Skip to Main Content

COVID-19

What people with cancer should know: https://www.cancer.gov/coronavirus

Get the latest public health information from CDC: https://www.cdc.gov/coronavirus/

Get the latest research information from NIH: https://covid19.nih.gov/

Principal Investigator
Name
Marcelo Oliveira
Degrees
Ph.D
Institution
Universidade Federal de Alagoas (UFAL)
Position Title
Full Professor
Email
About this CDAS Project
Study
NLST (Learn more about this study)
Project ID
NLST-685
Initial CDAS Request Approval
Jun 29, 2020
Title
Smart Radiology: an algorithm to predict the risk index of lung nodules in LDCT.
Summary
The main goal of this project is to develop an algorithm to identify the risk index of lung nodules capable of determining the ideal moment of the patient performs the next screening. This risk index will allow patients with nodules classified as low risk to undergo screening CT exams less frequently than high-risk patients. NLST nodules will be used for training, testing and validating our machine learning and deep learning algorithms. Generative adversarial networks will be used to produce synthetic nodules to enlarge our training set.Nodules and its perinodular zone will be segmented and their radiomics attributes extracted. Using a Convolutional Neural Networks deep-features will be obtained, and then will be fused with radiomics attributes to make only one dataset. Using the fused dataset, machine learning algorithms will be evaluated to predict the risk index of lung nodules.
Aims

- Identify the best radiomics attributes to quantify the perinodular zone.
- Identify the best method to fused deep-features and radiomics;
- Evaluate the real capacity of the IA algorithms to reduce the number of false-positive and false-negative in screening exams identified in the NLST dataset.
- Evaluate the hypothesis that the perinodular zone can improve predictions performance of the machine learning algorithms.
- Evaluate the hypothesis that perinodular zone improves the early nodules (< 10mm) classification accuracy.

Collaborators

Prof. Dr. Paulo Mazzoncini de Azevedo Marques, PhD – Universidade de São Paulo - USP
Dr. Marcel Koenigkam Santos, MD, PhD – Radiologist, Hospital das Clínicas de Ribeirão Preto - HCFMRP/USP