Why Artificial Intelligence for Children’s Medicine?

This is the first post of eight, which will make a case for a very different approach to building and deploying AI applications for children, irrespective of whether they were geographically and socially lucky.

Our story begins in my cloud computing class at Stanford University in 2015, and involves what at first glance might seem like an unlikely student — a board certified pediatric cardiologist named Dr. Anthony Chang. After watching with interest alongside the rest of the medical profession as Watson defeated humans on Jeopardy!, Dr. Chang made an interesting decision. He decided to enroll himself in Stanford’s bioinformatics program. Fast forward to today, and Dr. Chang is a recognized leader in AI in medicine, founding the AIMed Conference and authoring the book, Intelligence-Based Medicine: Artificial Intelligence and Human Cognition in Clinical Medicine and Healthcare.

Drawing on his own professional experiences, the story Dr. Chang tells to explain his decision relates to a little girl in Myanmar. The local team in Myanmar — located over 8,000 miles away from Dr. Chang’s place of work in Orange County, California — wanted to share her ultrasound image with him to get his expert opinion on how best to proceed. Time was of the essence, yet unfortunately, they were unable to share the image. Without locally available expertise, the girl sadly died on the operating room table. When Dr. Chang later saw the ultrasound image, he felt sure he could have diagnosed the problem and saved her life.

This story is just one of many that serves to exemplify the fact that there are many challenges facing pediatric healthcare today, both in the US and around the world.

Pediatric clinical expertise is not distributed.

The NY Times recently reported that hospitals around the country, from regional medical centers to smaller local facilities, are closing pediatric units. The reason is economics: institutions make more money from adult patients[1]. In April 2022, Henrico Doctors’ Hospital in Richmond, VA ended its pediatric inpatient services. In July, Tufts Children’s Hospital in Boston followed suit. Pediatric units in Colorado Springs, CO, Raleigh, NC, and Doylestown, PA, have closed as well. The result in the US is that pediatric clinical expertise is increasingly being geographically concentrated in the 500 specialized children’s hospitals leaving clinics, regional providers, and non-children’s healthcare providers with little capacity or capability.

Pediatric clinical expertise is not plentiful.

Consider pediatric cancer. Great strides have been made, and as a result there are far more pediatric cancer survivors than in decades past. All too familiar to pediatricians like Dr. Chang, however is the fact that many of these survivors are at increased risk of dying from heart disease unless they are under the care of a pediatric cardiologist. There are 3,000 pediatric cardiologists in the US., located mostly in the larger cities, with only 100 new pediatric cardiologists joining their ranks each year. Yet our population of children — including our population of pediatric cancer survivors — is growing by 4,000,000 per year. And this is not a problem of expertise limited to pediatric cardiology. Consider that three states (Montana, South Dakota, and Wyoming) don’t have a single pediatric emergency physician.

Outside the US, the challenges are orders of magnitude greater. Take a moment to consider just a few of the current and eye-opening global state of affairs:

· In Mexico, congenital heart disease (CHD) is recognized as the second most common cause of death in children under five years of age,[2] yet less than a third of children born in Mexico with CHD have access to any treatment, much less a pediatric cardiologist.

· India has only 300 pediatric cardiologists for a population of 1.39 billion, compared to the 3,000 that serve the U.S. population 330 million. That’s 1/10th the number of physicians for 4 times the population — a 40x difference.[3]

· In Africa, pneumonia is still the number one cause of death[4], not because there isn’t available treatment, but because there is no one to diagnose the condition. By 2050, Africa will be home to one billion children; two in every five children in the world will be born there. Their need for pediatric healthcare will be exponentially greater in the coming years.

Long tail conditions are not easy to diagnose by any one clinician.

More likely than not, you’ve never heard of focal cortical dysplasia (FCD) — a condition that is estimated to be present in 1 in 2,500 newborn infants. While knowing about FCD may be of no real consequence to you, it can have life-altering implications for those who suffer from it. Consider a child in the southern United States, who had multiple seizures every day for most of his life. When he had them at night, he would wake up screaming. While he had MRI scans, his clinicians were unable to recognize and diagnose the underlying cause as FCD. As a result, he was put on one ineffective drug after another for years. Had an AI application trained on a wealth of previously inaccessible MRI data from multiple locations detected his FCD sooner, he could have received the surgery necessary to end his epilepsy. When we’re talking about years worth of drugs, doctors and seizures, imagine how it would have changed his and his family’s life.

In short, today’s healthcare system — especially when it comes to pediatric care — is based in large part on you being geographically lucky, or what the British call “wining the post code lottery”. As a result, there is a gap in access and quality of healthcare between cities and rural areas. What’s true in first-world countries is even more so when compared to the developing world. What holds the key to eliminating this aspect of healthcare inequity globally isn’t just the building of more medical schools. Rather, Dr. Chang and other clinicians believe it is Artificial intelligence (AI) applications. With AI, clinical expertise can be widely distributed. With AI, clinicians who might never have hands on experience with rare conditions can be given access to global data that will allow them to make informed diagnoses. With AI, every clinician can save time and maybe even lives.

Consumer AI applications such as Siri and now ChatGPT, are already serving to demonstrate the remarkable progress AI has made in the past ten years. So how was this progress achieved? Read on to learn why we’ve seen rapid advances in consumer AI and not AI in medicine.

If you’d like to stay up to date on our progress to the moon, register for the newsletter at www.pediatricmoonshot.com, follow us on LinkedIn, subscribe to the Pediatric Moonshot podcast, listen to the Spotify playlist, and subscribe to the Pediatric Moonshot Youtube channel.

Many thanks for extensive editing by Laura Jana, Pediatrician, Social Entrepreneur & Connector of Dots; Leanne West, Chief Engineer of Pediatric Technology at Georgia Tech. Special thanks to Alberto Tozzi, Head of Predictive and Preventive Medicine Research Unit at Ospedale Pediatrico Bambino Gesù for the translation to Italian.