Emmanuel Zachariah

Ph.D.

UMDNJ-Robert Wood Johnson Medical School

Assistant Professor

NLST (Learn more about this study)

NLST-24

Jun 25, 2013

A novel class prediction algorithm for the analysis of nodules in NLST

Lung cancer is the leading cause of cancer death in the United States. Most of the patients are already in their advance stage of disease when they first show the symptoms. Identification patient at early stage will help oncologist to surgically remove the tumor. National Lung Screening Trail (NLST) results suggested that screening Individuals identified at high risk for lung cancer (i.e. current or former smokers, with at least 30 pack-years, aged between 55-74 years with no history of lung cancer) will reduce the mortality significantly. Currently CT scan is being used to identify the malignant nodules in high risk patient population. However, CT scan yields a high rate of false-positive results (20 to 40%).

The goal of the present study is to test and validate our proprietary class prediction algorithm to differentiate malignant nodules from benign nodules.

We already have expertise to analyze gene expression data using CAD based program. (Patent pending: Method and Kit for Classifying a Patient filed on 01.12.2012; PCT: WO 2012/006008 A1). In the present study we wish to test and validate our class prediction algorithm for the differentiation of malignant nodules from beginning nodules.

Currently we have been using two open source libraries (ITK and VTK) for the segmentation and visualization of the lungs. We have used the programs for the segmentation of the lungs with the DICOM files obtained from CT scan images. What we have done is called semi-automatic segmentation as it requires inputs from the user. In the next step we will explore automatic segmentation of lungs using ITK and VTK. Once the nodules are identified the voxel data will be analyzed using our class prediction algorithm.

Aim 1: To develop an automatic CAD program to identify nodules from background signals.
We will use freely available medical visualization tool kit (VTK) to identify the benign and malignant nodules

Aim 2: To test our proprietary class prediction algorithm to differentiate malignant nodules from benign nodules.
The Voxel data obtain from CT scan image will be analyzed with our proprietary class prediction algorithm.
We wish to test ~ 200 patients images obtained from the NLST to build the model, training, testing and validation of our algorithm.

Aim 3: To validate our proprietary class prediction algorithm in larger patients population.
The CT scan image will be analyzed with our proprietary class prediction algorithm.
We wish to test ~ 3,000 patients images obtained from NLST to validate our algorithm in large number of patients.

Judith K Amorosa, MD. Clinical Professor, UMDNJ/University Radiology Group, New Brunswick, NJ 08901.

Salma Jabbour. MD. Assistant Professor, UMDNJ-Robert Wood Johnson Medical School, New Brunswick, NJ 08901.

John Langenfeld, MD. Associate Professor, UMDNJ-Robert Wood Johnson Medical School, The Cancer Institute of New Jersey, New Brunswick, NJ 08901.

Joseph Aisner, MD. Professor, UMDNJ-Robert Wood Johnson Medical School, The Cancer Institute of New Jersey, New Brunswick, NJ 08901.