LuCAS

Introduction

Lung cancer is the leading cause of cancer-related deaths worldwide and the mortality rate is still rising rapidly. According to the American Cancer Society statistics reported in 2011, lung cancer has a 5-year survival rate of only 18%. This is because there are few early symptoms of lung cancer, and it is often found only in the late stage of poor prognosis. Five-year survival rates are 54% when detected and treated before metastasis, but only about 15% of all lung cancer patients are detected early.

Various screening programs for high-risk groups have been attempted for early detection of lung cancer, and low-dose CT screening programs have been shown to significantly reduce mortality in lung cancer compared to chest radiographs. In many countries around the world, including Korea and the United States, low-dose CT screening programs are underway or will soon be introduced. We introduce a system to detect, measure, categorize and track nodules in the screening CT images: LuCAS

Performance evaluation

Performance of our system is evaluated on NLST dataset. We chose 44 participants diagnosed with interval lung cancer, compared with participants who did not develop lung cancer (controls.) By retrospective review, cases had higher estimated likelihood of lung cancer. We evaluated the performance of our system in the detection of missed lung cancers.

Above figure shows adenocarcinoma in the right lower lobe. At the first screening, the lung cancer was small, it was missed by the interpreting radiologist but it was detected by our detection system. The third image shows the size of lung cancer increased at the second round of screening. The sensitivity of our system was 95-97% for positive screens and 95-98% for missed lung cancer detection, while the false-positive rate was limited.

Features

Smart nodule list

The system-detecting nodule list is provided with detaied information: nodule location (image number and estimated lobe), measurements (average/long/short diameters), type, interval growth, 4X opinion, and the lungRADs category. You can sort the list in various ways: lungRADs category score, average diameter, nodule volume, image number, or using our smart sorting option. You may choose to check only the most critical 6 nodules in the list.

Display options

The detected nodules are indicated by various annotations including: arrow mark, text breifing, segmentation boundary, and long/short diameters. You can always turn the display of each annotation on or off in the main view.

Handy access to details

The details of a nodule can be easily accessed by keyboard short-cuts or by right-clicking on the detected nodule. You can also apply one-click zoom to the nodule and edit the diameters of the nodule (long/short). The decision to approve or reject the detection can be made as well.

One-step nodule annotation

With our patented nodule annotation tool, you can easily measure long and short diameters of a nodule, and then register it to the system database. A possibly missed nodule can be handled in this way.

Automatic report generation

After reviewing all the images, the system generates a final report in one-click manner. The report can be customized to include clinical information, figures, tables, graphs, in various output formats such as PDF sheet, structured report or burned-in DICOM image. It can be sent to the hospital main PACS as well.

For more information..

We are constantly working on improving our system. Please feel free to contact us if you need more information.