Table 2 Application of the rare event detection algorithm to images from late-stage breast cancer subjects

From: Unsupervised detection of rare events in liquid biopsy assays

 

TPR (\(\bar{N}=2500\))

AUROC

Cell type

Mean

St. dev.

Min.

Max.

Mean

St. dev.

Min.

Max.

D∣CK

1.00

0.0

1.00

1.00

1.00

0.0

1.00

1.00

D∣CK∣V

1.00

0.0

1.00

1.00

1.00

1.00 × 10−4

1.00

1.00

D∣CK∣V∣CD

0.750

0.382

0.0

1.00

0.914

0.185

0.500

1.00

D∣V

0.389

0.448

0.0

1.00

0.978

0.0427

0.883

1.00

D∣V∣CD

0.798

0.339

0.0

1.00

0.991

0.0210

0.940

1.00

D∣CK∣CD

0.917

0.144

0.667

1.00

0.998

3.20 × 10−3

0.993

1.00

D-∣CK

1.00

0.0

1.00

1.00

1.00

1.00 × 10−4

1.00

1.00

All

0.746

0.265

0.0

1.00

0.982

0.0280

0.915

1.00

  1. Columns 2-5: statistics for the true positive rate for a cohort of 2500 rare tiles identified by the algorithm. Columns 6-9: statistics for the AUROC obtained by varying \(\bar{N}\) from 0 to N. Values are reported for different event types (Rows 1-7) and all types together (Row 8).