Bayesian probabilistic techniques for cancer growth prediction/ prognosis based on informative features containing significant medical details within the historical CT scans
Principal Investigator
Name
Haroon Rasheed
Degrees
PhD
Institution
Bahria University,Karachi - 75620 Pakistan
Position Title
Associate Professor
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-307
Initial CDAS Request Approval
Jun 8, 2017
Title
Bayesian probabilistic techniques for cancer growth prediction/ prognosis based on informative features containing significant medical details within the historical CT scans
Summary
Lung cancer is one of the leading causes of death from cancer among males and females. Recent research studies report a 5- year survival rate of approximately 54% for cancer detection at an early stage whereas it is only 4 % at advanced stages. Thus, early detection and characterization of lung nodules is essential because early diagnosis can help in targeted and effective cancer treatment.
In order to make a correct diagnosis of the tumor, radiologists are required to study a large number of medical images in a short period of time. The burden imposed on radiologists continues to increase with the rapidly growing use of low-dose CT for lung screening. This makes the decision-making of lung nodule classification difficult as malignant or benign on the basis of CT images and other clinical information. Thus, reliable and efficient automated nodule detection schemes can become an integral component of future lung cancer screening protocols.
Currently, even though Computer Aided Detection (CAD) systems are proven to improve the efficiency of radiologists in the detection of nodules, these are not widely used in clinical practice due to the deficiencies that need to be overcome. During the last decade, ample research has been directed towards machine learning techniques for lung cancer prediction with promising results achieved. However, there is a need to investigate probabilistic models for early cancer detection and reliable prognosis. A reliable prediction and prognosis would help determining the best possible medical treatment for lung cancer patients.
In order to make a correct diagnosis of the tumor, radiologists are required to study a large number of medical images in a short period of time. The burden imposed on radiologists continues to increase with the rapidly growing use of low-dose CT for lung screening. This makes the decision-making of lung nodule classification difficult as malignant or benign on the basis of CT images and other clinical information. Thus, reliable and efficient automated nodule detection schemes can become an integral component of future lung cancer screening protocols.
Currently, even though Computer Aided Detection (CAD) systems are proven to improve the efficiency of radiologists in the detection of nodules, these are not widely used in clinical practice due to the deficiencies that need to be overcome. During the last decade, ample research has been directed towards machine learning techniques for lung cancer prediction with promising results achieved. However, there is a need to investigate probabilistic models for early cancer detection and reliable prognosis. A reliable prediction and prognosis would help determining the best possible medical treatment for lung cancer patients.
Aims
1. To investigate and collect informative features containing significant medical details of a lung tumor within a CT scan.
2. To formulate a quantitative relationship between nodule growth and the investigated features of lung tumor using historical CT images.
3. To develop Bayesian probabilistic techniques for cancer growth prediction/prognosis based on the collected medical details in CT images.
Collaborators
Tariq Mairaj Rasool Khan, PhD (National University of Sciences and Technology)