Table 1.
NSCLC Patient and Tumor Characteristics.
Fig 1.
A 67-year-old woman with adenocarcinoma of the right upper lobe.
The outline of the tumor is drawn on the axial low-b value DW image (b = 0 s/mm2) (A). The tumor's boundaries are meticulously identified with reference to the coronal and sagittal reformatted DW images (B, C). The data acquired from each slice are summed to generate volumes of interest (VOIs) (D). The contours of the VOI are automatically copied to the exact same location of the corresponding ADC maps.
Table 2.
ADC Histogram Parameters of Pathologic Grades of NSCLC.
Table 3.
ADC Histogram Parameters of High- and Low-Grade of NSCLC.
Fig 2.
ROC curves of percentiles of ADC in predicting high-grade.
The AUC was highest for the 95th percentile ADC (AUC = 0.74, cut-off value of 1634.1 × 10−6 mm2/sec, sensitivity 84.6%, specificity 66.7%).
Table 4.
ADC Histogram Parameters of Lymphovascular Invasion and Pleural Invasion of NSCLC.
Fig 3.
ROC curves of percentiles of ADC (A), kurtosis and skewness (B) in predicting lymphovascular invasion. The AUC was highest for the kurtosis (AUC = 0.809, cut-off value of 1.0815×10-6mm2/sec, sensitivity 61.2%, specificity 90.9%) in predicting lymphovascular invasion.
Fig 4.
ROC curve of skewness in predicting pleural invasion.
The AUC of the skewness was 0.648 (cut-off value 0.824×10−6 mm2/sec, sensitivity 60.0%, specificity 73.3%) in predicting pleural invasion.