ASCO Breakthrough 2019: Lung Cancer Diagnosis Tool Takes CAD to New Dimensions
Posted: Thursday, October 24, 2019
With technology far beyond mere imaging, a team from China has developed a way to use computer-aided diagnosis (CAD) and deep-learning analysis to locate malignant lung lesions on CT scans, measure them, and qualitatively evaluate them for density. Xianling Liu, MD, PhD, of the Second Xiangya Hospital, Changsha, Hunan, presented these findings at the 2019 American Society of Clinical Oncology (ASCO) Breakthrough: A Global Summit for Oncology Innovators in Bangkok (Abstract 27).
The goal was to “generate quantitative morphology features for assisting lesion diagnosis,” explained Dr. Liu. Ultimately, the algorithmic advances allowed the team to calculate each lesion’s longest diameter, shortest diameter, volume, and largest cross-section. The density types that could be identified were calcified, solid, partial-solid, and ground-glass opacity. The data collected were from 3,956 lung CT series with multiple lung nodules from 15 hospital in China. Data on more than 1,100 lung CT scans from other data sets were included as well.
Not only can lung-lesion screening, as performed by radiologists, be time-consuming, explained Dr. Liu, “but its accuracy varies depending on a doctor’s individual experiences.” This CAD system yields free-response receiver operating characteristic of 0.4663, recall of 82.46%, and precision of 36.06% for nodules measuring between 5 and 30 mm, according to Dr. Liu.
“The proposed computer-aided diagnosis system offers a fast and convenient approach for assisting the diagnosis of lung nodule pathologies, and it is beneficial to relate our research to the current framework of lung cancer diagnosis,” Dr Liu stated.
Disclosure: Dr. Liu reported financial relationships with LinkDoc Technology.
