Radiomics-based risk prediction: survival in sarcopenic gastrointestinal and genitourinary cancer cohorts using a population-specific sarcopenia definition

OBJECTIVE: for females), along with other factors such as radiomic and clinical variables, could improve risk categorization for survival in sarcopenic, gastrointestinal (GI), and genitourinary (GU) cancer patients. MATERIALS AND METHODS: This retrospective study involved 2,015 patients who underwent major surgery at our institution between 2006 and 2018. From the axial CT image of the L3 vertebral region, the skeletal muscle was segmented to calculate the SMI values. 646 radiomic features were extracted from the image, and the Boruta algorithm was used to identify the most pertinent features. The final set of features consisted of these features and clinical variables. The dataset was split into training and testing sets comprising 1,512 patients (75%) and 503 patients (25%), respectively. Four models were evaluated in this study: (1) random survival forest (RSF)-based radiomic model, (2) Cox proportional hazard-based model, (3) Asian (Japanese)-specific SMI definition-based model, and (4) integrated model, which combines experiments (2) and (3). The Kaplan-Meier (KM) curves, log-rank tests, and Harrell's concordance index (C-index) were computed to assess the predictive performance of the models. RESULTS: The integrated model achieved the highest performance in terms of C-index (0.760, 95% CI: 0.686-0.835) among all other models. Survival stratification was statistically significant across all the experiments (p < 0.0001). CONCLUSION: This work demonstrates the efficacy of utilizing recently proposed Japanese population-based SMI definition values to optimize risk stratification for sarcopenic patients. KEY POINTS: Question Can population-specific sarcopenia definitions, clinical variables, and CT muscle radiomics together improve overall survival risk stratification in patients with gastrointestinal and genitourinary cancers? Findings In this cohort of 2,015 patients, combining sarcopenia, clinical variables, and CT muscle radiomics yielded an overall survival C-index of 0.760. Relevance statement The incorporation of the Asian (Japanese)-specific skeletal muscle index definition improves risk stratification for the survival of sarcopenic cancer patients when combined with an imaging-based approach, such as radiomics and other clinical parameters, highlighting the value of multimodal integration.

Authors

Institutions

Publication Details

Journal
European Radiology Experimental
Published
2026-11-13
DOI
https://doi.org/10.1186/s41747-026-00848-2
Primary Topic
Radiomics and Machine Learning in Medical Imaging
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Radiomics-based risk prediction: survival in sarcopenic gastrointestinal and genitourinary cancer cohorts using a population-specific sarcopenia definition

Haruki Kume, Kotaro Sugawara, Shohei Inui, Satoru Taguchi et al.
4 links
European Radiology Experimental
Radiomics and Machine Learning in Medical Imaging
article

Radiomics-based risk prediction: survival in sarcopenic gastrointestinal and genitourinary cancer cohorts using a population-specific sarcopenia definition

Haruki Kume, Kotaro Sugawara, Shohei Inui, Satoru Taguchi, Kenjiro Kishitani, T. Yoshikawa, Mariko Kurokawa, Wataru Gonoi, Yoshifumi Baba, Shouhei Hanaoka, Shinichiro Shiomi, Koichi Yagi, Nishta Letchumanan, Osamu Abe
article en

Abstract

OBJECTIVE: for females), along with other factors such as radiomic and clinical variables, could improve risk categorization for survival in sarcopenic, gastrointestinal (GI), and genitourinary (GU) cancer patients. MATERIALS AND METHODS: This retrospective study involved 2,015 patients who underwent major surgery at our institution between 2006 and 2018. From the axial CT image of the L3 vertebral region, the skeletal muscle was segmented to calculate the SMI values. 646 radiomic features were extracted from the image, and the Boruta algorithm was used to identify the most pertinent features. The final set of features consisted of these features and clinical variables. The dataset was split into training and testing sets comprising 1,512 patients (75%) and 503 patients (25%), respectively. Four models were evaluated in this study: (1) random survival forest (RSF)-based radiomic model, (2) Cox proportional hazard-based model, (3) Asian (Japanese)-specific SMI definition-based model, and (4) integrated model, which combines experiments (2) and (3). The Kaplan-Meier (KM) curves, log-rank tests, and Harrell's concordance index (C-index) were computed to assess the predictive performance of the models. RESULTS: The integrated model achieved the highest performance in terms of C-index (0.760, 95% CI: 0.686-0.835) among all other models. Survival stratification was statistically significant across all the experiments (p < 0.0001). CONCLUSION: This work demonstrates the efficacy of utilizing recently proposed Japanese population-based SMI definition values to optimize risk stratification for sarcopenic patients. KEY POINTS: Question Can population-specific sarcopenia definitions, clinical variables, and CT muscle radiomics together improve overall survival risk stratification in patients with gastrointestinal and genitourinary cancers? Findings In this cohort of 2,015 patients, combining sarcopenia, clinical variables, and CT muscle radiomics yielded an overall survival C-index of 0.760. Relevance statement The incorporation of the Asian (Japanese)-specific skeletal muscle index definition improves risk stratification for the survival of sarcopenic cancer patients when combined with an imaging-based approach, such as radiomics and other clinical parameters, highlighting the value of multimodal integration.

European Radiology ExperimentalVol. 10(1)
University of Tokyo Hospital (JP), The University of Tokyo (JP)
Japan Society for the Promotion of Science
Good health and well-being
Radiomics and Machine Learning in Medical Imaging
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon