Vietnamese students successfully ‘decode’ the word doctor using machine learning technology

Using technology to overcome the complexity of Vietnamese handwriting

In the newly released information, RMIT Vietnam said that, through working with the Hospital for Tropical Diseases and Oxford University Clinical Research Unit (OUCRU) in Ho Chi Minh City, Phung Minh Tuan, a student The University’s Software Engineering department has successfully developed a terminal set to recognize the writing on Vietnamese medical records. This is a potential technology to help accelerate the digitization of medical records that the health sector is deploying.

A male student studying his final year at the Faculty of Science and Technology, RMIT University said that handwriting recognition technology has improved a lot, but most existing methods have been developed to read English and Until now, there is still very little software specifically for Vietnamese ”.

“Recognizing Vietnamese handwriting is basically more challenging in English because of the presence of complex layers of characters, intonation and punctuation,” said Phung Minh Tuan.

Phung Minh Tuan spent more than three months “trial-and-error” to find the most effective way to convert the image of a paper medical record into an electronic version. “We address these challenges by promoting various tasks on the handwriting set,” said the male student.

Student Phung Minh Tuan spent more than three months “trial-and-error” to find the most effective way to convert the image of a paper medical record into an electronic version.

Tools to support data digitization in medical facilities

As explained by the author of the study: “We apply the process of reducing infection, breaking down writing to the word level and applying the Bigram language model to increase the probability of possible editing for common words. around. More important is to coordinate and implement a machine learning structure that encompasses a ResNet neural network for textual extraction and BiLSTM for frequency pattern modeling, and CTC for final transcription. At this point, the final output signal in string form along with the lexicon will help with more accurate results ”.

As a lecturer in the Faculty of Science and Technology at RMIT University and a direct teacher instructing Phung Minh Tuan students, Dr. Dinh Ngoc Minh emphasized on the promising results of the project.

Dr Minh said the collective could play an essential role in supporting the digital transformation of medical facilities and hospitals in Vietnam, making them more willing to switch to a disease management system. Modern electronic project.

“The work proposed by Tuan could accelerate the process of digitizing the medical record system. With the help of machines to handle the entire medical record, medical facilities can gradually switch to the electronic system without having to change the process abruptly, ”said Dr. Minh.

According to him, such a system will also allow medical facilities in remote areas or health workers who do not have access to computers to continue with the current paper system and be easily digitized later. there.

Student Phung Minh Tuan and direct instructor, Dr. Dinh Ngoc Minh, Lecturer of Science and Technology Faculty, RMIT University.

The teacher also believes that being able to easily share patient’s medical records between departments and departments in medical facilities will help reduce unnecessary tests and optimize treatment, thereby gradually improving. quality of medical care.

“Most importantly, Tuan’s work could create a digital medical record data set for various potential medical machine learning solutions. In fact, the parties working with us, Hospital for Tropical Diseases and OUCRU, plan to use the generated data to develop a system of diagnostic specialists, improve treatment procedures, and minimize errors in medical practice. faculty ”, Dr. Minh shared more.

With this work, student Phung Minh Tuan has got an internship at OUCRU. Not only that, male students were also given a presentation about their research at the international science conference class A – The ACIS2020, AHT seminar as well as online exhibition of RMIT students’ works.