A Data-Adaptive Loss Function for Incomplete Data and Incremental Learning in Semantic Image Segmentation

Abstract

In the last years, deep learning has dramatically improved the performances in a variety of medical image analysis applications. Among different types of deep learning models, convolutional neural networks have been among the most successful and they have been used in many applications in medical imaging.Training deep convolutional neural networks often requires large amounts of image data to generalize well to new unseen images. It is often time-consuming and expensive to collect large amounts of data in the medical image domain due to expensive imaging systems, and the need for experts to manually make ground truth annotations. A potential problem arises if new structures are added when a decision support system is already deployed and in use. Since the field of radiation therapy is constantly developing, the new structures would also have to be covered by the decision support system.In the present work, we propose a novel loss function to solve multiple problems: imbalanced datasets, partiallylabeled data, and incremental learning. The proposed loss function adapts to the available data in order to utilize all available data, even when some have missing annotations. We demonstrate that the proposed loss function also works well in an incremental learning setting, where an existing model is easily adapted to semi-automatically incorporate delineations of new organs when they appear. Experiments on a large in-house dataset show that the proposed method performs on par with baseline models, while greatly reducing the training time and eliminating the hassle of maintaining multiple models in practice.

Compressing the Activation Maps in Deep Convolutional Neural Networks and the Regularization Effect of Compression

“…We employed the soft Sørensen-Dice coefficient (DSC) loss in the segmentation tasks [13,55,56,57,58], defined as,…”

Ecosystem Health of the Beiyun River Basin (Beijing, China) as Evaluated by the Analytic Hierarchy Process Method

“…Rule Reference class-balanced selecting samples that covering every individual class [26], [103] loss-based selecting samples based on the highest, lowest or median value of the cross-entropy loss. [104] entropy-based the prediction uncertainty is estimated, selecting samples with the lowest, the highest uncertainty, and samples close to the average uncertainty.…”

Anomaly sound detection of industrial devices by using teacher-student incremental continual learning

“…Compared to traditional machine learning, where the entire model needs to be retrained, incremental learning allows training only a small portion of the model on new data, thus saving significant time and computational resources. Incremental learning is frequently applied in scenarios such as real-time data analysis [36], online prediction [37], adaptive systems [38,39], where the data is constantly changing, necessitating timely updates to the model to adapt to new data features. Incremental learning allows the model to swiftly adapt to new devices, increasing its flexibility and real-time capabilities, and thereby improving the model's accuracy and usefulness.…”