3D Segmentation with Exponential Logarithmic Loss for Highly Unbalanced Object Sizes
This work addresses the challenge of segmenting very small structures in 3D medical images, which is an incremental improvement for medical imaging applications.
The paper tackles the problem of 3D medical image segmentation with highly unbalanced object sizes by proposing a network architecture and an exponential logarithmic loss function, achieving an average Dice coefficient of 82% on brain segmentation with 20 labels and a smallest-to-largest object size ratio of 0.14%, with segmentation taking around 0.4 seconds per volume.
With the introduction of fully convolutional neural networks, deep learning has raised the benchmark for medical image segmentation on both speed and accuracy, and different networks have been proposed for 2D and 3D segmentation with promising results. Nevertheless, most networks only handle relatively small numbers of labels (<10), and there are very limited works on handling highly unbalanced object sizes especially in 3D segmentation. In this paper, we propose a network architecture and the corresponding loss function which improve segmentation of very small structures. By combining skip connections and deep supervision with respect to the computational feasibility of 3D segmentation, we propose a fast converging and computationally efficient network architecture for accurate segmentation. Furthermore, inspired by the concept of focal loss, we propose an exponential logarithmic loss which balances the labels not only by their relative sizes but also by their segmentation difficulties. We achieve an average Dice coefficient of 82% on brain segmentation with 20 labels, with the ratio of the smallest to largest object sizes as 0.14%. Less than 100 epochs are required to reach such accuracy, and segmenting a 128x128x128 volume only takes around 0.4 s.