Keywords:
Performed at one institution, Diagnostic or prognostic study, Retrospective, Cancer, Computer Applications-Detection, diagnosis, CT, Oncology, Lung, Chest
Authors:
H.-F. Zhu, J. Fu, X.-Y. Meng, X.-L. Mu, C.-B. Liu; Beijing/CN
DOI:
10.26044/ecr2020/C-02292
Methods and materials
Patients
We retrospectively analyzed patients who underwent chest CT at our institution between August 2017 and June 2019. All aspects of this retrospective study were approved by the institutional review board.
Inclusion criteria: (1) All patients underwent chest NECT scanning; (2) Pulmonary nodules with a diameter of longer than 0.6cm and less than 1.5cm;(3)There was no obvious calcification and cavity, and there was no atelectasis and satellite disease around the nodules; (4) With surgical proved pathology; (5) no previous radiotherapy or chemotherapy.
Forty-one patients were included(30 men and 11 women, age(46y-77y ). All nodules were categorized into two groups(pulmonary adenocarcinomas (n=24)vs. benign lesions (n=17)(pulmonary granulation(n=11), pulmonary leiomyoma(n=1),sarcoidosis(n=5)).
Image segmentation
Images were retrieved from the picture archiving and communication system and loaded into ITK-SNAP 2.2.0 software for further segmentation. Segmentation was done on the lung nodule based on lung window (J.F. with 1 years of experience) under the supervision of a board-certified radiologist (HF. Zhu. with 20 years of experience) blinded to the clinical profiles. The regions of interest (ROI) in the texture analysis was drawn freehand around the peripheral boundary of tumor on the slice by a single observer (Figure 2). The texture features were quantitatively extracted usingusing Pyradiomics(https://pyradiomics.readthedocs.io/en/
latest/), an open-source python package for the extraction of Radiomics features from clinical images(Figure 3).
Statistical analysis
The differences between pulmonary adenocarcinomas and benign nodules were statistically evaluated using multivariate analysis and nonparametric test.