Cross-attention fusion of biparametric MRI sequences with demographic and clinical variables for patient-level classification of clinically significant prostate cancer
Abstract
Accurate patient-level risk stratification of clinically significant prostate cancer (csPCa) on biparametric MRI remains challenging, as most existing approaches require costly lesion-level annotations or have not been evaluated in large temporally separated cohorts. We developed and validated a multimodal deep learning framework combining T2-weighted (T2w), diffusion-weighted (DWI), and apparent diffusion coefficient (ADC) volumes with age, prostate-specific antigen (PSA), and PSA density (PSAD) for patient-level csPCa classification, trained without lesion-level annotations. The study included a retrospective development cohort of 3, 939 examinations and an independent temporal validation cohort comprising 1, 409 prospectively acquired examinations from 13 centers. The model employs three sequence-specific encoders with asymmetric cross-attention fusion and a two-stage transfer-learning strategy. Adding demographic and clinical variables consistently improved performance over MRI-only models, with PSAD emerging as the most informative complementary factor, primarily by improving specificity. The best-performing model—combining MRI, age, PSA, and PSAD with cross-attention—achieved a mean AUC of 0.765 ± 0.006 on the retrospective held-out test set and 0.741 ± 0.006 on the independent temporal validation cohort. Post hoc subgroup analyses suggested broadly stable rank-order discrimination (AUC) across most strata, though fixed-threshold sensitivity and specificity varied substantially, while Grad-CAM maps offered a qualitative indication of more frequent overlap with suspicious regions under cross-attention. These findings support PSAD-informed multimodal patient-level models as a viable tool for refined csPCa risk stratification on biparametric MRI, with potential to reduce false-positive classification of ISUP grade group 1 disease as clinically significant.
Related
- Confidence-based classification of clinically significant prostate cancer on biparametric MRI using Monte Carlo dropout and Grad-CAM visualization.
- Characterization and Antimicrobial Susceptibility Patterns of Clinically Significant Non-fermenting Gram-Negative Bacilli Isolated From Clinical Specimens
- Transcatheter atrial septal defect closure in infants weighing 3 to 6 kg with clinically significant bronchopulmonary dysplasia
- Validation of the discriminatory performance of the iPREFACE score for clinically significant fetal acidemia: an outcome-enriched nested case–control study
- Clinically significant neck pain and tablet tilt angles in real-world educational settings: A nationwide cross-sectional study of university students
BibTeX
@article{dimitriadis2026crossattention,
title = {{Cross-attention fusion of biparametric MRI sequences with demographic and clinical variables for patient-level classification of clinically significant prostate cancer}},
author = {Avtantil Dimitriadis and Grigorios Kalliatakis and Richard Osuala and Dimitri Kessler and Oliver Diaz and Simone Mazzetti and Daniele Regge and Karim Lekadir and Dimitrios Fotiadis and Manolis Tsiknakis and Nikolaos Papanikolaou and the ProCAncer-I Consortium and Stelios Sfakianakis and Varvara Kalokyri and Eleftherios Trivizakis and Nikolaos Tachos and Eugenia Mylona and Dimitrios Zaridis and Charalampos Kalantzopoulos and José Guilherme de Almeida and Ana Castro Verde and Ana Carolina Rodrigues and Nuno Rodrigues and Miguel Chambel and Henkjan Huisman and Maarten de Rooij and Anindo Saha and Jasper J. Twilt and Jurgen Futterer and Luis Martí-Bonmatí and Leonor Cerdá-Alberich and Gloria Ribas and Silvia Navarro and Manuel Marfil and Emanuele Neri and Giacomo Aringhieri and Lorenzo Tumminello and Vincenzo Mendola and Deniz Akata and Mustafa Özmen and Ali Devrim Karaosmanoglu and Firat Atak and Musturay Karcaaltincaba and Joan C. Vilanova and Jurgita Usinskiene and Ruta Briediene and Audrius Untanas and Kristina Slidevska and Katsaros Vasilis and Georgiou Georgios and Dow-Mu Koh and Robby Emsley and Sharon Vit and Ana Ribeiro and Simon Doran and Tiaan Jacobs and Gracián García-Martí and Valentina Giannini and Giovanni Cappello and Giovanni Maimone},
journal = {Scientific Reports},
year = {2026},
month = oct,
doi = {10.1038/s41598-026-71998-x},
url = {https://www.nature.com/articles/s41598-026-71998-x},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Cross-attention fusion of biparametric MRI sequences with demographic and clinical variables for patient-level classification of clinically significant prostate cancer
AU - Avtantil Dimitriadis
AU - Grigorios Kalliatakis
AU - Richard Osuala
AU - Dimitri Kessler
AU - Oliver Diaz
AU - Simone Mazzetti
AU - Daniele Regge
AU - Karim Lekadir
AU - Dimitrios Fotiadis
AU - Manolis Tsiknakis
AU - Nikolaos Papanikolaou
AU - the ProCAncer-I Consortium
AU - Stelios Sfakianakis
AU - Varvara Kalokyri
AU - Eleftherios Trivizakis
AU - Nikolaos Tachos
AU - Eugenia Mylona
AU - Dimitrios Zaridis
AU - Charalampos Kalantzopoulos
AU - José Guilherme de Almeida
AU - Ana Castro Verde
AU - Ana Carolina Rodrigues
AU - Nuno Rodrigues
AU - Miguel Chambel
AU - Henkjan Huisman
AU - Maarten de Rooij
AU - Anindo Saha
AU - Jasper J. Twilt
AU - Jurgen Futterer
AU - Luis Martí-Bonmatí
AU - Leonor Cerdá-Alberich
AU - Gloria Ribas
AU - Silvia Navarro
AU - Manuel Marfil
AU - Emanuele Neri
AU - Giacomo Aringhieri
AU - Lorenzo Tumminello
AU - Vincenzo Mendola
AU - Deniz Akata
AU - Mustafa Özmen
AU - Ali Devrim Karaosmanoglu
AU - Firat Atak
AU - Musturay Karcaaltincaba
AU - Joan C. Vilanova
AU - Jurgita Usinskiene
AU - Ruta Briediene
AU - Audrius Untanas
AU - Kristina Slidevska
AU - Katsaros Vasilis
AU - Georgiou Georgios
AU - Dow-Mu Koh
AU - Robby Emsley
AU - Sharon Vit
AU - Ana Ribeiro
AU - Simon Doran
AU - Tiaan Jacobs
AU - Gracián García-Martí
AU - Valentina Giannini
AU - Giovanni Cappello
AU - Giovanni Maimone
PY - 2026
DA - 2026/10/08
JO - Scientific Reports
DO - 10.1038/s41598-026-71998-x
UR - https://www.nature.com/articles/s41598-026-71998-x
AB - Accurate patient-level risk stratification of clinically significant prostate cancer (csPCa) on biparametric MRI remains challenging, as most existing approaches require costly lesion-level annotations or have not been evaluated in large temporally separated cohorts. We developed and validated a multimodal deep learning framework combining T2-weighted (T2w), diffusion-weighted (DWI), and apparent diffusion coefficient (ADC) volumes with age, prostate-specific antigen (PSA), and PSA density (PSAD) for patient-level csPCa classification, trained without lesion-level annotations. The study included a retrospective development cohort of 3, 939 examinations and an independent temporal validation cohort comprising 1, 409 prospectively acquired examinations from 13 centers. The model employs three sequence-specific encoders with asymmetric cross-attention fusion and a two-stage transfer-learning strategy. Adding demographic and clinical variables consistently improved performance over MRI-only models, with PSAD emerging as the most informative complementary factor, primarily by improving specificity. The best-performing model—combining MRI, age, PSA, and PSAD with cross-attention—achieved a mean AUC of 0.765 ± 0.006 on the retrospective held-out test set and 0.741 ± 0.006 on the independent temporal validation cohort. Post hoc subgroup analyses suggested broadly stable rank-order discrimination (AUC) across most strata, though fixed-threshold sensitivity and specificity varied substantially, while Grad-CAM maps offered a qualitative indication of more frequent overlap with suspicious regions under cross-attention. These findings support PSAD-informed multimodal patient-level models as a viable tool for refined csPCa risk stratification on biparametric MRI, with potential to reduce false-positive classification of ISUP grade group 1 disease as clinically significant.
SN - 2045-2322
ER -