Estanislao OUBEL

Ph.D.

Median Technologies

Head Science & Image Processing

NLST (Learn more about this study)

NLST-362

Oct 18, 2017

Nodule identification and classification in Lung screening

Finding Lung cancer disease at an early and treatable stage remains the most formidable challenge. Computer-assisted early pulmonary nodule detection by imaging offers an opportunity to catch the tumor before it grows and spreads. In the hand of radiologists, these deep-learning algorithms could significantly reduce the false positive rate that currently afflicts early screening and improve cancer prediction efficiency. Even with the growing interest for CAD in the context of screening and diagnostic, no satisfactory system is available in clinical routine, probably because their performance and robustness are expected to be much higher due to liability concerns for delaying a biopsy recommendation.
Our extensive experience in developing lesion management solution aligns very well with such healthcare need. Therefore, we are planning in developing systems that would work 1- to improve accuracy in nodule detection and 2- to classify the detected nodules according to their likelihood to be malignant. We aim to use the unique NLST dataset to train and develop these computer systems. The CT dataset from NLST with its structured annotated demographics and outcomes will also be an important database to use that has the potential to aid our understanding in disease progression.