Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automated Prediction of Breast Cancer Response to Neoadjuvant Chemotherapy from DWI Data

Shir Nitzan, Maya Gilad, Moti Freiman

TL;DR

This study tackles automated prediction of pathological complete response (pCR) to neoadjuvant chemotherapy in breast cancer using diffusion-weighted MRI (DWI). It introduces Size-Adaptive Lesion Weighting within nnU-Net to achieve fully automatic tumor segmentation, which feeds a PD-DWI pCR predictor to estimate treatment response. At pre-NAC, the method achieves a pCR prediction performance close to expert levels (AUC around $0.76$ versus a human benchmark of $0.796$), and during NAC it outperforms standard automated approaches (AUC around $0.729$ compared with $0.654$ and $0.576$). By removing the need for manual segmentation while maintaining robustness to NAC-induced histopath changes, the approach has strong potential to improve clinical planning for NAC in breast cancer.

Abstract

Effective surgical planning for breast cancer hinges on accurately predicting pathological complete response (pCR) to neoadjuvant chemotherapy (NAC). Diffusion-weighted MRI (DWI) and machine learning offer a non-invasive approach for early pCR assessment. However, most machine-learning models require manual tumor segmentation, a cumbersome and error-prone task. We propose a deep learning model employing "Size-Adaptive Lesion Weighting" for automatic DWI tumor segmentation to enhance pCR prediction accuracy. Despite histopathological changes during NAC complicating DWI image segmentation, our model demonstrates robust performance. Utilizing the BMMR2 challenge dataset, it matches human experts in pCR prediction pre-NAC with an area under the curve (AUC) of 0.76 vs. 0.796, and surpasses standard automated methods mid-NAC, with an AUC of 0.729 vs. 0.654 and 0.576. Our approach represents a significant advancement in automating breast cancer treatment planning, enabling more reliable pCR predictions without manual segmentation.

Automated Prediction of Breast Cancer Response to Neoadjuvant Chemotherapy from DWI Data

TL;DR

This study tackles automated prediction of pathological complete response (pCR) to neoadjuvant chemotherapy in breast cancer using diffusion-weighted MRI (DWI). It introduces Size-Adaptive Lesion Weighting within nnU-Net to achieve fully automatic tumor segmentation, which feeds a PD-DWI pCR predictor to estimate treatment response. At pre-NAC, the method achieves a pCR prediction performance close to expert levels (AUC around versus a human benchmark of ), and during NAC it outperforms standard automated approaches (AUC around compared with and ). By removing the need for manual segmentation while maintaining robustness to NAC-induced histopath changes, the approach has strong potential to improve clinical planning for NAC in breast cancer.

Abstract

Effective surgical planning for breast cancer hinges on accurately predicting pathological complete response (pCR) to neoadjuvant chemotherapy (NAC). Diffusion-weighted MRI (DWI) and machine learning offer a non-invasive approach for early pCR assessment. However, most machine-learning models require manual tumor segmentation, a cumbersome and error-prone task. We propose a deep learning model employing "Size-Adaptive Lesion Weighting" for automatic DWI tumor segmentation to enhance pCR prediction accuracy. Despite histopathological changes during NAC complicating DWI image segmentation, our model demonstrates robust performance. Utilizing the BMMR2 challenge dataset, it matches human experts in pCR prediction pre-NAC with an area under the curve (AUC) of 0.76 vs. 0.796, and surpasses standard automated methods mid-NAC, with an AUC of 0.729 vs. 0.654 and 0.576. Our approach represents a significant advancement in automating breast cancer treatment planning, enabling more reliable pCR predictions without manual segmentation.
Paper Structure (16 sections, 8 equations, 3 figures, 1 table)

This paper contains 16 sections, 8 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. METHOD
  3. Data
  4. Fully Automatic pCR Prediction
  5. Segmentation
  6. PD-DWI
  7. Experimental Methodology
  8. Training Details
  9. Performance Evaluation
  10. Results
  11. Performance at pre-NAC
  12. Performance during NAC treatment
  13. CONCLUSION
  14. Compliance with ethical standards
  15. Conflict of Interest declaration
  16. ...and 1 more sections

Figures (3)

  • Figure 1: Automated Workflow for pCR Prediction: DWI data is processed by the nnU-Net for tumor segmentation using a tailored loss function. The resulting segmentation, combined with DWI, informs the PD-DWI model to forecast pCR.
  • Figure 2: DWI images with the reference manual and models segmentations.
  • Figure 3: Metrics at mid-NAC, with T0, T1, and T2 data