Labels in a haystack: Approaches beyond supervised learning in biomedical applications

Abstract

Summary Recent advances in biomedical machine learning demonstrate great potential for data-driven techniques in health care and biomedical research. However, this potential has thus far been hampered by both the scarcity of annotated data in the biomedical domain and the diversity of the domain's subfields. While unsupervised learning is capable of finding unknown patterns in the data by design, supervised learning requires human annotation to achieve the desired performance through training. With the latter performing vastly better than the former, the need for annotated datasets is high, but they are costly and laborious to obtain. This review explores a family of approaches existing between the supervised and the unsupervised problem setting. The goal of these algorithms is to make more efficient use of the available labeled data. The advantages and limitations of each approach are addressed and perspectives are provided.

Topic Segmentation of Research Article Collections

“…One possibility is to integrate knowledge from online classification systems of research works such as that of ACM. 5 Another possibility is to hire experts who could manually annotate hundreds of records, and then utilize semi-supervised learning techniques (Soleimani and Miller, 2016;Yakimovich et al, 2021) for scaling up to the entire collection.…”

“…To this end, we will try to integrate external knowledge from online publication libraries. Another possibility could be the manual annotation of a certain number of samples and the utilization of semi-supervised techniques for identifying the topic of all data collection samples (Soleimani and Miller, 2016;Yakimovich et al, 2021).…”

Reinforcing Interdisciplinary Collaborations To Unravel The Astrocyte "Calcium Code"

“…There is currently a major lack of annotations of astrocyte calcium activity images. Note that this is also true for other datasets of 3D images in live tissue [53]. For all of those reasons, a common and promising approach is to use realistic synthetic datasets with known ground-truths (i.e., all morphological and dynamical properties are known and controlled) to train and quantitatively assess the performance of analysis software.…”

Advocating For Interdisciplinary Collaborations To Unravel The Astrocyte "Calcium Code"

“…There is currently a major lack of annotations of astrocyte calcium activity images. Note that this is also true for other datasets of 3D images in live tissue [54]. For all of those reasons, a common and promising approach is to use realistic synthetic datasets with known ground-truths (i.e., all morphological and dynamical properties are known and controlled) to train and quantitatively assess the performance of analysis software.…”