Skip to Main Content

An official website of the United States government

Development of an assistive lung cancer screening system from chest X-ray images using an anatomically aware deep Convolutional Neural Networks (CNN)

Principal Investigator
Name
Taufiq Hasan
Degrees
B.Sc., M.Sc., Ph.D.
Institution
Department of Biomedical Engineering, Bangladesh University of Engineering and Technology
Position Title
Assistant Professor
Email
About this CDAS Project
Study
PLCO (Learn more about this study)
Project ID
PLCO-471
Initial CDAS Request Approval
Apr 10, 2019
Title
Development of an assistive lung cancer screening system from chest X-ray images using an anatomically aware deep Convolutional Neural Networks (CNN)
Summary
Lung cancer is one of the leading causes of death in Bangladesh [1] with a prevalence of 13.1% in the male population. In this country, 15.9% of all cancer-related deaths in males are due to lung cancer. The most effective method for early diagnosis of lung cancer is the Computed Tomography (CT) imaging [2], although traditional chest X-ray can also be used. However, only 24.2% of District and Upazila public health centers in Bangladesh have X-ray machines [3] while the availability of CT scan facility is about 1%. Even when the equipment is available, there is a severe shortage of expert radiologists who can perform the diagnosis [4]. This project aims to first develop a deep learning based automated X-ray image classification system, and afterward, generate anatomical heatmaps on digital X-ray images in order to assist the radiologist in a teleradiology framework. The PLCO dataset will be used in the first stage to train the deep learning system whereas local dataset acquired through collaborators in Bangladesh will be utilized for fine-tuning and adaptation of the proposed models.

[1] S. A. Hussain and R. Sullivan, “Cancer control in Bangladesh,” Jpn. J. Clin. Oncol., vol. 43, no. 12, pp. 1159–1169, Dec. 2013.
[2] K. M. Latimer and T. F. Mott, “Lung cancer: diagnosis, treatment principles, and screening,” Am. Fam. Physician, vol. 91, no. 4, pp. 250–256, Feb. 2015.
[3] Bangladesh Health Facility Survey 2014: Final Report. 2016.
[4] "Career building in Radiology, Imaging, Career building in Radiology, and Imaging”, The Daily Star, 18-Dec-2013. [Online]. Available: https://www.thedailystar.net/career-building-in-radiology-imaging-2865. [Accessed: 11-Sep-2018].
Aims

In this project, we aim to develop and evaluate a state-of-the-art deep neural network based algorithm for assistive lung cancer diagnosis system from X-ray images. Specific aims of the project are:

* Implementation of an anatomically aware deep learning framework for lung cancer screening from chest X-ray images.
* Collection of lung cancer X-ray images from local hospitals and utilize transfer learning of network weights for improved performance in the collected data.
* Generate heatmaps on X-ray images to be used as an assistive tool for radiologists, and perform the necessarily subjective and objective evaluation.

Collaborators

Dr. Taufiq Hasan, mHealth Lab, Department of Biomedical Engineering, Bangladesh University of Engineering and Technology (BUET).
Asif Shahriar Sushmit, mHealth Lab, Department of Biomedical Engineering, Bangladesh University of Engineering and Technology (BUET).