According to the American Cancer Society, it is estimated that about 13 percent of all cancers that are diagnosed are lung cancers . Thus, for this year, they estimate that in the United States alone, about 235,760 new cases will be diagnosed, while more than 130,000 people will die from this disease.
In our country, according to the Spanish Association Against Cancer, about 18,000 people die each year, of which 16,000 are men and 2,000 are women. In addition, it is considered the leading cause of cancer death in males , and the sixth in females. In turn, it is estimated that the mean age at death is 68 years of age for men and 66 years for women.
According to the World Health Organization (WHO), lung cancer is the leading cause of death worldwide , with an estimated 1.8 million deaths in 2020.
Recently, we have learned that an artificial intelligence (AI) program is able to accurately predict the risk that nodules detected in the lungs on screening CT will become cancerous . The study has been published in the specialized journal ‘Radiology’.
Low-dose radiation chest CT is used to detect people at high risk for lung cancer, such as regular smokers. It has been shown to help significantly reduce mortality from this cancer by helping to detect cancers at an early stage , precisely when medical treatment can be most successful.
Although lung cancer usually appears as pulmonary nodules on CT images, most of the nodules found are benign and do not require further clinical studies. Hence, the existence of this new program could be very useful when it comes to distinguishing with great precision between benign and malignant nodules , especially when it comes to detecting cancer early.
For this new study, the scientists developed an algorithm for the evaluation of pulmonary nodules using deep learning . This type of learning consists of an artificial intelligence (AI) application capable of finding certain patterns in the image data .
Thus, the researchers trained the algorithm with the use of CT images of more than 16,000 nodules, which included 1,249 malignancies.
The algorithm yielded excellent results , even exceeding the Pan-Canadian Model for Early Detection of Lung Cancer established for estimating the risk of lung nodule malignancy.
This is an unprecedented advance, since the algorithm can be extremely helpful for radiologists to accurately estimate the risk of malignancy of those pulmonary nodules that have been found in the test. Thus, “this can help optimize follow-up recommendations for lung cancer screening participants.”
The advantages and benefits that this program would offer
The experts indicate that, in addition, this algorithm could potentially provide additional benefits to the clinic. Since it does not require manual interpretation of the different characteristics of nodule images, it can reduce substantial inter-observer variability in interpreting CT results .
“This can lead to fewer unnecessary diagnostic interventions, reduce the workload for radiologists, and reduce lung cancer screening costs.”
What’s more, the scientists plan to continue improving the algorithm , incorporating different clinical parameters just as important, such as gender, age and smoking history .
In turn, they are working on the development of a deep learning algorithm that takes multiple CT exams, ideal for comparison of before and after tests, useful for nodules detected later.
Reference: “Deep Learning for Malignancy Risk Estimation of Pulmonary Nodules Detected at Low-Dose Screening CT” by Kiran Vaidhya Venkadesh, Arnaud AA Setio, Anton Schreuder, Ernst T. Scholten, Kaman Chung, Mathilde MW Wille, Zaigham Saghir, Bram van Ginneken , Mathias Prokop and Colin Jacobs, May 18, 2021, Radiology. DOI: 10.1148 / radiol.2021204433
