Table 7 The mean/std test-set confusion matrix for the 10-fold cross validation test-set results with videos labelled N (Section II.A2) excluded. Positives correspond to frames from LUS videos that contain signatures of consolidations (and associated imaging patterns) while negatives correspond to frames that do not. One of the 10 folds, which contained only healthy frames, was excluded from the results. Here Y* frames are frames Y* videos (Section II.A2).
Method | Mean/SD (%) with Y* frames excluded from the test set | |||
|---|---|---|---|---|
TP | FP | FN | TN | |
Frame-based | 5.68 ± 4.66 | 5.93 ± 7.18 | 4.08 ± 5.52 | 84.31 ± 28.82 |
All-or-nothing binary video based | 6.75 ± 3.90 | 9.74 ± 13.80 | 3.01 ± 3.46 | 80.50 ± 31.80 |
Sampled binary video based | 6.73 ± 3.98 | 9.51 ± 11.43 | 3.03 ± 3.52 | 80.73 ± 30.94 |
Sampled quaternary video based | 7.59 ± 4.08 | 8.16 ± 11.72 | 2.17 ± 3.42 | 82.08 ± 33.69 |
