In EDS ansehen

Enhancing prediction accuracy for muscle invasion in bladder cancer using a dual-energy CT-based interpretable model incorporating habitat radiomics and deep learning.

Gespeichert in:
Bibliographische Detailangaben
Titel: Enhancing prediction accuracy for muscle invasion in bladder cancer using a dual-energy CT-based interpretable model incorporating habitat radiomics and deep learning.
Autoren: Du C; Department of Radiology, First Affiliated Hospital of Dalian Medical University, Dalian, China., Wei W; Department of Radiology, First Affiliated Hospital of Dalian Medical University, Dalian, China., Hu M; Department of Radiology, First Affiliated Hospital of Dalian Medical University, Dalian, China., He J; Department of Radiology, First Affiliated Hospital of Dalian Medical University, Dalian, China., Shen J; Department of Radiology, First Affiliated Hospital of Dalian Medical University, Dalian, China., Liu Y; Department of Radiology, First Affiliated Hospital of Dalian Medical University, Dalian, China., Li J; CT Research, GE Healthcare, Dalian, China., Liu L; Department of Urology, First Affiliated Hospital of Dalian Medical University, Dalian, China. liuleidmu1989@163.com.
Quelle: BMC cancer [BMC Cancer] 2025 Dec 01; Vol. 25 (1), pp. 1842. Date of Electronic Publication: 2025 Dec 01.
Publikationsart: Journal Article
Sprache: English
Info zur Zeitschrift: Publisher: BioMed Central Country of Publication: England NLM ID: 100967800 Publication Model: Electronic Cited Medium: Internet ISSN: 1471-2407 (Electronic) Linking ISSN: 14712407 NLM ISO Abbreviation: BMC Cancer Subsets: MEDLINE
Imprint Name(s): Original Publication: London : BioMed Central, [2001-
MeSH-Schlagworte: Urinary Bladder Neoplasms*/pathology , Urinary Bladder Neoplasms*/diagnostic imaging , Urinary Bladder Neoplasms*/surgery , Tomography, X-Ray Computed*/methods , Deep Learning*, Humans ; Male ; Female ; Retrospective Studies ; Middle Aged ; Aged ; Neoplasm Invasiveness/diagnostic imaging ; ROC Curve ; Aged, 80 and over ; Radiomics
Abstract: Background: Accurately predicting muscle invasion status in bladder cancer (BCa) is essential for developing personalized treatment plans. This study integrates habitat analysis and 2.5-dimension (2.5D) deep learning (DL) to create an interpretable model using dual-energy computed tomography (DECT) images to enhance the preoperative assessment of muscle invasion status in BCa.
Methods: This study retrospectively recruited 200 BCa patients who underwent DECT urography and divided them into a training cohort (n = 140) and a test cohort (n = 60). Quantitative parameters from DECT images were measured, and independent predictors of muscle invasion were identified through stepwise regression analysis to develop a DECT model. Radiomics and 2.5D DL models were constructed using iodine-based material decomposition (IMD) images. Tumors were classified into distinct sub-regions using K-means clustering. Through feature extraction and selection, a traditional radiomics model, a habitat model, four 2.5D DL models, and the integrated model combining above models were developed. The predictive performance of the models was assessed using receiver operating characteristic (ROC) curve analysis and calibration curve analysis. The SHAP method was utilized to interpret the optimal model and visualize its decision-making process.
Results: The effective atomic number (Z eff ) was recognized as an independent predictor of muscle invasion in BCa. The integrated model, combining Z eff , habitat features, and ResNet 101-based DL features (DECT-DLH), represented the optimal model for predicting muscle invasion in BCa. The area under the curve (AUC) using the optima model was 0.981 (95% CI: 0.963-0.998) in the training cohort and 0.874 (95% CI: 0.785-0.964) in the test cohort. Calibration curves validated the model's reliability for clinical applications. Additionally, SHAP elucidated the decision-making processes associated with the model's predicted outcomes.
Conclusion: The integrated model, combining DECT quantitative parameters, habitat features, and 2.5D DL features, enhances the accuracy of preoperative predictions for muscle invasion status in BCa. The SHAP methodology improves the interpretability of decision-making within the model and provides valuable support for clinicians in developing personalized treatment strategies.
(© 2025. The Author(s).)
References: Cancer Imaging. 2025 Mar 12;25(1):31. (PMID: 40075494)
Front Physiol. 2022 Aug 25;13:948767. (PMID: 36091379)
Insights Imaging. 2023 Sep 19;14(1):151. (PMID: 37726599)
J Urol. 2024 Apr;211(4):533-538. (PMID: 38265030)
Abdom Radiol (NY). 2024 Nov;49(11):3883-3892. (PMID: 38937340)
Front Oncol. 2021 Feb 11;10:610691. (PMID: 33643912)
Insights Imaging. 2023 Jul 16;14(1):125. (PMID: 37454355)
Eur Radiol. 2020 Jun;30(6):3295-3305. (PMID: 32055949)
BMJ. 2024 Feb 12;384:e076743. (PMID: 38346808)
BMC Med Imaging. 2024 Apr 27;24(1):98. (PMID: 38678222)
BMC Med Imaging. 2024 Sep 16;24(1):245. (PMID: 39285354)
J Cancer Res Clin Oncol. 2021 Dec;147(12):3757-3767. (PMID: 33839938)
Tumori. 2023 Apr;109(2):148-156. (PMID: 35442120)
Nat Rev Cancer. 2018 Sep;18(9):576-585. (PMID: 29891961)
Eur Urol. 2024 Jan;85(1):17-31. (PMID: 37858453)
Clin Cancer Res. 2021 Jan 1;27(1):237-245. (PMID: 33028594)
Korean J Radiol. 2012 Nov-Dec;13(6):702-10. (PMID: 23118568)
Eur Urol. 2019 Jul;76(1):57-58. (PMID: 31000355)
EClinicalMedicine. 2023 Apr 04;58:101934. (PMID: 37090441)
PLoS One. 2021 Sep 15;16(9):e0257195. (PMID: 34525114)
J Hepatol. 2022 Jun;76(6):1348-1361. (PMID: 35589255)
J Clin Med. 2020 Aug 04;9(8):. (PMID: 32759779)
Cancer Imaging. 2024 Aug 15;24(1):107. (PMID: 39148139)
J Urol. 2016 Oct;196(4):1021-9. (PMID: 27317986)
Int J Surg. 2024 Feb 01;110(2):1039-1051. (PMID: 37924497)
Acad Radiol. 2025 Feb;32(2):821-833. (PMID: 39455346)
Cancer Imaging. 2025 Feb 26;25(1):19. (PMID: 40011960)
Cancer. 2018 Dec 15;124(24):4633-4649. (PMID: 30383900)
BMC Pulm Med. 2024 Oct 30;24(1):545. (PMID: 39478525)
Front Oncol. 2022 Dec 05;12:1019749. (PMID: 36544709)
Acad Radiol. 2024 Jun;31(6):2346-2355. (PMID: 38057182)
Eur J Radiol. 2025 Apr;185:112014. (PMID: 40031377)
Quant Imaging Med Surg. 2024 Jul 1;14(7):4617-4634. (PMID: 39022292)
Eur Urol. 2022 Nov;82(5):469-482. (PMID: 35965208)
Int J Cancer. 2022 Dec 15;151(12):2229-2243. (PMID: 36095154)
Radiology. 2018 Jul;288(1):26-35. (PMID: 29714680)
Ann Oncol. 2022 Mar;33(3):244-258. (PMID: 34861372)
Eur Radiol. 2022 Dec;32(12):8737-8747. (PMID: 35678859)
Clin Radiol. 2021 Aug;76(8):627.e23-627.e29. (PMID: 33985770)
Front Cell Dev Biol. 2022 Mar 02;10:833578. (PMID: 35309916)
Cancer Imaging. 2020 Jul 3;20(1):43. (PMID: 32620153)
Cancer Immunol Immunother. 2024 Jun 4;73(8):153. (PMID: 38833187)
EBioMedicine. 2019 Jun;44:289-297. (PMID: 31129097)
Contributed Indexing: Keywords: Bladder cancer; Computed tomography; Deep learning; Habitat radiomics; SHAP
Entry Date(s): Date Created: 20251201 Date Completed: 20251202 Latest Revision: 20251205
Update Code: 20251205
PubMed Central ID: PMC12667133
DOI: 10.1186/s12885-025-15107-7
PMID: 41327160
Datenbank: MEDLINE
Beschreibung
Abstract:Background: Accurately predicting muscle invasion status in bladder cancer (BCa) is essential for developing personalized treatment plans. This study integrates habitat analysis and 2.5-dimension (2.5D) deep learning (DL) to create an interpretable model using dual-energy computed tomography (DECT) images to enhance the preoperative assessment of muscle invasion status in BCa.<br />Methods: This study retrospectively recruited 200 BCa patients who underwent DECT urography and divided them into a training cohort (n = 140) and a test cohort (n = 60). Quantitative parameters from DECT images were measured, and independent predictors of muscle invasion were identified through stepwise regression analysis to develop a DECT model. Radiomics and 2.5D DL models were constructed using iodine-based material decomposition (IMD) images. Tumors were classified into distinct sub-regions using K-means clustering. Through feature extraction and selection, a traditional radiomics model, a habitat model, four 2.5D DL models, and the integrated model combining above models were developed. The predictive performance of the models was assessed using receiver operating characteristic (ROC) curve analysis and calibration curve analysis. The SHAP method was utilized to interpret the optimal model and visualize its decision-making process.<br />Results: The effective atomic number (Z <subscript>eff</subscript> ) was recognized as an independent predictor of muscle invasion in BCa. The integrated model, combining Z <subscript>eff</subscript> , habitat features, and ResNet 101-based DL features (DECT-DLH), represented the optimal model for predicting muscle invasion in BCa. The area under the curve (AUC) using the optima model was 0.981 (95% CI: 0.963-0.998) in the training cohort and 0.874 (95% CI: 0.785-0.964) in the test cohort. Calibration curves validated the model's reliability for clinical applications. Additionally, SHAP elucidated the decision-making processes associated with the model's predicted outcomes.<br />Conclusion: The integrated model, combining DECT quantitative parameters, habitat features, and 2.5D DL features, enhances the accuracy of preoperative predictions for muscle invasion status in BCa. The SHAP methodology improves the interpretability of decision-making within the model and provides valuable support for clinicians in developing personalized treatment strategies.<br /> (© 2025. The Author(s).)
ISSN:1471-2407
DOI:10.1186/s12885-025-15107-7