Technology must make more effective cardiovascular disease prognosis
An algorithm based on artificial intelligence, created by researchers from the State University of Campinas (Unicamp) in partnership with the Federal University of Santa Catarina (UFSC), was able to automatically identify changes in the thickness of the intimal sublayer of the carotid artery (the thinnest wall of the vessel that transports blood and oxygen to the brain) from common ultrasound images. The software for image processing and analysis uses techniques based on topological extinction values and mathematical morphology to limit regions of interest and highlight desired aspects.
According to professor and researcher at the Faculty of Medical Sciences (FCM-Unicamp) Wilson Nadruz Junior, the technology was capable of processing medical images, allowing specific measurement of the intimate sublayer, normally done by highly trained professionals and using imaging devices. research, be carried out automatically by any sonographer using routine clinical equipment. “We believe that this technology has enormous potential for cardiovascular prevention, early treatment and reduction of future cardiovascular diseases”, says the cardiologist.
In addition to traditional exams
Currently, the diagnosis of atherosclerosis is made using invasive imaging methods, such as catheterization, or indirect ones, such as magnetic resonance imaging and tomography angiography. Typically, these methods are expensive and time-consuming. Ultrasound examination of the neck region, in turn, is a non-invasive method, used for more than three decades to identify plaques that increase the risk of cardiovascular diseases. It allows you to measure the carotid wall of a structure known as the intima-media complex.
The intimate tissue is the thinnest layer, and the middle tissue represents a thicker part of the carotid wall. Today, by measuring the intima-media complex, it is not possible to differentiate whether the changes are caused by an increase in the intimal sublayer, a more specific measure of atherosclerosis, or by an increase in the sublayer media, which occurs due to the growth of muscle cells and has lower prognostic value. In this context, recent studies conducted at Unicamp, as well as by research groups abroad, indicate that changes in the intimate layer, which comes into contact with the blood, would be more associated with more serious cardiovascular problems, being more important in predicting the risk of illnesses.
According to Nadruz, several groups have already separated the layers that make up these important vessels of the human body manually in the images. The segmentation of the carotid wall fractions in routine exams, however, is hardly visible, even in high-resolution images, as the structure that divides them is very tenuous, requiring specialized equipment and the monitoring of a trained specialist. Therefore, artificial intelligence can be an interesting option for detecting more discreet anomalies that can quickly better direct treatment for diseases in a subclinical stage, when there are still no symptoms. “The intention is not to replace the work of specialists, but to pre-classify and characterize the images in a reliable way, for subsequent evaluation by the specialist. It would be an exam carried out before the patient reaches the cardiologist”, explains Nadruz Junior.
For the coordinator of the echocardiography service at Hospital de Clínicas da Unicamp, professor José Roberto Matos Souza, the algorithm could be the key to faster and more efficient diagnoses of cases of atherosclerosis, a silent disease characterized by the accumulation of fat in the arteries, which makes it difficult the passage of blood and which can lead to heart attacks, strokes and even sudden death. Atherosclerosis is considered the second biggest cause of death in the world, after covid-19, according to the WHO. “This is a slowly evolving disease that usually presents symptoms only in an advanced stage and that can manifest itself in an emergency situation. With new ways for early diagnosis we can save many lives”, says the cardiologist and professor at the Faculty of Medical Sciences at Unicamp.
Component Tree
The possibility of automatic monitoring of this measurement of the carotid intima layer in simple ultrasound examinations of the neck is the great advantage of the new technology. The resource uses artificial intelligence to process the collected images, without causing major impacts on laboratory routines. “This computational resource could be included in existing equipment and applied by any professional in the field”, explains professor and researcher at the Faculty of Technology (FT-Unicamp) Rangel Arthur.
The extracted ultrasound images are pre-processed to eliminate noise and reduce the difference in intensity generated by lighting asymmetry. Next, a so-called vertical filter is applied, used to highlight regions of interest. Finally, the algorithm is able to map and hierarchize the image surface as if it were the topography of a terrain, automatically limiting the regions of interest using topological extinction values. Therefore, the grayscale image can be seen as a relief, where the position of the pixel is its location, and the intensity represents its height. This way, the lighter points branch out to the darker points. This relationship between plateaus, valleys or mountains of this relief is useful for the computer vision system. Analysis of the results allows the identification of patterns and characteristics that can be used as early indicators of diseases and for new clinical trials. “Extinction values consist of excavations sufficient to completely remove higher mounds and flatten them. From this operation, it is possible to highlight the most significant mountain ranges, in terms of area or volume occupied, which exactly represent the patterns of the carotid sublayers”, says professor and researcher at the Federal University of Santa Catarina (UFSC) Alexandre Gonçalves Silva.
According to Arthur, assessments of changes in the thickness of the carotid intima layer made with the computer program were comparable to assessments by well-trained sonographers. “We were able to delineate the carotid intima layer with a higher correlation than that found among the selection specialists themselves and identify changes in the thickness of the intima layer with an accuracy of 85%”, he adds.
Technology transfer
The invention was protected by the Unicamp Innovation Agency (Inova Unicamp) with a patent deposited at the National Institute of Industrial Property (INPI). Currently, the technology for segmenting the intimal layer of the carotid artery is part of Unicamp's Technology Portfolio and is available for licensing. Negotiations are made directly with Inova Unicamp.
Companies and public or private institutions can license intellectual property developed at the University. In addition to access to cutting-edge technologies, technology transfer reduces risks associated with the development of new innovative products and processes and contributes to socioeconomic development based on scientific knowledge.
This report was originally published on the Innovates Unicamp.
