Machine learning with asymmetric abstention for biomedical decision-making

Mariem Gandouz,Hajo Holzmann,Dominik Heider

doi:10.1186/s12911-021-01655-y

Mariem Gandouz, Hajo Holzmann + Show 1 more

Open Access

https://doi.org/10.1186/s12911-021-01655-y

Copy DOI

Abstract

Machine learning and artificial intelligence have entered biomedical decision-making for diagnostics, prognostics, or therapy recommendations. However, these methods need to be interpreted with care because of the severe consequences for patients. In contrast to human decision-making, computational models typically make a decision also with low confidence. Machine learning with abstention better reflects human decision-making by introducing a reject option for samples with low confidence. The abstention intervals are typically symmetric intervals around the decision boundary. In the current study, we use asymmetric abstention intervals, which we demonstrate to be better suited for biomedical data that is typically highly imbalanced. We evaluate symmetric and asymmetric abstention on three real-world biomedical datasets and show that both approaches can significantly improve classification performance. However, asymmetric abstention rejects as many or fewer samples compared to symmetric abstention and thus, should be used in imbalanced data.

Highlights

Machine learning (ML) and artificial intelligence (AI) have entered many areas of life and will pave the way for a new era in biomedicine
Efficient, and novel method to optimally build an asymmetric type of abstaining binary classifiers using an asymmetric abstention interval around the intersection between the two distributions of positive samples and negative samples based on Pareto optimization, similar to the approach proposed by Herbei and Wegkamp [15]
In order to improve the Matthews correlation coefficient (MCC), the samples are rejected, which is significant for the symmetric abstention on imbalanced data

Summary

Introduction

Machine learning (ML) and artificial intelligence (AI) have entered many areas of life and will pave the way for a new era in biomedicine These methods can improve medical treatment or diagnosis, identify novel subtypes, or provide new insights into survival prognostics. Biomedical decision support systems based on ML and AI have entered many different studies and biomedical fields, e.g., Oncology [4], Pathology [10, 21, 35], Diabetes [6, 27], Human Genetics [20], and Infectious Diseases [14, 19, 28] as part of a growing trend towards precision medicine. There is great potential for biomedical decision support systems based on ML and AI techniques and they have become key players in disease diagnostics, prognostics, and therapeutics [34].

Objectives

Methods

Results

Conclusion