The system can help medical specialists in the development of their conclusive diagnoses.
Three students from Mexico's Superior School of Computer Sciences (Escom) have developed a system to detect lung cancer. In the Central American country, lung cancer is the second-highest cause of death among men and the fourth for women.
The methodology, created by the trio, analyzes images of computerized tomographies of the thorax and then classifies them in order to prepare a pre-diagnosis. The system is designed to help medical specialists in the development of conclusive diagnoses.
In order to analyze the images, the students use pattern recognition techniques (Artificial Neural Networks), which provides a basis for classifications.
Desarrollan en la #Escom sistema para detección de cáncer de pulmón. Con un banco de imágenes públicas de mil 400 pacientes, los estudiantes entrenaron algoritmos que reconocen anomalías en los pulmones. https://t.co/w5YWSdaIQI pic.twitter.com/NQR8zC5csd— IPN (@IPN_MX) 17 de marzo de 2019
The three inventors are Ximena Cortes, Isaac Aguirre and Sergio Martinez Avila. During development, the students worked with a database of 1,400 patients through U.S. site, The Lung Image Database Consortium Image Collection (LIDC-IDRI).
Aguirre explains that the group used one section of the images to train an algorithm, another for the evaluation process that allows one to measure the performance of the system, while the last section tests the system using images exclusive of those used to train the algorithm.
"Our system has reached a precision rate of 95 percent, referring to the percentage of pre-diagnostics that are classified correctly," Aguirre said.
Cortes stressed that the interpretation of a chest tomography, "is not simple, because of the amount of information it comes with, which can make it difficult to make decisions and cause erroneous diagnoses."
Martinez added that the system is not meant to substitute clinical analysis and traditional methods, but to prevent healthy patients from undergoing biopsies or any other unnecessary and invasive treatment.
"The program is able to identify images that present the slightest indication of cancer, which allows the specialist to suggest a biopsy only to those patients selected by the system."
According to the U.S. National Cancer Institute, 70% of cancer diagnoses in early clinical conditions have a better response to treatment.