Thoracic CT Lung Nodule Detection

Lung cancer screening and surveillance requires a detailed assessment of lung nodules from CT images. This project demonstrates the use of a deep learning model for locating, segmenting, measuring, and characterizing lung nodules in thoracic CT images. This tool also provides its native zero-footprint viewer to allow users to navigate, visualize, and inspect nodules in 3D. 

This model was trained and tested on lung CT images of thousands of patients in more than a dozen centers.

  • Automatic lung segmentation
  • Automatic lung nodule detection volume rendering of the 3D thoracic CT image and segmented lung nodules
  • Reports diameter and volume measurements on each detected lung nodule
  • Reports the following three nodule characteristics for each nodule: radiographic solidity (solid, mixed, or ground glass), lobulation (subtle or obious), and spiculation (subtle or obvious)
  • Reporting is based on Lung-RADS Guidelines and is exported to an editable report document