When he entered the Computer Engineering department at Hanoi University of Science and Technology, Son did not think he could create a tool to support fetal imaging diagnosis with the same accuracy as the leading ultrasound doctors in Vietnam.
Bui Van Son, 23 years old, graduated from the Computer Engineering major at the School of Information and Communication Technology, Hanoi University of Science and Technology in late October 2023 with a good degree. Son's average score of 3.3/4 is not high compared to many of his friends, but it is the expected result, because in addition to studying, he spent most of his time on scientific research.
"Researching at the Biomedical Informatics Laboratory, International Artificial Intelligence Research Center - BK.AI is the most memorable part of my youth during my 5 years at Bach Khoa," Son said.
Here, Son participated in AI application research projects to solve some problems in the medical industry, including the project "Determining the fetal nuchal translucency using ultrasound images" ordered by Hanoi Medical University Hospital. This project helped Son win second prize in the Student Scientific Research Competition of Hanoi University of Science and Technology and the excellent presentation award on the day of his graduation thesis defense.
Bui Van Son. Photo: Duong Tam
Son was a student at a village school in Quang Xuong district, Thanh Hoa province. Admitted to the Computer Engineering major at Hanoi University of Science and Technology in 2018, Son envisioned learning about both hardware and software, and being able to create websites and applications that he had previously dreamed of.
In his second year, when he was familiar with the learning method at Polytechnic, Son learned about artificial intelligence (AI) and participated in scientific research. His direction was different from his friends in the group, and he had not studied any subjects related to AI, which caused Son many difficulties.
The young man started with some free AI courses at school, then searched for more courses on AI, Data Science, Deep Learning online, and combined working at a Japanese company as an AI intern.
A year later, when he was more confident in what he had learned, Son applied to join the school's Biomedical Informatics lab. Here, Son learned about practical problems in the medical field such as decoding genes, using AI to predict suitable drugs for certain diseases... In addition to knowledge about AI, Son had to learn more about Biomedical knowledge such as decoding gene sequences, DNA, mRNA, PCR testing procedures...
In 2022, Dr. Nguyen Hong Quang, Head of the Biomedical Informatics Research Department, Hanoi University of Science and Technology, received an order from Hanoi Medical University Hospital to develop a method to automatically determine the width of the fetal nuchal translucency through 2D ultrasound images, helping to diagnose congenital malformations in the fetus before birth.
Receiving the topic, Mr. Quang immediately thought of Son because he saw that his student was passionate about research and good at image processing. He said that at the beginning of his second year, Son did an image processing problem on identifying defective solar panels from photos taken by drones.
"When Hanoi Medical University Hospital proposed, I assigned it to Son, even though I knew this was a difficult problem," said Mr. Quang.
Son said he was initially quite confused because he did not know what nuchal translucency or nuchal translucency was, or what the purpose of this measurement was. After being guided by the teacher, supported by the doctors, and shown ultrasound videos of fetuses to understand the ultrasound process, along with self-study, Son understood the meaning of this topic.
"The nuchal translucency is an area of fluid that accumulates behind the fetus's neck, appearing from weeks 11 to 14 of pregnancy. If the width of the nuchal translucency is larger than 3 mm, the fetus is at risk of Down syndrome and many other birth defects," Son explained.
Determining the width of the nuchal translucency helps doctors make early diagnoses of fetal abnormalities, thereby providing advice to the mother. However, with current techniques, this is done manually by highly specialized doctors. This technique depends on a number of subjective factors of the doctor, with many potential errors.
Results of measuring the fetal nuchal translucency from ultrasound images, using AI models. Photo: Character provided
Receiving a dataset of up to thousands of fetal ultrasound images, Son encountered difficulties in the beginning. Son tried some advanced image processing models but the results were not positive. The young man could not find a specific direction, and for 2-3 weeks he did not know what to report to his teacher, so he was under pressure.
After that, Son boldly shared with his instructor and was given many suggestions. Son followed those suggestions, consulted some research in the world, and tried many models over and over again.
As a result, the AI model researched by Son helps calculate the nuchal translucency through ultrasound images in about 5 seconds, doctors can measure many times without wasting much time. The website and application integrating the AI model built by Son also help them have more basis to make diagnoses about the fetus or use it for post-checking, minimizing errors during the ultrasound process.
According to a representative of Hanoi University of Science and Technology, the ordering party assessed Son's research results as having the potential for practical application and considered including them in the post-audit process.
"This is a topic but there are probably nearly 20 sub-problems inside and a lot of work to handle. Son has been very diligent and completed it well, creating a support tool that provides results with the same accuracy as the diagnosis of leading ultrasound doctors in Vietnam," said Mr. Quang. The two teachers and students have completed an article on this topic and sent it to the international journal of Biomedical Information.
Currently working as a data science engineer at VNPT, Son is always grateful for the time he devoted himself to scientific research during his university years. With the topic of determining the nuchal translucency of the fetus, Son hopes that the website and application he built will be widely used, bringing good results to the community.
Source link
Comment (0)