Early Prediction of Breast Cancer Recurrence for Patients Treated with Neoadjuvant Chemotherapy: A Transfer Learning Approach on DCE-MRIs

Abstract

Cancer treatment planning benefits from an accurate early prediction of the treatment efficacy. The goal of this study is to give an early prediction of three-year Breast Cancer Recurrence (BCR) for patients who underwent neoadjuvant chemotherapy. We addressed the task from a new perspective based on transfer learning applied to pre-treatment and early-treatment DCE-MRI scans. Firstly, low-level features were automatically extracted from MR images using a pre-trained Convolutional Neural Network (CNN) architecture without human intervention. Subsequently, the prediction model was built with an optimal subset of CNN features and evaluated on two sets of patients from I-SPY1 TRIAL and BREAST-MRI-NACT-Pilot public databases: a fine-tuning dataset (70 not recurrent and 26 recurrent cases), which was primarily used to find the optimal subset of CNN features, and an independent test (45 not recurrent and 17 recurrent cases), whose patients had not been involved in the feature selection process. The best results were achieved when the optimal CNN features were augmented by four clinical variables (age, ER, PgR, HER2+), reaching an accuracy of 91.7% and 85.2%, a sensitivity of 80.8% and 84.6%, a specificity of 95.7% and 85.4%, and an AUC value of 0.93 and 0.83 on the fine-tuning dataset and the independent test, respectively. Finally, the CNN features extracted from pre-treatment and early-treatment exams were revealed to be strong predictors of BCR.

A Multicentre Evaluation of Dosiomics Features Reproducibility, Stability and Sensitivity

“…The key idea behind radiomics is that we can mine images by extracting image descriptors, called radiomic features, which can provide rich information about the tumour or healthy tissue and can be used to build predictive or prognostic models. This method allows quantitative analysis of different Image modalities and identification of patterns and correlations among voxels that can be of interest for improving diagnosis, prognosis and prediction of treatment outcomes [2][3][4]. Clinical outcomes can be therefore predicted employing radiomics features, potentially changing the treatment paradigm.…”

Sensitivity of Contrast-Enhanced Breast MRI vs X-ray Mammography Based on Cancer Histology, Tumor Grading, Receptor Status, and Molecular Subtype: A Supplemental Analysis of 2 Large Phase III Studies

“…Sung Eun Song et al published a retrospective comparative study on CE-BMRI and XRM imaging features of HER2+ breast cancers according to hormone receptor status. While survival, pattern of recurrence, and treatment (neo-adjuvant) response differ between HER2+/HR+ vs HER+/HR– remarkably (and are hard to predict 43 , 44 ), they did not find any differences in mammographic imaging presentations and calcification features and magnetic resonance (MR) kinetic features by a computer-aided diagnosis (CAD). 45 However, no direct comparison of sensitivities of the 2 imaging tools was reported.…”

Robustness Evaluation of a Deep Learning Model on Sagittal and Axial Breast DCE-MRIs to Predict Pathological Complete Response to Neoadjuvant Chemotherapy

“…This aspect can be crucial to prove the generalizability of the method on data provided by multiple cancer institutions. As demonstrated elsewhere [ 26 , 41 ], the addition of clinical variables could contribute to improving the performances achieved by using an AI approach. By combining the CNN-features with the clinical variables, the overall performances were stable at varying MRI acquisition views: an AUC value of 80.3% and 78.0% was achieved on the independent tests conducted for the public DB and the private DB, respectively.…”