Neoadjuvant chemotherapy response prediction using pretreatment diffusion and contrast-enhanced magnetic resonance imaging with clinical variables

Pablo García Marcos; Paula Puerta González; Guillermo Lorenzo; Héctor Gómez; Covadonga del Camino; Adán Rodríguez; Ignacio Peláez; Angel Rio-Alvarez

cs.AIAug 18, 2026

Neoadjuvant chemotherapy response prediction using pretreatment diffusion and contrast-enhanced magnetic resonance imaging with clinical variables

Authors: Pablo García Marcos, Paula Puerta González, Guillermo Lorenzo, Héctor Gómez, Covadonga del Camino, Adán Rodríguez, Ignacio Peláez, Angel Rio-Alvarez,

Organizations: Computer Science Department, University of Oviedo, Asturias, Spain · Biomedical Engineering Center (BME), University of Oviedo, Asturias, Spain · Electrical Engineering Department, University of Oviedo, Asturias, Spain · Group of Numerical Methods in Engineering, Department of Mathematics, University of A Coruña, A Coruña, Spain · Oden Institute for Computational Engineering and Sciences, The University of Texas at Austin, Austin, Texas, USA · School of Mechanical Engineering, Purdue University, West Lafayette, Indiana, USA · Weldon School of Biomedical Engineering, Purdue University, West Lafayette, Indiana, USA · Purdue Center for Cancer Research, Purdue University, West Lafayette, Indiana, USA · Radiodiagnostic Service, Asturias Central University Hospital (HUCA) · Oncology Service, Cabueñes University Hospital (CAHU), Asturias, Spain

Abstract

Prediction of pathological complete response before neoadjuvant chemotherapy may facilitate more tailored therapeutic planning for breast cancer patients. This work proposes a deep-learning model for pretreatment data only, combining apparent diffusion coefficient maps, dynamic contrast-enhanced magnetic resonance imaging, and clinical variables. The study uses the public ACRIN 6698/I-SPY2 multicenter dataset. The architecture employs EfficientNet-B0 pretrained encoders for image feature extraction and late fusion with clinical information. Multiple clinical variables were evaluated, including age, race, histological type, HR/HER2 subtype, SBR grade, and maximum diameter. Only HR/HER2 subtype improved the average area under the receiver operating characteristic curve (AUC) and was retained in the final model. Using stratified five-fold cross-validation, standalone apparent diffusion coefficient maps achieved a mean AUC of 0.79, whereas dynamic contrast-enhanced magnetic resonance imaging achieved 0.74. Adding HR/HER2 subtype improved performance to 0.83 and 0.81, respectively. The final configuration, using both imaging modalities and HR/HER2 subtype, achieved an AUC of 0.86. These results support pretreatment multimodal learning for response prediction, although external validation is required before clinical use.

Figures & tables

Figure 1: Diagram of baseline preprocessing and multimodal fusion. All modalities are restricted to a comparable tumor patch.

Figure 2: Fusion architecture used by the multimodal configurations.

Input	Mean AUC $\pm$ SD	Accuracy (%) $\pm$ SD	F1 $\pm$ SD
ADC	$0.7936\pm 0.0940$	$75.2406\pm 8.1322$	$0.7570\pm 0.0793$
DCE-MRI	$0.7428\pm 0.0943$	$70.6085\pm 10.4020$	$0.7107\pm 0.1007$
ADC + HR/HER2 subtype	$0.8297\pm 0.0721$	$74.6346\pm 6.7194$	$0.7499\pm 0.0654$
DCE-MRI + HR/HER2 subtype	$0.8054\pm 0.0751$	$73.5714\pm 3.9633$	$0.7371\pm 0.0428$
ADC + DCE-MRI + HR/HER2 subtype	$0.8567\pm 0.0643$	$76.6667\pm 4.3979$	$0.7641\pm 0.0438$

Table 1: Average cross-validation performance using five folds. Rows including HR/HER2 subtype report results with that variable incorporated; rows without it correspond to imaging-only configurations.

Study	Dataset	Input data	Prediction time point	AUC	Main comment
[ 19 ]	ACRIN 6698	DW-MRI/ADC	Mid-/post-treatment	0.60 / 0.61	Monitoring biomarker; not baseline-only prediction
[ 12 ]	BMMR2 / ACRIN 6698-derived	DCE-MRI and/or DW-MRI, optional clinical variables	Mostly post-baseline	0.803–0.840	Challenge setting using information after treatment onset
[ 3 ]	ACRIN 6698 + external cohort	T1-DCE + DW + ADC + clinicopathological variables	Pre- and early-treatment longitudinal MRI	0.903 / 0.861	Not baseline-only; includes external testing
[ 6 ]	BMMR2 / ACRIN 6698-derived	Physiologically decomposed DW-MRI radiomics	Longitudinal early-/mid-treatment MRI	0.89	Early- and mid-treatment scans are used
Current study	ACRIN 6698	ADC + DCE-MRI + HR/HER2 subtype	Pretreatment	0.86	Strictly baseline prediction

Table 2: Comparison with MRI-based studies using post-baseline or longitudinal information for pCR prediction.

Study	Dataset	Input data	AUC	Validation	Main comment
[ 14 ]	Multicenter non-ACRIN cohort	DCE + DWI + T2 radiomics + clinical variables	0.86	Multicenter	Higher input complexity and handcrafted radiomics
[ 8 ]	Single-center non-ACRIN cohort (536 patients)	T1-weighted subtraction MRI + T2-weighted MRI + clinical variables	0.888	Single-center	Pretreatment deep learning; large single-center cohort; no multicenter validation
[ 23 ]	Small non-ACRIN cohort	DCE-MRI CRBM-radiomics	0.87	None (preprint)	Small cohort; no multicenter validation
Current study	ACRIN 6698 (136 patients)	ADC + DCE-MRI + HR/HER2 subtype	0.86	Multicenter CV	Compact model; multicenter public dataset

Table 3: Comparison with MRI-based studies using strictly pretreatment information for pCR prediction.

Cancer Drug Response Prediction

See all 21 papers

Early Prediction of Pathological Complete Response to Neoadjuvant Chemotherapy Using Temporal Deep Learning on DWI
Aug 18, 2026Pablo García Marcos, Md. Tarequl Islam, Paula Puerta González +5Clinical Outcome Prediction Breast Cancer
Computer Sciences Department and the Biomedical Engineering Center (BME), University of Oviedo, 33003 Asturias, Spain · Rabindra University, Sirajganj-6770, Bangladesh · Electrical Engineering Department and the Biomedical Engineering Center (BME), University of Oviedo, 33003 Asturias, Spain +4
Longitudinal 3D Foundation Modeling for Neoadjuvant Breast Cancer Response Prediction from Serial DCE-MRI
Aug 7, 2026Fidel Omar Tito Cruz, Neda Ghafouri, Zengyan Wang +3Breast Cancer Longitudinal Medical Image Analysis
University of Central Florida, Orlando, FL, USA
Transcriptome-informed multi-modal AI for predicting neoadjuvant therapy response from breast cancer biopsies
Oct 2, 2026Jungkyu Park, Dhruva Biswas, Joseph Cappadona +22Clinical Outcome Prediction Computational Pathology
Ataraxis AI, New York, NY, USA. · Breast Oncology Program, Dana-Farber Cancer Institute, Boston, MA, USA. · Harvard Medical School, Boston, MA, USA. +13

Aug 18, 2026J/K move · Enter open · S save

Early Prediction of Pathological Complete Response to Neoadjuvant Chemotherapy Using Temporal Deep Learning on DWI

Pablo García Marcos, Md. Tarequl Islam, Paula Puerta González, Guillermo Lorenzo+4

Computer Sciences Department and the Biomedical Engineering Center (BME), University of Oviedo, 33003 Asturias, Spain · Rabindra University, Sirajganj-6770, Bangladesh · Electrical Engineering Department and the Biomedical Engineering Center (BME), University of Oviedo, 33003 Asturias, Spain+4

Clinical Outcome Prediction Breast Cancer

Open paper AskPDF arXiv

Early identification of non-responders to neoadjuvant chemotherapy (NACT) is crucial for timely treatment adaptation in breast cancer. However, many existing predictive models rely on multiparametric magnetic resonance imaging (MRI), late treatment time points, or extensive clinical data, which limits their applicability. This study proposes a deep learning framework for early prediction of pathological complete response (pCR) using only diffusion-weighted MRI (DW-MRI) acquired at baseline and after the first NACT cycle. This framework feeds cropped tumor-centered patches to an EfficientNet-based temporal model that directly learns tumor shape and local tissue characteristics without explicit radiomic feature engineering. The model, trained with 10-fold cross-validation, achieved an area under the receiver operating characteristic curve (AUC) of 0.90 for pCR prediction after one cycle, providing actionable information after a single treatment cycle while avoiding gadolinium administration and reducing dependence on heterogeneous clinical data. By focusing on the baseline-to-first-cycle window instead of later stages, the approach supports earlier escalation or de-escalation of NACT, and its exclusive reliance on DW-MRI facilitates protocol standardization, multi-centre deployment and privacy-preserving data sharing. These results demonstrate that DW-MRI-based deep learning on tumor-centered patches constitutes a minimally invasive, clinically deployable strategy for early pCR prediction, with direct implications for personalized treatment adaptation in neoadjuvant breast cancer therapy.

Abstract

Cancer Drug Response Prediction

Explore similar work

Early Prediction of Pathological Complete Response to Neoadjuvant Chemotherapy Using Temporal Deep Learning on DWI

Longitudinal 3D Foundation Modeling for Neoadjuvant Breast Cancer Response Prediction from Serial DCE-MRI

Transcriptome-informed multi-modal AI for predicting neoadjuvant therapy response from breast cancer biopsies