Page 76 - AIH-2-2

Artificial Intelligence in Health Improved liver tumor segmentation with dense networks

4. Conclusion Availability of data

In this article, we present novel, end-to-end FCNs Data are available from the corresponding author upon
designed for liver tumor segmentation in CT volumes. reasonable request.
Extensive experiments were conducted on two public
datasets. Our method achieved an average improvement References
of 5.9% in the Dice score and 6% in the Lesion Dice 1. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA,
score compared to previous methods, demonstrating the Jemal A. Global cancer statistics 2018: GLOBOCAN
effectiveness of our approach in segmenting multiscale estimates of incidence and mortality worldwide for 36 cancers
tumors and more accurately delineating tumor in 185 countries. CA Cancer J Clin. 2018;68(6):394-424.
boundaries. Moreover, our model uses the raw CT scans doi: 10.3322/caac.21492
as input without the need for separate pre-processing 2. Dou Q, Yu L, Chen H, et al. 3D deeply supervised network
procedures and directly outputs the segmentation map. for automated segmentation of volumetric medical images.
Our method can be easily extended to other medical Med Image Anal. 2017;41:40-54.
image segmentation tasks within the CT modality.
In future work, we will focus on the segmentation of doi: 10.1016/j.media.2017.05.001
small tumors, which affects the eventual segmentation 3. Seo H, Huang C, Bassenne M, Xiao R, Xing L. Modified
performance largely. Considering the generalization U-Net (mU-Net) With incorporation of object-dependent
of computer-aided diagnosis, we will explore the high level features for improved liver and liver-tumor
application of our improved pre-processing module to segmentation in CT images. IEEE Trans Med Imaging.
2020;39(5):1316-1325.
other image modalities such as MRI and other diverse
segmentation tasks. doi: 10.1109/TMI.2019.2948320
4. Lippi G, Banfi G, Church S, et al. Preanalytical quality
Acknowledgments improvement. In pursuit of harmony, on behalf of European

We would like to express our gratitude to our guide, Federation for Clinical Chemistry and Laboratory Medicine
Dr. Peng, Associate Professor, School of Electronic (EFLM) Working group for preanalytical phase (WG-PRE).
Clin Chem Lab Med CCLM. 2015;53(3):357-370.
Information and Communications, Huazhong University
of Science and Technology, for his invaluable guidance doi: 10.1515/cclm-2014-1051
and support throughout the course of this research. His 5. Hawkins R. Managing the pre- and post-analytical phases of
expertise and insightful feedback significantly contributed the total testing process. Ann Lab Med. 2012;32(1):5-16.
to the development of this work. doi: 10.3343/alm.2012.32.1.5
Funding 6. Compton CC, Robb JA, Anderson MW, et al. Preanalytics and
precision pathology: Pathology practices to ensure molecular
None. integrity of cancer patient biospecimens for precision
medicine. Arch Pathol Lab Med. 2019;143(11):1346-1363.
Conflict of interest
doi: 10.5858/arpa.2019-0009-SA
The authors declare they have no competing interests.
7. Gündisch S, Hauck S, Sarioglu H, et al. Variability of protein
Author contributions and phosphoprotein levels in clinical tissue specimens during
the preanalytical phase. J Proteome Res. 2012;11(12):5748-5762.
Conceptualization: Shen Huang doi: 10.1021/pr300560y
Formal analysis: Yiyang Zhang, Ziyang Huang
Investigation: Shanshan Zhao 8. Moghbel M, Mashohor S, Mahmud R, Saripan MIB. Review
Methodology: Yizhuo Xu of liver segmentation and computer assisted detection/
diagnosis methods in computed tomography. Artif Intell
Writing – original draft: Shen Huang, Yizhuo Xu Rev. 2018;50(4):497-537.
Writing – review & editing: Yizhuo Xu, Shen Huang
doi: 10.1007/s10462-017-9550-x
Ethics approval and consent to participate 9. Yin S, You X, Yang X, Peng Q, Zhu Z, Jing XY. A joint space-
Not applicable. angle regularization approach for single 4D diffusion image
super-resolution. Magn Reson Med. 2018;80(5):2173-2187.
Consent for publication doi: 10.1002/mrm.27184
Not applicable. 10. Soler L, Delingette H, Malandain G, et al. Fully automatic

Volume 2 Issue 2 (2025) 70 doi: 10.36922/aih.5001