Two-step artificial intelligence algorithm for liver segmentation automates anatomic virtual hepatectomy

BACKGROUND

Anatomic virtual hepatectomy with precise liver segmentation for hemilivers, sectors, or Couinaud's segments using conventional three-dimensional simulation is not automated and artificial intelligence (AI)-based algorithms have not yet been applied.

METHODS

Computed tomography data of 174 living-donor candidates for liver transplantation (training data) were used for developing a new two-step AI algorithm to automate liver segmentation that was validated in another 51 donors (validation data). The Pure-AI (no human intervention) and ground truth (GT, full human intervention) data groups were compared.

RESULTS

In the Pure-AI group, the median Dice coefficients of the right and left hemilivers were highly similar, 0.95 and 0.92, respectively; sectors, posterior to lateral: 0.86-0.92, and Couinaud's segments 1-8: 0.71-0.89. Labeling of the first-order branch as hemiliver, right or left portal vein perfectly matched; 92.8% of the second-order (sectors); 91.6% of third-order (segments) matched between the Pure-AI and GT data.

CONCLUSIONS

The two-step AI algorithm for liver segmentation automates anatomic virtual hepatectomy. The AI-based algorithm correctly divided all hemilivers, and more than 90% of the sectors and segments.

Automatic liver segmentation using 3D convolutional neural networks with a hybrid loss function

PURPOSE

Automatic liver segmentation from abdominal computed tomography (CT) images is a fundamental task in computer-assisted liver surgery programs. Many liver segmentation algorithms are very sensitive to fuzzy boundaries and heterogeneous pathologies, especially when the data are scarce. To solve these problems, we propose an automatic liver segmentation framework based on three-dimensional (3D) convolutional neural networks with a hybrid loss function.

METHODS

Two networks are incorporated in our method with the first being a liver shape autoencoder that is trained to obtain compressed codes of liver shapes, and the second being a liver segmentation network that is trained with a hybrid loss function. The design of the hybrid loss function is comprised of three parts. The first part is an adaptively weighted cross-entropy loss, which pays more attention to misclassified pixels. The second part is an edge-preserving smoothness loss, which guarantees that the adjacent pixels with the same label have similar outputs, while dissimilar for pixels with different labels. The third part of the loss is a shape constraint to model high-level structural differences based on the learned shape codes. Both networks use 3D operations for data processing. In our experiments, data augmentation is performed at both the training and the test stage.

RESULTS

We extensively evaluated our method on two datasets: the Segmentation of the Liver Competition 2007 (Sliver07), and the Combined (CT-MR) Healthy Abdominal Organ Segmentation (CHAOS) Challenge. Finally, with only 20 training scans, we achieved the best score of 82.55 on the Sliver07 challenge, and a score of 83.02 on the CHAOS challenge.

CONCLUSIONS

In this study, we proposed a novel hybrid loss to overcome the difficulties in liver segmentation. The quantitative and qualitative results demonstrate that our method is highly suited for pathological liver segmentation, even when trained with a small dataset.