Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Development and External Validation of a Combined Clinical–Radiomic Model for Predicting Insufficient Hypertrophy of the Future Liver Remnant following Portal Vein Embolization

Wang, Qiang ; Brismar, Torkel B. ; Björk, Dennis ; Baubeta, Erik LU orcid ; Lindell, Gert LU ; Björnsson, Bergthor and Sparrelid, Ernesto (2024) In Annals of Surgical Oncology
Abstract
Objectives
This study aimed to develop and externally validate a model for predicting insufficient future liver remnant (FLR) hypertrophy after portal vein embolization (PVE) based on clinical factors and radiomics of pretreatment computed tomography (CT)

Patients and methods
Clinical information and CT scans of 241 consecutive patients from three Swedish centers were retrospectively collected. One center (120 patients) was applied for model development, and the other two (59 and 62 patients) as test cohorts. Logistic regression analysis was adopted for clinical model development. A FLR radiomics signature was constructed from the CT images using the support vector machine. A model combining clinical factors and FLR... (More)
Objectives
This study aimed to develop and externally validate a model for predicting insufficient future liver remnant (FLR) hypertrophy after portal vein embolization (PVE) based on clinical factors and radiomics of pretreatment computed tomography (CT)

Patients and methods
Clinical information and CT scans of 241 consecutive patients from three Swedish centers were retrospectively collected. One center (120 patients) was applied for model development, and the other two (59 and 62 patients) as test cohorts. Logistic regression analysis was adopted for clinical model development. A FLR radiomics signature was constructed from the CT images using the support vector machine. A model combining clinical factors and FLR radiomics signature was developed. Area under the curve (AUC) was adopted for predictive performance evaluation

Results
Three independent clinical factors were identified for model construction: pretreatment standardized FLR (odds ratio (OR): 1.12, 95% confidence interval (CI): 1.04–1.20), alanine transaminase (ALT) level (OR: 0.98, 95% CI: 0.97–0.99), and PVE material (OR: 0.27, 95% CI: 0.08–0.87). This clinical model showed an AUC of 0.75, 0.71, and 0.68 in the three cohorts, respectively. A total of 833 radiomics features were extracted, and after feature dimension reduction, 16 features were selected for FLR radiomics signature construction. When adding it to the clinical model, the AUC of the combined model increased to 0.80, 0.76, and 0.72, respectively. However, the increase was not significant.

Conclusions
Pretreatment CT radiomics showed added value to the clinical model for predicting FLR hypertrophy following PVE. Although not reaching statistically significant, the evolving radiomics holds a potential to supplement traditional predictors of FLR hypertrophy. (Less)
Please use this url to cite or link to this publication:
author
; ; ; ; ; and
organization
publishing date
type
Contribution to journal
publication status
epub
subject
in
Annals of Surgical Oncology
publisher
Springer
external identifiers
  • pmid:39658716
  • scopus:85211794449
ISSN
1534-4681
DOI
10.1245/s10434-024-16592-z
language
English
LU publication?
yes
id
8f82135d-9c59-4be4-97ce-a6323a7a129b
date added to LUP
2024-12-12 22:20:15
date last changed
2025-04-04 14:16:23
@article{8f82135d-9c59-4be4-97ce-a6323a7a129b,
  abstract     = {{Objectives<br/>This study aimed to develop and externally validate a model for predicting insufficient future liver remnant (FLR) hypertrophy after portal vein embolization (PVE) based on clinical factors and radiomics of pretreatment computed tomography (CT)<br/><br/>Patients and methods<br/>Clinical information and CT scans of 241 consecutive patients from three Swedish centers were retrospectively collected. One center (120 patients) was applied for model development, and the other two (59 and 62 patients) as test cohorts. Logistic regression analysis was adopted for clinical model development. A FLR radiomics signature was constructed from the CT images using the support vector machine. A model combining clinical factors and FLR radiomics signature was developed. Area under the curve (AUC) was adopted for predictive performance evaluation<br/><br/>Results<br/>Three independent clinical factors were identified for model construction: pretreatment standardized FLR (odds ratio (OR): 1.12, 95% confidence interval (CI): 1.04–1.20), alanine transaminase (ALT) level (OR: 0.98, 95% CI: 0.97–0.99), and PVE material (OR: 0.27, 95% CI: 0.08–0.87). This clinical model showed an AUC of 0.75, 0.71, and 0.68 in the three cohorts, respectively. A total of 833 radiomics features were extracted, and after feature dimension reduction, 16 features were selected for FLR radiomics signature construction. When adding it to the clinical model, the AUC of the combined model increased to 0.80, 0.76, and 0.72, respectively. However, the increase was not significant.<br/><br/>Conclusions<br/>Pretreatment CT radiomics showed added value to the clinical model for predicting FLR hypertrophy following PVE. Although not reaching statistically significant, the evolving radiomics holds a potential to supplement traditional predictors of FLR hypertrophy.}},
  author       = {{Wang, Qiang and Brismar, Torkel B. and Björk, Dennis and Baubeta, Erik and Lindell, Gert and Björnsson, Bergthor and Sparrelid, Ernesto}},
  issn         = {{1534-4681}},
  language     = {{eng}},
  publisher    = {{Springer}},
  series       = {{Annals of Surgical Oncology}},
  title        = {{Development and External Validation of a Combined Clinical–Radiomic Model for Predicting Insufficient Hypertrophy of the Future Liver Remnant following Portal Vein Embolization}},
  url          = {{http://dx.doi.org/10.1245/s10434-024-16592-z}},
  doi          = {{10.1245/s10434-024-16592-z}},
  year         = {{2024}},
}