When it is not used to create images of Jesus made in shrimp or pornographic deepfakes, artificial intelligence can be an incredible tool. This is particularly the case in medicine, where AI revolutionizes research in many areas. An international team of scientists, including researchers from Charles Darwin University (CDU) in Australia, has developed a new IA model called ECGMPL, whose advent could lead to a real revolution in the fight against cancer. A detailed advance in a study published in the journal Computer Methods and Programs in Biomedicine Update.
This new model is capable of analyzing microscopic images of cells and tissues in order to detect endometrial cancer – one of the most frequent tumors of the female reproductive system – with exceptional precision of 99.26%, as reported by New Atlas magazine. By way of comparison, we reach current tools to reach a precision of 78.91% to 80.93% of cases. According to the researchers, ECGMPL could be adapted to diagnose other diseases, such as colon cancers and those in the mouth.
Doctor Asif Karim, study co-author and CDU researcher, does not fail to share his satisfaction: “The proposed ECGMLP model surpasses existing methods with an accuracy of 99.26%, thus exceeding transfer by transfer and personalized models mentioned in the study, while being efficient in terms of calculation.”
Endometrial cancer, but not just
To fully measure the power of this AI, it must be understood that cancer cells are on microscopic level, making the medical diagnosis complex.
A well -trained model is capable of analyzing these images (called histopathologicals in scientific jargon), improving its quality and identifying the first signs of cancer. Its ability to target specific areas in order to identify suspicious anomalies, often invisible to the naked eye, represents a major advance in the fight against cancer.
In the case of that of the endometrium, this advance is all the more crucial because an early detection makes it possible to set up an effective treatment to treat patients. In the United States, 600,000 women are currently affected by this disease, while in France, it is the fourth cause of cancer in women.
But the researchers did not stop there. They tested their model on other images of cancer cells from different parts of the body in order to assess its performance in other contexts. And the results are more than promising: ECGMPL has detected colon cancer with 98.57%precision, breast cancer with a reliability of 98.20%and mouth cancer with an accuracy of 97.34%.
