Neural networks and intelligent technologies are better than pediatricians have learned to diagnose childhood diseases
This conclusion was found by researchers from the University of Science and Technology in Hong Kong. Experts in the field of AI and machine learning decided to test the developed program for the diagnosis of childhood diseases. The finding was striking – the AI system demonstrated the best results compared to young pediatricians whose work experience is from 3 to 15 years. However, the “computer” could not give points to experienced doctors with 15 years of knowledge and practice.
What constitutes experiment’s idea?
Testing took place for 18 months at the Children’s Medical Center, Guangzhou. To efficacy evaluation of the system, data from 600,000 children were used. Data were taken using electronic medical records (HER). The task of the system is to analyze information on its issues, as well as a comparison with the evaluators of pediatricians themselves. For accuracy, it was decided to share the results of diagnoses of inexperienced and experienced doctors, so that testing could take into account human factors and professional experience.
AI worked by dividing diseases into specific categories by organ groups (cardiovascular system, gastrointestinal tract, upper and lower anatomical airway, etc.). The main categories were divided into subcategories, thereby creating an imitation of the program for the real mentality of pediatricians, who make a diagnosis based on specific initial information. In total, the program analyzed more than 100 million factors, and the classification by machine learning methods made it possible to generate about 55 different diagnoses for outpatients.
The result was logical. The program with AI was able to more accurately make the correct diagnoses in comparison with inexperienced doctors. The results of the intellectual system and experienced pediatricians were almost identical.
What conclusion can be drawn from this?
In the next 5-10 years, AI systems and machine learning will become a key medical appliance for diagnosing diseases and treating diseases. However, this does not mean the machines and AI programs will be able to completely take the place of specialists.
The perspective of such projects is to combine the experience of doctors and simplify the processing of a huge amount of data. Machine learning will allow doctors to more quickly and efficiently perform their duties, eliminating errors in the chosen methods of treatment.
The attention of the world community will continue to be drawn to research in China. The Hong Kong University of Science and Technology’s experiment has once again proved that learning neural networks will soon make a breakthrough in medicine.
Editor of IMD News