4MO - A novel AI solution for auto-segmentation of multi-origin liver neoplasms

Background

Accurate, swift and reproducible liver lesion segmentation would benefit radiologists and oncologists with enhanced detection and diagnosis, treatment planning, clinical decision making, and monitoring.

Methods

The LiTS-Challenge dataset [1] containing 91 CT scans with different types of liver lesions was used for training. A U-Net based model adapted with three-channel input and aggregated residual transformations [2] to extract important features was trained on the axial slices. The first channel input contained slice with intensities clipped to the interval [375,425] HU. The second channel input contained slice with intensities normalized between -1000 and 400 HU. As an input to the third channel, histogram equalization was applied on the normalized axial slice. The model was optimized based on its predictions at multiple resolution levels (512, 256 and 128) which encouraged the network to predict correctly not only at the last layer, but also at deep layers with low resolution output. The performance of the model was evaluated on an external validation dataset from Maastricht University Medical Centre (MUMC) containing 170 patients with liver metastases of colorectal origin. The reference ground truth delineation was provided by a trained medical doctor.

Results

The proposed model achieved an average dice score coefficient (DSC) of 0.72, an average sensitivity of 85% and an average False Positive Rate (FPR) <1% compared to manual delineation. Automated segmentation took on an average, 10s per case on a GTX 1070 GPU while it took around 15 min for manual segmentation.

Conclusions

We present a novel auto-segmentation solution based on routine clinical imaging of metastatic liver cancer patients. The performance of the network is in line with human experts. The solution increases efficiency via swifter reporting times, especially in clinical and image analysis workflows. Finally, this automated solution can be used to obtain information on the changes in the tumor size in response to treatment which can be compared with RECIST estimates used in clinical practice. [1] P. Bilic et al., “The Liver Tumor Segmentation Benchmark (LiTS),” 2019. [2] S. Xie, R. Girshick, P. Dollár, Z. Tu, K. He, and U. San Diego, “Aggregated Residual Transformations for Deep Neural Networks.”.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

Oncoradiomics.

Funding

PREDICT Grant.

Disclosure

A. Vaidyanathan: Research grant/Funding (self), PREDICT: Oncoradiomics. F. Zerka: Research grant/Funding (institution), PREDICT: Oncoradiomics. H.C. Woodruff: Shareholder/Stockholder/Stock options: Oncoradiomics. R. Leijenaar: Shareholder/Stockholder/Stock options: Oncoradiomics. W. Vos: Shareholder/Stockholder/Stock options: Oncoradiomics. S. Walsh: Research grant/Funding (self), Research grant/Funding (institution), Shareholder/Stockholder/Stock options: Oncoradiomics. P. Lambin: Research grant/Funding (self), Research grant/Funding (institution), Shareholder/Stockholder/Stock options: Oncoradiomics. All other authors have declared no conflicts of interest.