Lung tumor segmentation in MRI mice scans using 3D nnU-Net with minimum annotations

TL;DR

This work demonstrates the importance of 3D input over 2D input images for lung tumor segmentation in MRI scans, and demonstrates that the nnU-Net model outperforms the U-Net, U-Net3+, and DeepMeta models.

Abstract

In drug discovery, accurate lung tumor segmentation is an important step for assessing tumor size and its progression using \textit{in-vivo} imaging such as MRI. While deep learning models have been developed to automate this process, the focus has predominantly been on human subjects, neglecting the pivotal role of animal models in pre-clinical drug development. In this work, we focus on optimizing lung tumor segmentation in mice. First, we demonstrate that the nnU-Net model outperforms the U-Net, U-Net3+, and DeepMeta models. Most importantly, we achieve better results with nnU-Net 3D models than 2D models, indicating the importance of spatial context for segmentation tasks in MRI mice scans. This study demonstrates the importance of 3D input over 2D input images for lung tumor segmentation in MRI scans. Finally, we outperform the prior state-of-the-art approach that involves the combined segmentation of lungs and tumors within the lungs. Our work achieves comparable results using only lung tumor annotations requiring fewer annotations, saving time and annotation efforts. This work (https://anonymous.4open.science/r/lung-tumour-mice-mri-64BB) is an important step in automating pre-clinical animal studies to quantify the efficacy of experimental drugs, particularly in assessing tumor changes.

Figures (4)

• Figure 1: The schematic diagram summarizing our work on lung tumor segmentation. Three different data variants were acquired from public MRI dataset, pre-processed, and used for training one of the four U-Net-like architectures. The architectures differed in terms of loss functions, number of kernels, activation functions, batch-normalization layers, and full-scale skip connections. A - Activation; C - Convolutional Layer; D - Dropout; BN - Batch Normalization; SC - Skip Connections; L - Loss Function.
• Figure 2: Qualitative results for lung and tumor segmentation. Showing great performance of nnU-Net for both lung and tumor segmentation. Images were obtained without post-processing and were cropped for better visualization purposes.
• Figure 3: Qualitative results for tumor segmentation, indicating high performance of nnU-Net. Images are cropped for better visualization purposes.
• Figure 4: F1-score results on a sequence of slices for 2D and 3D nnU-Nets on a few slices and the full 3D stack of one image. 3D nnU-Net outperforming other models.