Benchmarking Deep Learning for Future Liver Remnant Segmentation in Colorectal Liver Metastasis

Anthony T. Wu; Arghavan Rezvani; Kela Liu; Pooya Khosravi; Roozbeh Houshyar; Whitney Li; Xiaohui Xie

arxiv: 2604.07999 · v1 · submitted 2026-04-09 · 💻 cs.LG

Benchmarking Deep Learning for Future Liver Remnant Segmentation in Colorectal Liver Metastasis

Anthony T. Wu , Arghavan Rezvani , Kela Liu , Roozbeh Houshyar , Pooya Khosravi , Whitney Li , Xiaohui Xie This is my paper

Pith reviewed 2026-05-10 17:58 UTC · model grok-4.3

classification 💻 cs.LG

keywords future liver remnant segmentationcolorectal liver metastasesCT imagingdeep learning benchmarknnU-NetSTU-Netcascaded segmentationsurgical planning

0 comments

The pith

Refining labels on 197 public CT volumes creates the first validated benchmark for future liver remnant segmentation, with cascaded nnU-Net reaching 0.767 Dice.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper fills a critical data gap for automated segmentation of the future liver remnant, the portion surgeons must preserve during operations to remove colorectal liver metastases. Poor segmentation risks post-hepatectomy liver failure, a major surgical complication. The authors manually refine all 197 volumes in the existing CRLM-CT-Seg dataset to produce accurate ground-truth labels that capture complex boundaries and lesions. They then run the first systematic comparisons of cascaded and end-to-end deep learning pipelines using nnU-Net, SwinUNETR, and STU-Net. Results show cascaded nnU-Net performs best on the final remnant task while a pretrained STU-Net leads on metastasis detection and tolerates upstream errors more effectively, supplying a reproducible starting point for AI surgical planning tools.

Core claim

By manually refining all 197 volumes from the public CRLM-CT-Seg dataset to produce high-fidelity, validated ground-truth labels, the authors create the first open-source benchmark for future liver remnant segmentation. They establish baselines showing that a cascaded nnU-Net achieves the highest FLR Dice score of 0.767 while a pretrained STU-Net delivers superior CRLM segmentation at 0.620 Dice and markedly greater robustness when used in cascaded pipelines.

What carries the argument

The manually refined CRLM-CT-Seg dataset with expert-validated labels, deployed in cascaded (liver to CRLM to FLR) and end-to-end segmentation pipelines across nnU-Net, SwinUNETR, and STU-Net models.

Load-bearing premise

The authors' manual refinement of every volume produces ground-truth labels that accurately trace complex resection boundaries, vasculature, and diffuse lesions without introducing systematic bias.

What would settle it

Independent radiologists re-annotating a random subset of the 197 volumes and obtaining average Dice overlap below 0.80 with the refined labels, or public code failing to reproduce the reported 0.767 FLR Dice score on the released dataset.

read the original abstract

Accurate segmentation of the future liver remnant (FLR) is critical for surgical planning in colorectal liver metastases (CRLM) to prevent fatal post-hepatectomy liver failure. However, this segmentation task is technically challenging due to complex resection boundaries, convoluted hepatic vasculature and diffuse metastatic lesions. A primary bottleneck in developing automated AI tools has been the lack of high-fidelity, validated data. We address this gap by manually refining all 197 volumes from the public CRLM-CT-Seg dataset, creating the first open-source, validated benchmark for this task. We then establish the first segmentation baselines, comparing cascaded (Liver->CRLM->FLR) and end-to-end (E2E) strategies using nnU-Net, SwinUNETR, and STU-Net. We find a cascaded nnU-Net achieves the best final FLR segmentation Dice (0.767), while the pretrained STU-Net provides superior CRLM segmentation (0.620 Dice) and is significantly more robust to cascaded errors. This work provides the first validated benchmark and a reproducible framework to accelerate research in AI-assisted surgical planning.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper refines an existing CT dataset into the first open benchmark for future liver remnant segmentation and runs some standard model baselines, but the manual label refinements lack any reported validation metrics.

read the letter

The core offering here is taking the public CRLM-CT-Seg volumes and manually refining all 197 of them to produce what they describe as the first validated open benchmark specifically for future liver remnant segmentation in colorectal liver metastasis cases. They then compare cascaded pipelines against end-to-end ones using nnU-Net, SwinUNETR, and STU-Net, with the best reported FLR Dice at 0.767 and some notes on robustness to errors in the cascaded path. That dataset release and the straightforward baseline comparisons are the actual new pieces; prior work had the raw volumes but not this targeted refinement for the FLR task. It is useful for anyone building tools for surgical planning where accurate remnant volume matters for avoiding liver failure. The numbers give a concrete starting point for people who want to test new architectures on this exact clinical segmentation problem. The soft spot is the manual refinement itself. The abstract states they refined every volume but supplies no inter-rater agreement scores, no overlap statistics between the original and refined masks, and no description of how difficult cases with diffuse lesions or convoluted vessels were handled. Without those checks it is hard to judge whether the ground truth is stable enough for the reported Dice scores to be taken as reliable benchmarks. Label noise could easily move those numbers up or down. This work is aimed at medical imaging groups working on liver oncology and surgical AI. A reader who needs an open FLR dataset with some off-the-shelf model numbers will find it practical. It is worth sending to peer review because the dataset creation addresses a real data gap and the experiments are reproducible in principle, but referees will need to see the missing validation details before the benchmark can be treated as solid.

Referee Report

1 major / 1 minor

Summary. The paper manually refines all 197 volumes of the public CRLM-CT-Seg dataset to create the first open-source benchmark for future liver remnant (FLR) segmentation in colorectal liver metastases (CRLM). It compares cascaded (Liver→CRLM→FLR) and end-to-end strategies with nnU-Net, SwinUNETR, and STU-Net, reporting that cascaded nnU-Net achieves the highest FLR Dice of 0.767 while pretrained STU-Net yields superior CRLM segmentation (0.620 Dice) and greater robustness to cascaded errors.

Significance. If the refined labels accurately delineate complex resection planes, vasculature, and diffuse lesions, the work supplies the first validated public benchmark for a clinically high-stakes task, enabling reproducible progress on AI tools that could reduce post-hepatectomy liver failure risk.

major comments (1)

[Methods] Methods section: the manual refinement of the 197 volumes is presented as producing a 'validated benchmark,' yet no quantitative checks are reported (inter-rater agreement, Dice overlap between original and refined masks, multi-expert overlap on difficult cases, or time-per-volume statistics). Because the central claim of benchmark superiority rests on label fidelity for convoluted FLR boundaries, the absence of these metrics leaves the reported Dice scores (0.767 FLR, 0.620 CRLM) vulnerable to unquantified label noise.

minor comments (1)

[Abstract] Abstract: the reported Dice scores are given without accompanying dataset split sizes, number of training/validation/test cases, or statistical significance tests between models.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback, which highlights an important aspect of benchmark validation. We address the major comment point by point below.

read point-by-point responses

Referee: [Methods] Methods section: the manual refinement of the 197 volumes is presented as producing a 'validated benchmark,' yet no quantitative checks are reported (inter-rater agreement, Dice overlap between original and refined masks, multi-expert overlap on difficult cases, or time-per-volume statistics). Because the central claim of benchmark superiority rests on label fidelity for convoluted FLR boundaries, the absence of these metrics leaves the reported Dice scores (0.767 FLR, 0.620 CRLM) vulnerable to unquantified label noise.

Authors: We agree that additional quantitative details on the refinement process would strengthen the presentation of the benchmark. In the revised manuscript we will expand the Methods section to report (i) average time per volume for refinement, (ii) Dice overlap between the original public CRLM-CT-Seg masks and the refined versions (which we have computed), and (iii) a more explicit description of the single-expert protocol used. We will also add a dedicated limitations paragraph that acknowledges the absence of inter-rater agreement and multi-expert overlap statistics, explains that refinement was performed by one experienced radiologist for consistency with surgical planning criteria, and discusses the potential impact of residual label noise on the reported Dice scores. These changes directly mitigate the concern about unquantified label fidelity without requiring new multi-rater annotations. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical benchmarking with manual dataset refinement and model evaluation

full rationale

The paper performs direct empirical work: manual refinement of an existing public dataset (CRLM-CT-Seg) to create ground-truth labels, followed by training and evaluating standard segmentation models (nnU-Net, SwinUNETR, STU-Net) under cascaded and end-to-end strategies. Reported metrics are Dice scores from held-out test volumes. No equations, fitted parameters, predictions derived from inputs, uniqueness theorems, or self-citations appear in the derivation chain. All results are falsifiable measurements on the refined data; the central claims rest on experimental outcomes rather than any reduction to prior assumptions or self-referential definitions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an applied empirical benchmarking study that introduces no new free parameters, mathematical axioms, or invented entities; it relies on standard deep learning architectures applied to a refined public dataset.

pith-pipeline@v0.9.0 · 5526 in / 1305 out tokens · 73234 ms · 2026-05-10T17:58:36.635791+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We address this gap by manually refining all 197 volumes from the public CRLM-CT-Seg dataset, creating the first open-source, validated benchmark... cascaded nnU-Net achieves the best final FLR segmentation Dice (0.767)
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

comparing cascaded (Liver→CRLM→FLR) and end-to-end (E2E) strategies using nnU-Net, SwinUNETR, and STU-Net

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

23 extracted references · 23 canonical work pages · 1 internal anchor

[1]

How- ever, this segmentation task is technically challenging due to complex resection boundaries, convoluted hepatic vasculature and diffuse metastatic lesions

ABSTRACT Accurate segmentation of the future liver remnant (FLR) is critical for surgical planning in colorectal liver metastases (CRLM) to prevent fatal post-hepatectomy liver failure. How- ever, this segmentation task is technically challenging due to complex resection boundaries, convoluted hepatic vasculature and diffuse metastatic lesions. A primary ...

work page
[2]

vanishing

INTRODUCTION Colorectal cancer (CRC) is the second leading cause of cancer-related mortality, with nearly two million new cases reported annually [1]. Colorectal liver metastases (CRLM) is a major driver of this mortality, affecting∼50%of CRC pa- tients [2, 3] and is the principal cause of death in nearly half of those cases [4]. Surgical liver resection,...

work page
[3]

Benchmarking Deep Learning for Future Liver Remnant Segmentation in Colorectal Liver Metastasis

MATERIALS AND METHODS 3.1 Dataset and Manual Refinement:We manually refined all 197 volumetric segmentations from the CRLM-CT-Seg dataset [12] using ITK-Snap [13]. All refinements were per- formed by a radiological researcher with final refinements confirmed by an abdominal radiologist with over ten years of subspecialty experience (Fig. 1a). Refinement f...

work page internal anchor Pith review Pith/arXiv arXiv doi:10.5281/zenodo.17574862 2026
[4]

Pipelined

RESULTS AND DISCUSSION We present 5-fold CV results to demonstrate model stability and held-out test set results to establish final baseline perfor- mance (Tables 1 and 2, respectively). 4.1 Baseline Performance:The CV results (Table 1) demon- strate low standard deviations, confirming training stability. STU-Net achieved the highest mean Dice across most...

work page
[5]

We es- tablish the first segmentation baselines for FLR prediction, demonstrating that a cascaded approach is slightly superior to an E2E one

CONCLUSION AND FUTURE DIRECTIONS In this work, we present the first fully-manual, high-fidelity segmentation dataset for CRLM and FLR analysis. We es- tablish the first segmentation baselines for FLR prediction, demonstrating that a cascaded approach is slightly superior to an E2E one. Pretrained models like STU-Net show high per- formance and greater rob...

work page
[6]

ACKNOWLEDGMENTS The authors report no conflicts of interest

work page
[7]

Ethical approval was not required as confirmed by the license attached with the open access data

COMPLIANCE WITH ETHICAL STANDARDS This research study was conducted retrospectively using hu- man subject data made available in open access by CRLM- CT-Seg [12]. Ethical approval was not required as confirmed by the license attached with the open access data

work page
[8]

Global cancer statis- tics 2022: Globocan estimates of incidence and mortal- ity worldwide for 36 cancers in 185 countries,

Freddie Bray, Mathieu Laversanne, Hyuna Sung, Jacques Ferlay, Rebecca L Siegel, Isabelle Soerjo- mataram, and Ahmedin Jemal, “Global cancer statis- tics 2022: Globocan estimates of incidence and mortal- ity worldwide for 36 cancers in 185 countries,”CA: a cancer journal for clinicians, vol. 74, no. 3, pp. 229– 263, 2024

work page 2022
[9]

New trends in surgery for colorectal liver metastasis,

Philipp Kron and Peter Lodge, “New trends in surgery for colorectal liver metastasis,”Annals of Gastroentero- logical Surgery, vol. 8, no. 4, pp. 553–565, 2024

work page 2024
[10]

Ar- tificial intelligence in the diagnosis and management of colorectal cancer liver metastases,

Gianluca Rompianesi, Francesca Pegoraro, Carlo DL Ceresa, Roberto Montalti, and Roberto Ivan Troisi, “Ar- tificial intelligence in the diagnosis and management of colorectal cancer liver metastases,”World Journal of Gastroenterology, vol. 28, no. 1, pp. 108, 2022

work page 2022
[11]

Cause of death from liver metastases in colorectal cancer,

Thomas S Helling and Magdeline Martin, “Cause of death from liver metastases in colorectal cancer,”Annals of surgical oncology, vol. 21, no. 2, pp. 501–506, 2014

work page 2014
[12]

Col- orectal liver metastases: An update on multidisciplinary approach,

Felix Che-Lok Chow and Kenneth Siu-Ho Chok, “Col- orectal liver metastases: An update on multidisciplinary approach,”World journal of hepatology, vol. 11, no. 2, pp. 150, 2019

work page 2019
[13]

Pushing the limits of surgical resection in col- orectal liver metastasis: how far can we go?,

Francisco Calderon Novoa, Victoria Ardiles, Eduardo de Santibanes, Juan Pekolj, Jeremias Goransky, Oscar Mazza, Rodrigo S ´anchez Claria, and Martin de San- tibanes, “Pushing the limits of surgical resection in col- orectal liver metastasis: how far can we go?,”Cancers, vol. 15, no. 7, pp. 2113, 2023

work page 2023
[14]

Artifi- cial intelligence in hepatology, liver surgery and trans- plantation: Emerging applications and frontiers of re- search,

Fadl H Veerankutty, Govind Jayan, Manish Kumar Ya- dav, Krishnan Sarojam Manoj, Abhishek Yadav, Sindhu Radha Sadasivan Nair, TU Shabeerali, Varghese Yeldho, Madhu Sasidharan, and Shiraz Ahmad Rather, “Artifi- cial intelligence in hepatology, liver surgery and trans- plantation: Emerging applications and frontiers of re- search,”World Journal of Hepatology...

work page 1977
[15]

Contribution of 3d virtual modeling in locating hepatic metastases, particularly “vanishing tumors

Mike Salavracos, Etienne Danse, Nicolas Michoux, Alexandre de Hemptinne, Tancr `ede De Poortere, and Laurent Coubeau, “Contribution of 3d virtual modeling in locating hepatic metastases, particularly “vanishing tumors”: a pilot study,”Artificial Intelligence Surgery, vol. 4, no. 4, pp. 331–347, 2024

work page 2024
[16]

Post-hepatectomy liver failure,

Rondi Kauffmann and Yuman Fong, “Post-hepatectomy liver failure,”Hepatobiliary surgery and nutrition, vol. 3, no. 5, pp. 238, 2014

work page 2014
[17]

Novel machine learning algorithm can identify patients at risk of poor overall survival following curative resection for colorectal liver metastases,

Iakovos Amygdalos, Gustav M ¨uller-Franzes, Jan Bed- narsch, Zoltan Czigany, Tom Florian Ulmer, Philipp Bruners, Christiane Kuhl, Ulf Peter Neumann, Daniel Truhn, and Sven Arke Lang, “Novel machine learning algorithm can identify patients at risk of poor overall survival following curative resection for colorectal liver metastases,”Journal of Hepato-Bilia...

work page 2023
[18]

Meta-analysis of the prognostic role of peri- operative platelet count in posthepatectomy liver failure and mortality,

Arianeb Mehrabi, Mohammad Golriz, Elias Khajeh, Omid Ghamarnejad, Pascal Probst, Hamidreza Fonouni, Sara Mohammadi, Karl Heinz Weiss, and Markus W B¨uchler, “Meta-analysis of the prognostic role of peri- operative platelet count in posthepatectomy liver failure and mortality,”Journal of British Surgery, vol. 105, no. 10, pp. 1254–1261, 2018

work page 2018
[19]

Preoperative ct and survival data for patients undergoing resection of col- orectal liver metastases,

Amber L Simpson, Jacob Peoples, John M Creasy, Gabor Fichtinger, Natalie Gangai, Krishna N Ke- shavamurthy, Andras Lasso, Jinru Shia, Michael I D’Angelica, and Richard KG Do, “Preoperative ct and survival data for patients undergoing resection of col- orectal liver metastases,”Scientific Data, vol. 11, no. 1, pp. 172, 2024

work page 2024
[20]

User-guided 3D active contour segmen- tation of anatomical structures: Significantly improved efficiency and reliability,

Paul A. Yushkevich, Joseph Piven, Heather Cody Ha- zlett, Rachel Gimpel Smith, Sean Ho, James C. Gee, and Guido Gerig, “User-guided 3D active contour segmen- tation of anatomical structures: Significantly improved efficiency and reliability,”Neuroimage, vol. 31, no. 3, pp. 1116–1128, 2006

work page 2006
[21]

nnu-net: a self- configuring method for deep learning-based biomedical image segmentation,

Fabian Isensee, Paul F Jaeger, Simon AA Kohl, Jens Petersen, and Klaus H Maier-Hein, “nnu-net: a self- configuring method for deep learning-based biomedical image segmentation,”Nature methods, vol. 18, no. 2, pp. 203–211, 2021

work page 2021
[22]

Swin unetr: Swin transformers for semantic segmentation of brain tumors in mri images,

Ali Hatamizadeh, Vishwesh Nath, Yucheng Tang, Dong Yang, Holger R Roth, and Daguang Xu, “Swin unetr: Swin transformers for semantic segmentation of brain tumors in mri images,” inInternational MICCAI brain- lesion workshop. Springer, 2021, pp. 272–284

work page 2021
[23]

Huang, H

Ziyan Huang, Haoyu Wang, Zhongying Deng, Jin Ye, Yanzhou Su, Hui Sun, Junjun He, Yun Gu, Lixu Gu, Shaoting Zhang, et al., “Stu-net: Scalable and trans- ferable medical image segmentation models empowered by large-scale supervised pre-training,”arXiv preprint arXiv:2304.06716, 2023

work page arXiv 2023

[1] [1]

How- ever, this segmentation task is technically challenging due to complex resection boundaries, convoluted hepatic vasculature and diffuse metastatic lesions

ABSTRACT Accurate segmentation of the future liver remnant (FLR) is critical for surgical planning in colorectal liver metastases (CRLM) to prevent fatal post-hepatectomy liver failure. How- ever, this segmentation task is technically challenging due to complex resection boundaries, convoluted hepatic vasculature and diffuse metastatic lesions. A primary ...

work page

[2] [2]

vanishing

INTRODUCTION Colorectal cancer (CRC) is the second leading cause of cancer-related mortality, with nearly two million new cases reported annually [1]. Colorectal liver metastases (CRLM) is a major driver of this mortality, affecting∼50%of CRC pa- tients [2, 3] and is the principal cause of death in nearly half of those cases [4]. Surgical liver resection,...

work page

[3] [3]

Benchmarking Deep Learning for Future Liver Remnant Segmentation in Colorectal Liver Metastasis

MATERIALS AND METHODS 3.1 Dataset and Manual Refinement:We manually refined all 197 volumetric segmentations from the CRLM-CT-Seg dataset [12] using ITK-Snap [13]. All refinements were per- formed by a radiological researcher with final refinements confirmed by an abdominal radiologist with over ten years of subspecialty experience (Fig. 1a). Refinement f...

work page internal anchor Pith review Pith/arXiv arXiv doi:10.5281/zenodo.17574862 2026

[4] [4]

Pipelined

RESULTS AND DISCUSSION We present 5-fold CV results to demonstrate model stability and held-out test set results to establish final baseline perfor- mance (Tables 1 and 2, respectively). 4.1 Baseline Performance:The CV results (Table 1) demon- strate low standard deviations, confirming training stability. STU-Net achieved the highest mean Dice across most...

work page

[5] [5]

We es- tablish the first segmentation baselines for FLR prediction, demonstrating that a cascaded approach is slightly superior to an E2E one

CONCLUSION AND FUTURE DIRECTIONS In this work, we present the first fully-manual, high-fidelity segmentation dataset for CRLM and FLR analysis. We es- tablish the first segmentation baselines for FLR prediction, demonstrating that a cascaded approach is slightly superior to an E2E one. Pretrained models like STU-Net show high per- formance and greater rob...

work page

[6] [6]

ACKNOWLEDGMENTS The authors report no conflicts of interest

work page

[7] [7]

Ethical approval was not required as confirmed by the license attached with the open access data

COMPLIANCE WITH ETHICAL STANDARDS This research study was conducted retrospectively using hu- man subject data made available in open access by CRLM- CT-Seg [12]. Ethical approval was not required as confirmed by the license attached with the open access data

work page

[8] [8]

Global cancer statis- tics 2022: Globocan estimates of incidence and mortal- ity worldwide for 36 cancers in 185 countries,

Freddie Bray, Mathieu Laversanne, Hyuna Sung, Jacques Ferlay, Rebecca L Siegel, Isabelle Soerjo- mataram, and Ahmedin Jemal, “Global cancer statis- tics 2022: Globocan estimates of incidence and mortal- ity worldwide for 36 cancers in 185 countries,”CA: a cancer journal for clinicians, vol. 74, no. 3, pp. 229– 263, 2024

work page 2022

[9] [9]

New trends in surgery for colorectal liver metastasis,

Philipp Kron and Peter Lodge, “New trends in surgery for colorectal liver metastasis,”Annals of Gastroentero- logical Surgery, vol. 8, no. 4, pp. 553–565, 2024

work page 2024

[10] [10]

Ar- tificial intelligence in the diagnosis and management of colorectal cancer liver metastases,

Gianluca Rompianesi, Francesca Pegoraro, Carlo DL Ceresa, Roberto Montalti, and Roberto Ivan Troisi, “Ar- tificial intelligence in the diagnosis and management of colorectal cancer liver metastases,”World Journal of Gastroenterology, vol. 28, no. 1, pp. 108, 2022

work page 2022

[11] [11]

Cause of death from liver metastases in colorectal cancer,

Thomas S Helling and Magdeline Martin, “Cause of death from liver metastases in colorectal cancer,”Annals of surgical oncology, vol. 21, no. 2, pp. 501–506, 2014

work page 2014

[12] [12]

Col- orectal liver metastases: An update on multidisciplinary approach,

Felix Che-Lok Chow and Kenneth Siu-Ho Chok, “Col- orectal liver metastases: An update on multidisciplinary approach,”World journal of hepatology, vol. 11, no. 2, pp. 150, 2019

work page 2019

[13] [13]

Pushing the limits of surgical resection in col- orectal liver metastasis: how far can we go?,

Francisco Calderon Novoa, Victoria Ardiles, Eduardo de Santibanes, Juan Pekolj, Jeremias Goransky, Oscar Mazza, Rodrigo S ´anchez Claria, and Martin de San- tibanes, “Pushing the limits of surgical resection in col- orectal liver metastasis: how far can we go?,”Cancers, vol. 15, no. 7, pp. 2113, 2023

work page 2023

[14] [14]

Artifi- cial intelligence in hepatology, liver surgery and trans- plantation: Emerging applications and frontiers of re- search,

Fadl H Veerankutty, Govind Jayan, Manish Kumar Ya- dav, Krishnan Sarojam Manoj, Abhishek Yadav, Sindhu Radha Sadasivan Nair, TU Shabeerali, Varghese Yeldho, Madhu Sasidharan, and Shiraz Ahmad Rather, “Artifi- cial intelligence in hepatology, liver surgery and trans- plantation: Emerging applications and frontiers of re- search,”World Journal of Hepatology...

work page 1977

[15] [15]

Contribution of 3d virtual modeling in locating hepatic metastases, particularly “vanishing tumors

Mike Salavracos, Etienne Danse, Nicolas Michoux, Alexandre de Hemptinne, Tancr `ede De Poortere, and Laurent Coubeau, “Contribution of 3d virtual modeling in locating hepatic metastases, particularly “vanishing tumors”: a pilot study,”Artificial Intelligence Surgery, vol. 4, no. 4, pp. 331–347, 2024

work page 2024

[16] [16]

Post-hepatectomy liver failure,

Rondi Kauffmann and Yuman Fong, “Post-hepatectomy liver failure,”Hepatobiliary surgery and nutrition, vol. 3, no. 5, pp. 238, 2014

work page 2014

[17] [17]

Novel machine learning algorithm can identify patients at risk of poor overall survival following curative resection for colorectal liver metastases,

Iakovos Amygdalos, Gustav M ¨uller-Franzes, Jan Bed- narsch, Zoltan Czigany, Tom Florian Ulmer, Philipp Bruners, Christiane Kuhl, Ulf Peter Neumann, Daniel Truhn, and Sven Arke Lang, “Novel machine learning algorithm can identify patients at risk of poor overall survival following curative resection for colorectal liver metastases,”Journal of Hepato-Bilia...

work page 2023

[18] [18]

Meta-analysis of the prognostic role of peri- operative platelet count in posthepatectomy liver failure and mortality,

Arianeb Mehrabi, Mohammad Golriz, Elias Khajeh, Omid Ghamarnejad, Pascal Probst, Hamidreza Fonouni, Sara Mohammadi, Karl Heinz Weiss, and Markus W B¨uchler, “Meta-analysis of the prognostic role of peri- operative platelet count in posthepatectomy liver failure and mortality,”Journal of British Surgery, vol. 105, no. 10, pp. 1254–1261, 2018

work page 2018

[19] [19]

Preoperative ct and survival data for patients undergoing resection of col- orectal liver metastases,

Amber L Simpson, Jacob Peoples, John M Creasy, Gabor Fichtinger, Natalie Gangai, Krishna N Ke- shavamurthy, Andras Lasso, Jinru Shia, Michael I D’Angelica, and Richard KG Do, “Preoperative ct and survival data for patients undergoing resection of col- orectal liver metastases,”Scientific Data, vol. 11, no. 1, pp. 172, 2024

work page 2024

[20] [20]

User-guided 3D active contour segmen- tation of anatomical structures: Significantly improved efficiency and reliability,

Paul A. Yushkevich, Joseph Piven, Heather Cody Ha- zlett, Rachel Gimpel Smith, Sean Ho, James C. Gee, and Guido Gerig, “User-guided 3D active contour segmen- tation of anatomical structures: Significantly improved efficiency and reliability,”Neuroimage, vol. 31, no. 3, pp. 1116–1128, 2006

work page 2006

[21] [21]

nnu-net: a self- configuring method for deep learning-based biomedical image segmentation,

Fabian Isensee, Paul F Jaeger, Simon AA Kohl, Jens Petersen, and Klaus H Maier-Hein, “nnu-net: a self- configuring method for deep learning-based biomedical image segmentation,”Nature methods, vol. 18, no. 2, pp. 203–211, 2021

work page 2021

[22] [22]

Swin unetr: Swin transformers for semantic segmentation of brain tumors in mri images,

Ali Hatamizadeh, Vishwesh Nath, Yucheng Tang, Dong Yang, Holger R Roth, and Daguang Xu, “Swin unetr: Swin transformers for semantic segmentation of brain tumors in mri images,” inInternational MICCAI brain- lesion workshop. Springer, 2021, pp. 272–284

work page 2021

[23] [23]

Huang, H

Ziyan Huang, Haoyu Wang, Zhongying Deng, Jin Ye, Yanzhou Su, Hui Sun, Junjun He, Yun Gu, Lixu Gu, Shaoting Zhang, et al., “Stu-net: Scalable and trans- ferable medical image segmentation models empowered by large-scale supervised pre-training,”arXiv preprint arXiv:2304.06716, 2023

work page arXiv 2023