An international team of researchers has developed an artificial intelligence model capable of detecting endometrial cancer with unprecedented precision: 99.26%, widely exceeding current automated diagnostic methods, which barely reach between 78% and 81% accuracy. This advance, published in the magazine Computer Methods and Program in Biomedicine Updatecould revolutionize the early detection of this disease, which represents the most common gynecological cancer in Australia and one of the main oncological threats for women.
The model, called ECGMLPanalyze microscopic tissue images through a sophisticated system that improves the quality of the samples, automatically identifies the suspicious areas and processes the information with self -learning algorithms. The most remarkable is its computational efficiency, which allows a quick and precise analysis without the need for complex additional equipment.
This artificial intelligence acts as a medical assistant with a very high technology that perfects and examines microscopic tissue images with amazing precision. First, he works as an expert photographic editor: Adjusts the lighting of the images, eliminates imperfections such as spots or distortions, and highlights the key details, similar to how we improve a photo on the mobile to see all its details.
Then, it becomes a cell detective, using intelligent algorithms to automatically identify and mark the suspicious areas in the tissue, as if it had a fluorescent marker that only illuminates hazardous cells.
Dr. Asif Karim, an computer expert at Charles Darwin University and co -author of the study, explains that “ECGMLP not only exceeds existing methods in precision, but does it maintaining a clinical efficiency that facilitates its practical implementation.” For her part, Professor Niusha Shafiabady, of the Australian Catholic University, points out that “this technology could be integrated into support systems to support the clinical decision, helping doctors to diagnose faster and more with greater certainty.”
The most impressive is your ability to learn and improve continuously. Through exhaustive training, the system has perfected its ability to ignore irrelevant information and concentrate only on the important thing, achieving an accuracy of 99.26% in its diagnoses. This means that it could become an invaluable tool for doctors, offering rapid and extremely reliable results that would allow to detect endometrial cancer in earlier stages and with greater certainty. Imagine as an infallible assistant who never tires, working in the background to support human specialists and help save lives through more timely and precise diagnoses.
Cancer is the main cause of death in the world. Only in 2020 were almost 10 million deaths attributed to this disease. Wired in Spanish talks with Alberto Hegewisch, medical director of Astrazeneca for the Latin American region, to talk about technological advances that involve artificial intelligence and robots for the diagnosis, treatments and recovery of cancer patients.
“Our study presents an innovative method for automated endometrial cancer diagnosis through histopathological images, standing out for its high precision and minimum processing time, overcoming existing techniques,” explain the authors of the investigation. The approach combines various image preprocessing techniques that guarantee high quality entry data for subsequent analysis. In addition, it uses a multiple stages segmentation method that allows you to precisely extract the regions of interest from tissue samples. “
Research uses the model ECGMLPrecognized for its excellent performance in image classification tasks. This system is capable of identifying complex patterns in the data and selectively filtering the information through control mechanisms, highlighting for its efficiency by requiring less parameters than other comparative models.
“To optimize the model, we make a Detailed ablation study In 12 clinical cases. This systematic evaluation allowed the method to be perfected and identify the configuration of parameters that reaches the greatest diagnostic precision, “the researchers explain.
But the truly promising this technology is its versatility. When scientists tested the system in other types of cancer, the results were equally impressive: 98.57% precision in colorectal cancer, 98.20% in breast cancer and 97.34% in oral cancer. This suggests that the same artificial intelligence architecture could be adapted to improve the diagnosis of multiple oncological diseases.
The potential impact of this advance is particularly relevant to endometrial cancer, whose symptoms – as irregular bleeding or pelvic pain – are usually confused with other less serious conditions, which frequently delays their diagnosis. In Australia, where this disease affects one in 52 women according to Cancer Council, a more precise diagnostic tool could save numerous lives, especially considering that early detection raises the five -year survival rate from 17% to 95%.
Although researchers warn that clinical trials are still necessary before their generalized implementation, this development marks a milestone in the field of artificial intelligence applied to oncology, following the steps of other successful systems for the detection of melanomas or lung cancer.
What began as a project to improve the diagnosis of specific cancer could end up transforming the way we detect and fight multiple oncological diseases worldwide.
#artificial #intelligence #diagnoses #cancer #precision #key #advance #female #health
