Reading Time: 4 minutes
A quick evolution of AI in pediatrics with some of its current uses in pediatric medicine today.
The incorporation of artificial intelligence into pediatric medicine has been primarily focused on brain mapping, developmental disorders, oncology, gene profiling, emergency care, and pattern recognition.
These processes deal with a glut of data, the amount of which is expansive.
One of the earliest papers about AI in pediatrics was published in 1984, and it introduced a computer-assisted medical decision-making system known as SHELP.
SHELP, created to diagnose inborn errors related to metabolism even in rare cases, played an important role in the clinical diagnoses and treatment of pediatric diseases.
Before 2008, the major research work on AI and pediatrics focused on the use of applications controlled by ruled-based systems (knowledge-based expert systems), artificial neural networks, genetic algorithms, and decision trees.
These applications came in handy in knowledge extraction and decision making relating to mortality and survival prediction, preterm birth, melanoma, lesion treatment, cancer, and neuroblastoma.
Between 2009 to 2012, the use of AI in pediatrics advanced, becoming more complex. It began to feature logistic regression models, discriminant analysis, and support vector machines for prediction, diagnosis, and care.
AI also assisted with signal, speech, and image processing. Some of the common disease conditions treated in collaboration with these innovations included those related to pathology, genetics, seizures, and infections, and they significantly targeted premature infants and young children.
The period from 2013 up until now has seen the creation of applications imbued with machine learning aimed at tackling the diagnosis and treatment of epilepsy, asthma, pneumonia, schizophrenia, and other neurological conditions like autism.
The newer AI-based tools also incorporate data representation, computer imaging, and other algorithm-based processes. Some of the more current uses of AI in pediatric medicine is summarized below:
a. Eliminating false alarms and alarm fatigue
False alarms are very common in hospitals and they lead to “alarm fatigue,” a situation where caregivers are overwhelmed by the sheer number of alarm signals and become desensitized. Naturally, this leads to delayed responses and sometimes even missing alarms altogether.
It will interest you to note that about 72% to 99% of clinical alarms are false, thus the likelihood of alarm fatigue is high. As a sad result, some patients in need suffer.
Before the advent of AI-based techniques, physicians depended solely on quality improvement projects to combat this: daily electrocardiogram electrode changes, proper skin preparation education, and customization of alarm parameters were some of the only ways to decrease the number of false alarms.
Enter machine-learning algorithms, which can play an important role in automatically classifying vital sign alerts as either real or artifact.
Researchers were able to build these successfully by training them with expert-labeled vital sign data streams. Coupled with the creation and implementation of data-driven vital sign parameters, alarm fatigue can be greatly reduced in a pediatric acute care unit.
b. Biomedical diagnosis
Medical science is nothing without unbiased diagnoses.
Pediatric medicine requires paying the utmost attention to even the tiniest details, details which can be missed by even the most skilled physicians when exhaustion comes into play.
AI-based processes like artificial neural networks, support vector machines, classification trees, and ensemble methods like random forests have been successfully applied to molecular imaging modalities in disease conditions, including neurodegenerative diseases.
Researchers have also employed machine-learning algorithms in the prediction of periventricular leukomalacia in neonates after cardiac surgery.
Additionally, AI is playing a leading role in radiology, including in the automated detection of diseases, the segmentation of lesions, and quantitation.
Machines now have the ability to diagnose diseases on images at a level comparable to skilled physicians. Researchers have even reported the use of models to assess skeletal maturity on pediatric hand radiographs.
Studying genotype-phenotype interrelationships among syndromes can be a nightmare for medical geneticists. This is due to the tedious nature of the job, especially when rare syndromes are involved.
Well, the use of visual diagnostic decision support systems powered by machine-learning algorithms and digital image processing can bring these nightmares to an end.
It offers geneticists a hybrid approach to the automated diagnoses in medical genetics.
c. Wearable technology
Wearable technologies are making waves in several medical procedures like at-home data collection of blood oxygenation, recording patient visits, and keeping close tabs on heart rate, respiration, and ECGs.
These wearable technologies, also used in sleep studies, psychosocial applications, and obesity intervention, are assisting children with movement disorders.
d. Robotic technology and virtual assistants
Robots can help children with neurological disorders like autism with several learning tasks. In general, these children tend to enjoy the tasks more when they interact with a robot compared to when they interact with an adult.
Physicians can also utilize electromechanical and robot-assisted arm training for the improvement of muscular activities after a stroke. Robots and virtual assistants are the future of physical rehabilitation and psychiatric therapy as well as healthcare education and the management of chronic diseases.
Artificial intelligence is changing every aspect of healthcare delivery via the introduction of highly efficient algorithm-controlled processes. These processes are less prone to errors and are capable of detecting even the tiniest detail that may otherwise be missed by physicians.
Contact me or comment if you have questions about AI in pediatric medicine. I would love to hear from you!