AI platforms aim to ease information overload in healthcare and improve patient care

Read about two startups that are using machine learning to analyze medical data and distill the information for healthcare professionals, ultimately leading to more effective outcomes for patients.

Top 5: Things to know about machine learning There are several different types of machine learning and it pays to know a little about each. Here's a quick guide.

Many digital health companies build dashboards for doctors to share data from a device or an app. The entrepreneurs often see this as an important product feature; in reality, most doctors have no time or interest in yet another source of patient information.

Doctors and nurses are already drowning in information. The problem is not a lack of information but knowing how to best use it. Doctors and nurses need technology platforms that make it easier to analyze health data and make treatment decisions.

This is the real role for artificial intelligence in healthcare: Helping doctors and nurses make decisions. As two Babson College professors write on NEJM Catalyst, AI in healthcare is about augmentation not automation.

SEE: Artificial intelligence: Trends, obstacles, and potential wins (Tech Pro Research)

Two startups—Arterys and Astarte Medical—are taking this approach to AI, specifically machine learning. Each company is focusing on a different slice of the information overload problem and aiming to "reduce the cognitive workload" for doctors and improve care for patients.

How Arterys is analyzing medical images for patterns

Healthcare systems have extensive libraries of X-rays, CT scans, MRIs, ultrasounds, and PET scans; researchers are using these scans to train algorithms to spot skin spots that may be melanoma or nodules in lungs that could be cancer. This analysis of thousands of scans can spot patterns that humans miss.

The idea is to have the algorithm do the first pass on the multiple images generated from one scan. The next step is to look for patterns identified in the larger analysis in an individual's records. The deep learning platform then makes a recommendation to the doctor about what to review more closely or what treatment to consider.

Arterys is building these "AI assistants" to inform treatment decisions and automate certain tasks. The company has six products for heart scans, one product for analyzing liver lesions, and one product for lung nodules.

What does Arterys do?

Arterys' 4D Flow software reads an MRI of the heart and provides an analysis of how blood flows through the four chambers of the heart. The software also calculates other heart health data points that an MD usually calculates after drawing by hand the contours of the four chambers as shown in the MRI. Unsurprisingly, the software is much quicker at this math. Arterys has similar products for lung and liver scans; the focus with those products is to track change over time as a measure of disease progression.

How does Arterys help?

Engineers don't use slide rules for complex equations any more, and cardiologists now have a replacement for their manual tools, though the software does not factor in a patient's medical history, co-morbidities, or medication regimen. Doctors can now spend the 60 - 90 minutes previously dedicated to these calculations to other tasks, such as considering the complex interactions of other illnesses that might be affecting the heart or consulting a colleague on a particularly challenging case. By automating some of the work of analyzing scans, a doctor can spend more time with patients explaining a treatment plan or discussing the treatment options.

This technology also has the potential to reduce variation in healthcare. While there are standards of care that doctors are expected to follow, anyone who has sought out a second opinion knows that two doctors can look at the same medical record and make two completely different treatment recommendations. Some variation is good and to be expected, but many healthcare leaders are hoping that deep learning technologies will reduce this variation and reduce costs and improve patient outcomes as a result.

Who is Arterys's target customer?

Hospitals are the most likely purchasers. Adoption by radiologists will determine the success of this approach to treatment. The fear is that hospital executives will want to replace humans with these deep learning platforms; the reality is that these analytical platforms will change the workflow and task lists for doctors, not replace them entirely.

SEE: AI and health: Using machine learning to understand the human immune system (ZDNet)

How NICUtrition is protecting preemies' tummies

Another company using AI as decision support in healthcare is focused on premature babies. Doctors know a lot about how to help preemies breathe but less about how to make sure the digestive system is functioning correctly. As it turns out, gut health directly influences brain health and a baby's risk of learning disabilities later in life.

Astarte Medical is developing a machine learning system to help doctors customize feeding and antibiotic treatment plans for each baby based on each individual's gut health. This technology has the potential to reduce unnecessary antibiotic use (which is a huge issue in public health) and to help tiny babies get the right nutrition at the right time.

What does it do?

Astarte's NICUtrition platform takes existing data from the medical record and analyzes it to measure the health of the preterm infant's gut. The software extracts 200 data points from medical records to create a gut health score and the appropriate feeding recommendations. NICUtrition Guidance makes the feeding suggestions, and NICUtrition MAGI calculates the score.

Treatment recommendations could include an increase or decrease in antibiotics or in food.

How does it help?

Many babies born early are at risk of necrotizing enterocolitis. This illness causes the gut biome to malfunction and eat away at the intestine, which in turn causes systemic infection. The disease occurs in nearly 10% of premature infants. The MAGI gut health score could help identify infants who are at the biggest risk for this problem.

Also, preemies often have developmental delays because a lot of brain development happens in the third trimester. In the last 13 weeks of pregnancy, the fetal brain triples in size. Brain development in a neonatal intensive care unit is very different than brain development in utero. The quality and quantity of a preemie's diet can have a direct effect on brain development.

Who is the target customer?

Children's hospitals and hospitals with neonatal intensive care units. Reducing the number of days in the NICU is good for everyone. Healthcare costs for premature babies can be up to $6,000 per day. That doesn't even count the emotional and physical cost to family members and the babies themselves. Personalizing treatment could help babies grow faster and go home sooner, increasing the family's quality of life and reducing healthcare costs for everyone.

Also, as much of the gut biome research is new, doctors do not have the expertise yet to assess this health risk and take action. NICUtrition has the potential to bring this cutting-edge research into the hospital quickly.

Also see

medicalai.jpg

By Veronica Combs

Veronica is an independent journalist and communications strategist. For more than 10 years, she has covered health and healthcare with a focus on innovation and patient engagement. She led AIR Louisville, a three-year digital health project focused ...