Applying Artificial Intelligence, Smart Health Center (under VinBigData) has become the first in Vietnam to successfully develop two modules that automatically diagnose brain lesions on computed tomography (CT) and Magnetic resonance imaging (MRI) images. Thereby, 07 modules of VinDr – a comprehensive AI solution for medical imaging has been officially completed.
Based on advanced Machine Learning technology, VinDr-BrainCT and VinDr-BrainMR are two newly deployed modules of VinDr-AI which aim at assisting radiologists in detecting and localizing brain lesions. Not only localizing, VinDr solution is capable of classifying diseases accurately and efficiently, thus supporting rapid screening in overloaded hospitals at the provincial and central levels. The introduction of these two new modules has officially made VinBigData the first organization in Vietnam to apply AI in interpreting CT and MRI studies. Currently, VinDr-BrainCT and VinDr-BrainMR have been implemented at four hospitals including Hanoi Medical University Hospital, Phu Tho General Hospital, Hai Phong International General Hospital and Binh Dinh General Hospital.
Specifically, with VinDr-BrainCT, radiologists have an additional effective tool to diagnose 19 common types of abnormalities such as cerebral atrophy, cerebral edema, skull fracture, subdural hematoma, etc. with 82% accuracy. Besides CT images, VinDr AI model is also capable of reading MRI to detect more complex brain lesions such as infiltration, sinus lesions, gliosis with an accuracy of up to 90% and the average speed of 35 seconds for each scan, 10 times faster than the doctor’s diagnosis speed.
By successfully developing VinDr-BrainCT and VinDr-BrainMR, the VinBigData’s Smart Health Center has officially completed all seven modules of the comprehensive AI solution for medical imaging – VinDr. The solution is capable of processing a variety of images from X-ray to CT and MRI, thereby providing diagnostic suggestions for nearly 70 common abnormalities in chest, liver, mammo, brain and spine.
Along with the new features, VinDr has also been integrated with the DICOM Imaging Router for classification of body parts from DICOM X-ray scans. This algorithm acts as a classifier, helping identify the image to be predicted and send it to a suitable AI module. It is considered as a step towards gradually turning VinDr into a smart, synchronous and seamless system which minimizes interruptions and manual manipulations in the diagnostic process.
The VinDr solution (with two groups of AI and PACS features) developed by the Smart Health Center, VinBigData, has been put into practice in dozens of clinics and hospitals across the country. The solution is expected to contribute to enhancing the quality of medical examination and treatment as well as improving public health.