Study
NLST
(Learn more about this study)
Project ID
NLST-268
Initial CDAS Request Approval
Dec 27, 2016
Title
Intelligent Lung Cancer Diagnostic Aid System: a RADIOMICS approach applied to histology
Summary
Cancer is characterized as an abnormal growth of cells that invade and destroy their surrounding tissues. In view of the estimated 600 thousand new cases estimated for 2016, cancer is the second cause of death of the Brazilian population. According to INCA data, among the various types of cancer, lung cancer is the one with the highest incidence in the world (1.8 million) and the one that kills the most. In Brazil alone, 17,330 new cases of lung cancer (PC) were estimated for the year 2016. The diagnosis from the histological images remains the gold standard in the diagnosis of COP. In order to understand the content and shape of the distribution of tissue type along the biomass of the nodule, the specialist needs to identify and classify the cellular contents of the nodule. However, histological images are complex, and diagnosing CP after visually analyzing various tissue slides throughout the day represents a challenging, extremely tiring and error-prone activity. Recent advances in computational techniques aim to reduce subjectivity in the diagnosis of SC, since they provide systematic and quantitative information that can help pathologists to be more productive, objective and consistent in the diagnostic process. However, segmentation of the CP cell nucleus is a major challenge to traditional imaging techniques. Another challenge in the area is the combination of quantitative representation techniques, called RADIOMICS, with machine learning representation (AM) techniques to determine CP subclassification. Quantitative characterization is important, not only for reliability in diagnosis, but also to guarantee the reproducibility and comparison between the computational tools used in research, providing an improvement of efficiency and accuracy in the decisions of the pathologists and thus, benefiting the patient.
According to the World Health Organization, it is unquestionable that cancer is a public health problem, especially among developing countries. According to data from INCA 2016, lung cancer (PC) is the one with the highest incidence in the world (1.8 million) and the one that kills the most. In Brazil alone, 17,330 new PC cases were estimated for the year 2016. The CP is the one with the highest incidence estimate for 2016 in the state of Alagoas, considering men and women.
Aims
The objective of this project is to develop a computational solution to aid the diagnosis of lung cancer using RADIOMICS descriptors in histological images of resected lung tissue. This project proposes an innovative methodology for the detection and automatic segmentation of lung cancer cell nuclei. Computational algorithms will be developed to extract RADIOMICS descriptors from segmented structures and the entire tissue. Machine Learning algorithms will use the descriptors to classify a histological image into subclassifications of lung cancer: adenocarcinoma, squamous cell carcinomas, and large cell carcinomas. Therefore, it is expected that the result of this project will help the pathologist to improve the prognosis and therapeutic process, defining from the subclassifications the conduction of the best treatment to the patient.
Collaborators
Federal University of Alagoas
