Skip to main content

Table 1 Performance of the developed methods for recognizing off-sample ion images as evaluated on the gold standard of 23,238 ion images showing F1-score (F1), precision (P), and recall (R). For each measure, we show the average and confidence intervals (+ − two standard deviations) over five folds of the cross validation

From: OffsampleAI: artificial intelligence approach to recognize off-sample mass spectrometry images

 

off-sample

on-sample

F1

P

R

F1

P

R

Deep residual learning

.97 (+/−.01)

.96 (+/−.03)

.98 (+/−.03)

.96 (+/−.04)

.97 (+/−.03)

.94 (+/−.07)

Semi-automated spatio-molecular biclustering, clusters curated for 2 datasets

.96 (+/−.03)

.96 (+/−.07)

.96 (+/−.04)

.97 (+/−.01)

.97 (+/−.03)

.97 (+/−.03)

Spatio-molecular biclustering

.93 (+/−.10)

.92 (+/−.10)

.94 (+/−.11)

.95 (+/−.06)

.95 (+/−.06)

.95 (+/−.06)

Molecular co-localization

.90 (+/− .07)

.95 (+/− .08)

.86 (+/− .15)

.93 (+/− .05)

.91 (+/− .11)

.96 (+/− .07)