DI-UMONS : Dépôt institutionnel de l’université de Mons

Recherche transversale
Rechercher
(titres de publication, de périodique et noms de colloque inclus)
2020-01-24 - Article/Dans un journal avec peer-review - Anglais - page(s)

Braman Nathaniel, El Adoui Mohammed , Drisis Stylianos, Vulchi Manasa, Benjelloun Mohammed , Madabhushi Anant, (en collaboration avec 6 autres personnes ) , "Deep learning-based prediction of response to HER2-targeted neoadjuvant chemotherapy from pre-treatment dynamic breast MRI: A multi-institutional validation study" in Nature Communications, 2020, 2020, 2001-2034

  • Edition : Nature Publishing Group (United Kingdom)
  • Codes CREF : Techniques d'imagerie et traitement d'images (DI2770), Intelligence artificielle (DI1180), Théorie de l'information (DI1161), Informatique mathématique (DI1160)
  • Unités de recherche UMONS : Informatique (F114)
  • Instituts UMONS : Institut de Recherche en Technologies de l’Information et Sciences de l’Informatique (InforTech)
Texte intégral :

Abstract(s) :

(Anglais) Predicting response to neoadjuvant therapy is a vexing challenge in breast cancer. In this study, we evaluate the ability of deep learning to predict response to HER2-targeted neo-adjuvant chemotherapy (NAC) from pre-treatment dynamic contrast-enhanced (DCE) MRI acquired prior to treatment. In a retrospective study encompassing DCE-MRI data from a total of 157 HER2+ breast cancer patients from 5 institutions, we developed and validated a deep learning approach for predicting pathological complete response (pCR) to HER2-targeted NAC prior to treatment. 100 patients who received HER2-targeted neoadjuvant chemotherapy at a single institution were used to train (n=85) and tune (n=15) a convolutional neural network (CNN) to predict pCR. A multi-input CNN leveraging both pre-contrast and late post-contrast DCE-MRI acquisitions was identified to achieve optimal response prediction within the validation set (AUC=0.93). This model was then tested on two independent testing cohorts with pre-treatment DCE-MRI data: 28 patients who received HER2-targeted NAC at a second institution and a 29-patient clinical trial dataset with imaging data from 3 institutions. This model achieved strong performance in a 28 patient testing set from a second institution (AUC=0.85, 95% CI 0.67-1.0, p=.0008) and a 29 patient multicenter trial including data from 3 additional institutions (AUC=0.77, 95% CI 0.58-0.97, p=0.006). Deep learning-based response prediction model was further found to exceed both a multivariable model incorporating predictive clinical variables (AUC < .65 in testing cohorts) and a model utilizing semi-quantitative DCE-MRI pharmacokinetic measurements (AUC < .60 in testing cohorts) in performance and robustness. The results presented in this work across multiple sites suggest that with further validation deep learning could provide an effective and reliable tool to guide targeted therapy in breast cancer, thus reducing overtreatment among HER2+ patients.