Abstract
Deep learning increasingly accelerates biomedical research, deploying neural networks for multiple tasks, such as image classification, object detection, and semantic segmentation. However, neural networks are commonly trained supervised on large-scale, labeled datasets. These prerequisites raise issues in biomedical image recognition, as datasets are generally small-scale, challenging to obtain, expensive to label, and frequently heterogeneously labeled. Furthermore, heterogeneous labels are a challenge for supervised methods. If not all classes are labeled for an individual sample, supervised deep learning approaches can only learn on a subset of the dataset with common labels for each individual sample; consequently, biomedical image recognition engineers need to be frugal concerning their label and ground truth requirements. This paper discusses the effects of frugal labeling and proposes to train neural networks for multi-class semantic segmentation on heterogeneously labeled data based on a novel objective function. The objective function combines a class asymmetric loss with the Dice loss. The approach is demonstrated for training on the sparse ground truth of a heterogeneous labeled dataset, training within a transfer learning setting, and the use-case of merging multiple heterogeneously labeled datasets. For this purpose, a biomedical small-scale, multi-class semantic segmentation dataset is utilized. The heartSeg dataset is based on the medaka fish’s position as a cardiac model system. Automating image recognition and semantic segmentation enables high-throughput experiments and is essential for biomedical research. Our approach and analysis show competitive results in supervised training regimes and encourage frugal labeling within biomedical image recognition.
Highlights
Today, biomedical image recognition and semantic segmentation are successfully driven by deep learning approaches and supervised training of neural networks [1]
The novel approaches to deploying heterogeneous labels within semantic segmentation have been presented and examined in multiple experiments, being common use-cases within biomedical semantic segmentation
A combined objective function for tasks with heterogeneous labels is introduced to cope with and utilize heterogeneous labels during neural network training, ready to be adapted to further semantic segmentation tasks and datasets
Summary
Biomedical image recognition and semantic segmentation are successfully driven by deep learning approaches and supervised training of neural networks [1]. Multi-class semantic segmentation on heterogeneous labels segmentation neural networks [3, 4]. Deep symmetric convolutional neural networks, such as V-net [5], and U-Net [6] dominate the algorithmic state-of-the-art. In supervised deep learning approaches, these architectures are driven by large-scale, homogeneous labeled datasets [7–9]. Homogeneity means that the ground truth for each sample of the dataset includes a label for each of the specified classes. Heterogeneity defines a ground truth in which different samples might miss labels for a present class or include labels for specific classes only. As a result, supervised training on heterogeneous data means reducing the number of available data samples for training, limiting the task’s scope to a reduced set of classes, or conclusively an unfeasible training
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
