Computer Tomography (CT) Scan image classification using Artificial Intelligence
Principal Investigator
Name
Manish Chaturvedi
Degrees
Ph.D.
Institution
Pandit Deendayal Petroleum University
Position Title
Assistant Professor
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-372
Initial CDAS Request Approval
Nov 9, 2017
Title
Computer Tomography (CT) Scan image classification using Artificial Intelligence
Summary
A pulmonary nodule is a small round or oval-shaped growth in the lung. Pulmonary nodules are smaller than three centimeters (around 1.2 inches) in diameter. If the growth is larger than that, it is called a pulmonary mass and is more likely to represent a cancer than a nodule. There are two main types of pulmonary nodules: malignant (cancerous) and benign (noncancerous). Over 90% of pulmonary nodules that are smaller than two centimeters (around 3/4 inch) in diameter are benign. A nodule is considered as suspicious for lung cancer if it is greater than 4 mm in diameter.
The Computer Tomography (CT) Scan generates images of chest/lung which are manually examined by the radiologists to detect the abnormalities. Aim of this study is to design an artificial intelligence based model for classifying the Computer Tomography (CT) Scan images into the following categories:
1) no significant abnormalities
2) minor abnormalities not suspicious for lung cancer
3) clinically significant abnormalities not suspicious for lung cancer
4) abnormalities suspicious for lung cancer
5) images with poor quality (Inadequate exams)
The CT scanned images from NLST will be used to develop the model. The set of images will be divided into training and testing data set. The training data set will comprise of nearly 80 % of and the remaining images will be used for testing purpose. The CT images from all the different age groups and gender will be uniformly distributed in both the data set. The model will be evaluated for the classification accuracy (false positives, false negatives) and recall (precision), computation requirement, and feasibility of deployment on a hand-held device. The five-fold cross validation will be carried out for evaluating the model.
We aim to use the open source machine learning platforms such as theano, keras, tensorflow, etc. for implementation of the model.
Deliverable:
1) an artificial intelligence model for CT scan image classification with appropriately tuned parameters
2) performance analysis of the model
3) open source software for CT scan image classification based on the model
The Computer Tomography (CT) Scan generates images of chest/lung which are manually examined by the radiologists to detect the abnormalities. Aim of this study is to design an artificial intelligence based model for classifying the Computer Tomography (CT) Scan images into the following categories:
1) no significant abnormalities
2) minor abnormalities not suspicious for lung cancer
3) clinically significant abnormalities not suspicious for lung cancer
4) abnormalities suspicious for lung cancer
5) images with poor quality (Inadequate exams)
The CT scanned images from NLST will be used to develop the model. The set of images will be divided into training and testing data set. The training data set will comprise of nearly 80 % of and the remaining images will be used for testing purpose. The CT images from all the different age groups and gender will be uniformly distributed in both the data set. The model will be evaluated for the classification accuracy (false positives, false negatives) and recall (precision), computation requirement, and feasibility of deployment on a hand-held device. The five-fold cross validation will be carried out for evaluating the model.
We aim to use the open source machine learning platforms such as theano, keras, tensorflow, etc. for implementation of the model.
Deliverable:
1) an artificial intelligence model for CT scan image classification with appropriately tuned parameters
2) performance analysis of the model
3) open source software for CT scan image classification based on the model
Aims
1) an artificial intelligence model for CT scan image classification with appropriately tuned parameters
2) performance analysis of the model
3) open source software for CT scan image classification based on the model
Collaborators
Manish Chaturvedi, Pandit Deendayal Petroleum University