Deep Learning at Chest Radiography and CT: Automated Detection and Diagnosis of Cancers.
Principal Investigator
Name
Daiju Ueda
Degrees
M.D.
Institution
Department of Radiology and Interventional Radiology, Graduate School of Medicine, Osaka City University
Position Title
Resercher
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-384
Initial CDAS Request Approval
Jan 2, 2018
Title
Deep Learning at Chest Radiography and CT: Automated Detection and Diagnosis of Cancers.
Summary
Lung cancer is an aggressive and heterogeneous disease. Advances in surgical, radiotherapeutic and chemotherapeutic approaches have been made, but the long-term survival rate remains low.The best way to reduce its mortality is early detection.
Chest radiography and CT are used as screening tools to detect lung cancers. Chest radiography is the most commonly used technique in clinical practice to lung cancer screening. The advantages of chest radiography is the low cost, low radiation dose, and easy accessibility.Chest CT can reduce in mortality from lung cancer by 20.0% as compared with the radiography. The advantages of chest CT is accuracy.
In the chest radiography, small lung cancers are hard to detect. And it sometimes difficult to distinguish cancer as a mass lesion to normal bronchi and bone benign lesions like fracture and bone island. In the chest CT, it’s pernickety and time-wasting to find subtle nodules from all of lung areas.
One solution of these problems is an assist by algorithms of automatic detecting system. The system probably help radiologists. To develop the algorithms, deep learning techniques can be applied which is the state-of-the-art in visual object recognition. Deep learning is a subfield of machine learning concerned with algorithms inspired by the structure and function of the brain. The deep learning methods extract features from images owing to optimization algorithms called backpropagation which adjust its internal parameters by themselves to best accuracy.
The purpose of our study it to develop the deep learning algorithms to detect lung cancers in chest radiography and CT. After developing the algorithms, we will apply them in our daily medical practice to evaluate.
Chest radiography and CT are used as screening tools to detect lung cancers. Chest radiography is the most commonly used technique in clinical practice to lung cancer screening. The advantages of chest radiography is the low cost, low radiation dose, and easy accessibility.Chest CT can reduce in mortality from lung cancer by 20.0% as compared with the radiography. The advantages of chest CT is accuracy.
In the chest radiography, small lung cancers are hard to detect. And it sometimes difficult to distinguish cancer as a mass lesion to normal bronchi and bone benign lesions like fracture and bone island. In the chest CT, it’s pernickety and time-wasting to find subtle nodules from all of lung areas.
One solution of these problems is an assist by algorithms of automatic detecting system. The system probably help radiologists. To develop the algorithms, deep learning techniques can be applied which is the state-of-the-art in visual object recognition. Deep learning is a subfield of machine learning concerned with algorithms inspired by the structure and function of the brain. The deep learning methods extract features from images owing to optimization algorithms called backpropagation which adjust its internal parameters by themselves to best accuracy.
The purpose of our study it to develop the deep learning algorithms to detect lung cancers in chest radiography and CT. After developing the algorithms, we will apply them in our daily medical practice to evaluate.
Aims
The aim of our study it to develop the deep learning algorithms to detect lung cancers in chest radiography and CT. After developing the algorithms, we will apply them in our daily medical practice to evaluate.
Collaborators
Choppin Antoine LPixel Inc.
Dennis Romero LPixel Inc.