Most parents are familiar with the pediatric chart that tracks their child’s height to ensure it is progressing along the expected curve. A child’s place on the chart, coupled with predictions of their future adult height, can determine whether hormonal or other interventions are required.
But these future height predictions, based on X-rays of the child’s hand, are not always accurate — and in some cases may adversely impact a child’s treatment.
Now, an innovative machine learning technique developed by a team of international researchers can predict height accurately within 1.5 inches. “Predictive algorithms generated by machine learning, which can already automatically drive cars and recognize spoken language, are the keys to unlocking data that can precisely inform the pediatrician for real-time decision making,” wrote the scientists, led by Professor Emeritus Zeev Hochberg from the Technion’s Rappaport Faculty of Medicine and Dr. Michael Shmoish, head of the Bioinformatics Unit at the Lorry I. Lokey Interdisciplinary Center for Life Sciences and Engineering.
Published in the Journal of Clinical Endocrinology & Metabolism, the research used data from 1,596 adults whose measurements were documented from their infancy to early adulthood in three long-term studies carried out by the team’s Swedish and Scottish partners. The goal was to predict the children’s adult height based on the growth data up to age 6 years. The researchers used the data to train a computer model to discover various predictive patterns that would probably not be noticed otherwise.
The researchers said their approach mitigates the significant deviations that occur with existing prediction methods and discovers complex patterns that improve the adult height predictions.
Born and raised in Haifa, Prof. Hochberg has served as the director of pediatric endocrinology at the Rambam Medical Center in Haifa, and a full professor of pediatrics and endocrinology at the Technion. He received his Ph.D. from the Technion, his M.D. from Tel Aviv University, and is a certified endocrinologist, pediatrician, and pediatric endocrinologist. Prof. Hochberg is one of two editors-in chief of the Yearbook of Paediatric Endocrinology, and has edited 8 books and authored of more than 300 peer-reviewed articles. He has received numerous awards including the prestigious Andrea Prader Prize for world leadership in pediatric endocrinology.