Fig. 6: Performance of the DNN trained exclusively with high- and standard-resolution scans.
From: Harnessing deep learning to detect bronchiolitis obliterans syndrome from chest CT
The performance of the network trained with high-resolution scans (slice thickness ≤ 1.25 mm, lung kernel) (a), thin-slice scans (slice thickness ≤ 1.25 mm, standard kernel) (b) and standard-resolution scans (slice thickness > 1.25 mm, standard kernel) (c), and tested on scans of each type separately, is shown in respective panels. For reference, performance of the DNN trained with all scans is shown in each panel in blue. In each subfigure, the Left plot shows the aggregated ROC curves. The Right plot shows the AUC attained in individual splits. The box extends from the first quartile (Q1) to the third quartile (Q3) of the data, with a line at the median. The whiskers extend from the box to the farthest data point lying within 1.5x the interquartile range (IQR) from the box. Flier points are those past the end of the whiskers. The number of data points for each box plot is 5 (n = 5). BOS bronchiolitis obliterans syndrome.
