I have limited experience, but I think that usually you would take this into account when building your loss function and heavily penalize false negatives during training.