI will augment — not replace — humans, including doctors, nurses and administration personnel for easier, simpler and faster diagnostics, treatment plans and patient monitoring to improve quality of care."

Ori Geva. Co-Founder and CEO of Medial EarlySign (Israel). "AI and machine learning will revolutionize healthcare as we know it. Recent technological developments enable us to automatically apply mathematical calculations to vast amounts of medical data — taking into consideration health conditions, genetic factors and lifestyle to identify complex trends, patterns and interrelationships over time.

Coupled with the explosion of data ranging from existing and ubiquitous EMR data to the thousands of data points in microbiome and proteome — these algorithms will utterly change the way research is done, expediting some of the discovery processes. AI will empower physicians and providers by providing tools and clinical insights that facilitate personalized, predictive and proactive outreach to high-risk patients, resulting in better adapted treatments, increased quality, reduced costs and improved outcomes."

Mini Peiris. Chief Marketing Officer of Ambra Health (New York City). "There has been much talk of AI's potential to streamline operating efficiency and improve accuracy. Particularly, machine learning has shown promise in optimizing read times in radiology and other image-intensive subspecialties such as cardiology, oncology and neurosurgery. And with government programs like Cancer Moonshot calling for large data set mining, AI will become a necessity to achieve patient care goals.

In a recent Ambra Health webinar event on automated medical image analysis and AI, 50 percent of polled audience attendees believe that in three years, using deep learning in radiology could help reduce imaging errors, and 30 percent believe artificial intelligence can work to automate workflows such as patient matching. The goal is not to remove the physician from the care cycle, but rather, help them make the best possible decision for a patient. In high pressure situations, like in the [emergency room] or surgical suite, fast analyzing programs can save critical time.

But to leverage tools like deep learning, you need to have your imaging stored in one central place. The cloud's role as a secure, cost-effective data repository with anytime, anywhere access and business continuity built in, fits that need. Healthcare systems should prepare now for the AI tools of the future by beginning to move their imaging operations to the cloud. With that central archive in place, physicians and hospitals should use AI today when it works well and offer feedback on programs when there is room for improvement. There must be a constant conversation between user and technology vendor to produce the most efficient software possible."

G. Philip Reger. Chief Technology Officer of Radiology Partners (California). "Machine learning and AI-based computer tools are going to have a big impact on radiology. We don't expect something disruptive to hit within the next few years, but it is hard to imagine a scenario in which radiology is not being significantly enhanced by software using these techniques 10 years from now.

Many are concerned that AI will eventually replace human radiologists, but we don't think that is anywhere close to likely. Instead, we expect AI to play a role in making it possible for radiologists to keep up with ever-increasing workloads and demands that they manage those workloads, all while providing high-quality reads.

I believe there are two ways that we will see this happen: The first is that AI techniques will evolve to the point that we're able to create a 'virtual resident' or 'rad assistant' that looks at radiological study along with the radiologist. The software, based on learning the specific preferences and idiosyncrasies of that particular physician, will surface findings and identify errors that would otherwise be missed. And by applying AI techniques to natural language processing, the computer will be able to read medical histories buried inside of EMRs as well as generate output that can go directly into the radiologist report."

Jas Grewal. CEO of CareSkore (Mountain View, Calif.). "AI will be taking on a huge role in healthcare over the next few decades. This will be driven by two things: shortage of skilled providers and advances in technology. AI is already being deployed to handle more and more of the 'busy' work that health professionals have to process, particularly in population health: discharged patient pain assessments, medication check-ins [and] appointment confirmations/modifications.

Soon they will be doing initial patient intake, asking for symptoms, occurrence, intensity, frequency, etc., to prep the doctor for the patient encounter. Then it will play a greater role in care coordination across the care team, assembly and processing results of all [of] the patient's team encounters, and accessing machine learning-driven analytics to compare observed data to patient risk assessments and direct next steps. This is all part of the administrative oversight of the patient, which takes up to 50 percent of a provider's time."

Darren Schulte, MD. CEO of Apixio (San Mateo, Calif.). "In the next 20 to 30 years we'll see AI and machine learning transform the practice of medicine from an art to a science. We'll see AI assist healthcare providers, and, in some instances, replace specialties such as general radiologists or pathologists. We can reduce care variation and inappropriate care. AI will provide a much better understanding of underlying individual risk to improve healthcare payment and financing.

Currently, U.S. healthcare occupies the worst of many worlds. Most care decisions made are based upon flimsy evidence that is not specific to individuals. Only a minority of all clinical care guidelines are from randomized clinical trials — the gold standard of evidence. Studies suggest that a third of the money spent on healthcare is wasted, inappropriate or harmful. It is estimated that nearly 100,000 people die each year from medical errors. We are not using knowledge from real-world care described in healthcare records to improve care even though about 75 percent of all physicians use electronic medical records. And we create insurance models and benefit designs based upon inaccurate pictures of risk.

By aggregating medical records and using computers to decipher the contents of these records, we can assemble individual portraits of patient care. Machines will be able to sift through millions of records to learn what treatments work in similar patient profiles. Physicians can then be matched with patients based on their past success in treating similar patient profiles.

Robots will become sophisticated enough to assist physicians with procedures, surgeries, diagnosis and treatment. AI-assisted drug development can result in compounds tailored for specific patient genetic, proteomic and phenotypic profiles. Furthermore, insurance products can also be fit for individuals using knowledge about predictions of future risk and what works in similar people. Think of this as a plan dynamically designed for each person based upon value and outcomes."

Copyright © 2024 Becker's Healthcare. All Rights Reserved. Privacy Policy. Cookie Policy. Linking and Reprinting Policy.

 

Featured Whitepapers

Featured Webinars