Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Ensemble of radiomics and ConvNeXt for breast cancer diagnosis

Jorge Alberto Garza-Abdala, Gerardo Alejandro Fumagal-González, Beatriz A. Bosques-Palomo, Mario Alexis Monsivais Molina, Daly Avedano, Servando Cardona-Huerta, José Gerardo Tamez-Pena

TL;DR

Breast cancer screening via mammography faces sensitivity limitations, particularly in dense breasts. The paper proposes a radiomics–ConvNeXtV1-small ensemble to improve diagnostic accuracy, evaluated on two large datasets (RSNA 2023 and TecSalud). The ensemble outperforms radiomics and DL alone, achieving an AUC of approximately 0.878 and demonstrating robust cross-dataset calibration via a leave-one-year-out strategy. This approach offers a practical, computation-friendly path toward more accurate, generalizable mammography-based breast cancer detection with potential clinical impact.

Abstract

Early diagnosis of breast cancer is crucial for improving survival rates. Radiomics and deep learning (DL) have shown significant potential in assisting radiologists with early cancer detection. This paper aims to critically assess the performance of radiomics, DL, and ensemble techniques in detecting cancer from screening mammograms. Two independent datasets were used: the RSNA 2023 Breast Cancer Detection Challenge (11,913 patients) and a Mexican cohort from the TecSalud dataset (19,400 patients). The ConvNeXtV1-small DL model was trained on the RSNA dataset and validated on the TecSalud dataset, while radiomics models were developed using the TecSalud dataset and validated with a leave-one-year-out approach. The ensemble method consistently combined and calibrated predictions using the same methodology. Results showed that the ensemble approach achieved the highest area under the curve (AUC) of 0.87, compared to 0.83 for ConvNeXtV1-small and 0.80 for radiomics. In conclusion, ensemble methods combining DL and radiomics predictions significantly enhance breast cancer diagnosis from mammograms.

Ensemble of radiomics and ConvNeXt for breast cancer diagnosis

TL;DR

Breast cancer screening via mammography faces sensitivity limitations, particularly in dense breasts. The paper proposes a radiomics–ConvNeXtV1-small ensemble to improve diagnostic accuracy, evaluated on two large datasets (RSNA 2023 and TecSalud). The ensemble outperforms radiomics and DL alone, achieving an AUC of approximately 0.878 and demonstrating robust cross-dataset calibration via a leave-one-year-out strategy. This approach offers a practical, computation-friendly path toward more accurate, generalizable mammography-based breast cancer detection with potential clinical impact.

Abstract

Early diagnosis of breast cancer is crucial for improving survival rates. Radiomics and deep learning (DL) have shown significant potential in assisting radiologists with early cancer detection. This paper aims to critically assess the performance of radiomics, DL, and ensemble techniques in detecting cancer from screening mammograms. Two independent datasets were used: the RSNA 2023 Breast Cancer Detection Challenge (11,913 patients) and a Mexican cohort from the TecSalud dataset (19,400 patients). The ConvNeXtV1-small DL model was trained on the RSNA dataset and validated on the TecSalud dataset, while radiomics models were developed using the TecSalud dataset and validated with a leave-one-year-out approach. The ensemble method consistently combined and calibrated predictions using the same methodology. Results showed that the ensemble approach achieved the highest area under the curve (AUC) of 0.87, compared to 0.83 for ConvNeXtV1-small and 0.80 for radiomics. In conclusion, ensemble methods combining DL and radiomics predictions significantly enhance breast cancer diagnosis from mammograms.
Paper Structure (11 sections, 3 figures, 2 tables)

This paper contains 11 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Dataset description
  3. Methodology
  4. Radiomics Image Preprocessing
  5. Radiomics Features And Machine Learning
  6. Deep Learning Workflow
  7. Ensemble workflow (Radiomics + ConvNeXtV1-small model)
  8. Results
  9. Discussion
  10. Limitations and future work
  11. Conclusion

Figures (3)

  • Figure 1: Sample images from TecSalud dataset before and after applying the preprocessing steps (including histogram matching).
  • Figure 2: Workflow of the proposed ensemble. The orange squares represent the radiomics subensemble while the blue squares are for the ConvNeXtV1-small.
  • Figure 3: AUC ROC for the ensemble. The dotted red line represents the radiomics sub ensemble, while the dotted blue line is the ConNeXt.