Table 2.
Performance parametera | Tumor resections evaluated (n) | Percentage pass | Parameter pass/fail |
---|---|---|---|
Analytical specificityb | 40 | 90% (36/40) | Pass |
Analytical sensitivityc | 40 | 100% (40/40) | Pass |
Accuracyd | 36 | r = 0.74, p value < 0.0001 | Pass |
Precisione | 12 | 92% (11/12) | Pass |
aDetails of the pre-defined acceptance criteria are described in the methods.
bAlgorithm accurately identified cells as objects, accurately classified tumor and stromal compartments of the tissue, and accurately classified DKK1 negative cells.
cAlgorithm accurately identified cells when present and classified DKK1 positive cells correctly.
dPearson correlation (p value < 0.05) with DKK1 manual H-scores.
eResults across 3 separate staining days within the same binned category of expression (negative bin: H-score = 0, low bin: H-score < 34, and high bin: H-score ≥ 35). In cases of discordant binning of negative, low or high categories, a ± 20 point H-score discrepancy was still considered acceptable.