Related Experiment Video
Updated: Jul 19, 2026
13:45
Hybrid µCT-FMT imaging and image analysis
Published on: June 4, 2015
13.8K
Radiomics and deep learning fusion model based on multiphasic CT for predicting HER2 expression status in bladder
Wenbang Pan1, Liu Yang1, Lin Yang1
1Department of General Surgery, Dalian Second People's Hospital, Dalian, China.
Translational Andrology and Urology
|April 13, 2026
Summary
A new tri-phasic CT imaging model accurately predicts Human Epidermal growth factor Receptor 2 (HER2) expression in bladder cancer. This noninvasive approach aids in selecting patients for targeted antibody-drug conjugate (ADC) therapy.
Area of Science:
- Oncology
- Radiology
- Medical Imaging
Background:
- Human Epidermal growth factor Receptor 2 (HER2) overexpression is a critical therapeutic target in bladder urothelial carcinoma (BLCA), particularly for antibody-drug conjugates (ADCs).
- Current immunohistochemistry (IHC)-based HER2 assessment in BLCA faces limitations due to intratumoral heterogeneity and sampling bias.
- There is a significant need for noninvasive, reliable imaging-based methods to predict HER2 expression status in BLCA.
Purpose of the Study:
- To develop and validate a noninvasive, imaging-based multimodal model for predicting HER2 expression status in BLCA.
- To utilize tri-phasic computed tomography (CT) imaging features for HER2 status prediction.
- To assess the performance of the multimodal model against single-modality approaches.
Main Methods:
- A cohort of 411 patients with BLCA from three institutions underwent tri-phasic contrast-enhanced CT and HER2 IHC.
- Patients were categorized as HER2-positive (IHC 2+/3+) or HER2-negative (IHC 0/1+).
- Radiomics, 2.5D/3D ResNet50 deep learning, clinical, and multimodal fusion models were developed and evaluated using ROC analysis, calibration curves, and decision curve analysis.
Main Results:
- Tumor architecture, hydronephrosis, pelvic pain, and lymph node status were identified as independent predictors of HER2 positivity.
- The 2.5D ResNet50 deep learning model achieved an external AUC of 0.827, outperforming the 3D model (AUC=0.608).
- The multimodal fusion model demonstrated superior performance with external AUCs of 0.873, accuracy 0.819, sensitivity 0.857, and specificity 0.784, significantly outperforming single-modality models.
Conclusions:
- A tri-phasic CT-based multimodal model provides a noninvasive method for assessing HER2 expression status in BLCA.
- This imaging-based model shows promise for identifying patients who may benefit from HER2-targeted ADC therapy.
- The multimodal approach offers a more accurate and reliable prediction of HER2 status compared to individual imaging or clinical factors.
