Table 1.
Demographics of the patients.
Fig 1.
Normal and malignant airway tissue.
EB-OCT images of (A) normal and (B) malignant airway tissue. ROI started from lumen boundary and ended at a depth of 100 pixels along A-line, corresponding to the region between two red circles. A-line was indicated by the yellow arrow. The average value of attenuation coefficient and intensity for each depth of 100 randomly selected A-lines were extracted from EB-OCT images of normal (C) and malignant (D) lesions, respectively.
Table 2.
Attenuation coefficient and image features of normal and malignant airway tissues with significant p-values.
Fig 2.
The ROC curve of the testing set of the classifier.
Classification accuracy reached up to 81.4%, and sensitivity and specificity were 76.4% and 84.8% respectively.
Fig 3.
Pulmonary nodules of different cases on CT scan and corresponding OCT images.
Although final pathology confirmed that they were different type lesions, including pneumonia (A&B), adenocarcinoma (C&D), squamous cell carcinoma (E&F) and small cell lung cancer (H&I), they have similar manifestations on CT scan (Black arrows point to lesions). OCT images demonstrated that normal lesion appeared homogeneous and had clear structure (B). In OCT images of malignant lesions (D F I), the lesions appear as unevenly distributed areas of high backscatter, resulting in the loss of layer structure and glandular tissue. Red arrows indicate the lesion areas.
Table 3.
Image features of normal, malignant and inflammatory airway tissues with significant p-values.