Public Release: 

Trained technicians using CV software improved the accuracy and quality of LDCT scans

International Association for the Study of Lung Cancer

DENVER - Trained technician screeners with assisted computer-aided nodule detection or computer vision (CV) screening workstations can efficiently and accurately review and triage abnormal low-dose computed topography (LDCT) scans for radiologist review.

The National Lung Cancer Screening Trial (NLST) reported that three annual low-dose computed topography (LDCT) scans, in contrast to standard lung X-rays, can decrease lung cancer mortality by 20% and overall mortality by 7% in high risk individuals. However, the poor specificity of LDCT leads to a high rate of false positive abnormalities and additional scans are often required after baseline screening. Consequently, implementing LDCT screening is often cost-prohibitive. Training a technician to review scans with the assistance of CV software has the potential to improve both the efficiency and accuracy of LDCT scan interpretation.

Investigators selected 828 baseline LDCT scans from the previously reported Pan-Canadian Early Detection of Lung Cancer Study. Of the 828 baseline scans, 136 scans had proven malignant lung nodules, 536 had proven benign nodules with a range of nodule sizes, and 136 did not have any nodules or calcified nodules 1 mm or larger. CV screening workstations were used to identify solid and subsolid lung nodules and to highlight nodule features and assess a malignancy risk score. Nonradiology technicians were trained to operate the screening workstations and to review the scans for abnormalities to categorize the scans as abnormal (requiring radiologist review) or normal (not requiring review).

###

The results published in the Journal of Thoracic Oncology, the official journal of the International Association for the Study of Lung Cancer (IASLC), reported that of the 136 scans that were classified as normal, 85 were found to have noncalcified nodules. Of the 556 screens classified as having benign nodules, 15 (2.7%) were misclassified as normal by the technician. With regard to the 136 scans that showed lung cancer, the trained technician assisted by CV software accurately triaged 100% of the scans as abnormal. The overall sensitivity and specificity of the technician assisted by CV in identifying an abnormal scan, using 1 mm or larger nodule size cutoff, were 97.8% (95% CI: 96.4-98.8) and 98% (95% CI: 89.5-99.7), respectively. The average time for a trained technician to review the scan and generate a report was 208 seconds.

The authors comment that, "Significant CT scan reader variability was found in the NLST, with false-negative rates ranging from 3.8% to 8.1% and false-positive rates ranging from 3.8% to 69%. The strategy of having a technician aided by CV technology as first reader will likely improve the consistency and quality of scan interpretation. It should be noted that the NLST, Lung-RADS, and PanCan criteria are not entirely validated. Prospective evaluation of these lung nodule management guidelines is needed to define the optimal action threshold."

Co-authors Alexander Ritchie, Martin Tammemagi, Ming Sound Tsao, and Stephen Lam are members of IASLC.

About the IASLC:

The International Association for the Study of Lung Cancer (IASLC) is the only global organization dedicated to the study of lung cancer. Founded in 1974, the association's membership includes more than 5,000 lung cancer specialists in over 100 countries. Visit http://www.iaslc.org for more information.

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.