https://biometry.nci.nih.gov/cdas/approved-projects/2130/

Machine learning based lung cancer identification and characterization

Principal Investigator

Name: Richard Vlasimsky
Degrees: B.S., M.B.A.
Institution: IMIDEX
Position Title: CEO
Email: richard.vlasimsky@imidex.com

Project Information
Study: NLST
Project ID: NLST-495
Title: Machine learning based lung cancer identification and characterization

Summary

Radiologist are under significant pressure to increase their reading efficiency, while at the same time to improve their diagnostic accuracy. This is a particularly the case with cancer, where early detection and treatment can make a big difference in both morbidity and mortality. As a result, radiological examinations for cancer often lead to false positives, putting the patient at un-necessary risk from invasive procedures such as lung biopsies.

This project will use machine learning technology to develop 3D image recognition algorithms that intercept CT images for early cancer detection and treatment guidance. Evaluation of the algorithms will be performed on a hold out sample of data and sensitivity / specificity ROC curves will be generated to assess accuracy.

Aims

The aims of this project are:
-provide early detection of cancer from CT,
-more accurately characterize and classify lesions from CT based on pathology.
-predict the response of different treatment modalities based on the radiological CT,
-predict the progression of cancer based on series of radiological CT’s and ultimately mortality.

Collaborators

Kenneth Bellian, MD
Jake Gelfand
Roger Nichols, MD
Tom Suby-Long, MD