Introducing image classification efficacies

Abstract

Accuracy assessment is essential in all image classification-related fields, ranging from molecular imaging to earth observation. However, existing accuracy metrics are too sensitive to class imbalance or lack explicit interpretations for assessing classification performance. Consequently, their scores may be misleading when they are applied to compare classification algorithms that address different image data sources. These limitations jeopardize the widespread application of deep learning classification methods for classifying different image types. We introduce the metrics of image classification efficacy from medicine and pharmacology to overcome the limitations of accuracy metrics. We include a baseline classification to derive the metrics of image classification efficacy and apply real-world and hypothetical examples to further examine their usefulness. Image classification efficacies can be applied at the map and class levels and for binary and multiclass classifications. The interpretability and comparability of image classification efficacies facilitate reliable classification method evaluation across data sources. We detail the procedures of classification efficacy assessment for image classification researchers and classifier users.