A Noninvasive Tool Based on Magnetic Resonance Imaging Radiomics for the Preoperative Prediction of Pathological Complete Response to Neoadjuvant Chemotherapy in Breast Cancer
Purpose This study aimed to identify patients with pathological complete response (pCR) and make better clinical decisions by constructing a preoperative predictive model based on tumoral and peritumoral volumes of multiparametric magnetic resonance imaging (MRI) obtained before neoadjuvant chemothe...
Gespeichert in:
Veröffentlicht in: | Annals of surgical oncology 2022-11, Vol.29 (12), p.7685-7693 |
---|---|
Hauptverfasser: | , , , , , , , , , , , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | Purpose
This study aimed to identify patients with pathological complete response (pCR) and make better clinical decisions by constructing a preoperative predictive model based on tumoral and peritumoral volumes of multiparametric magnetic resonance imaging (MRI) obtained before neoadjuvant chemotherapy (NAC).
Methods
This study investigated MRI before NAC in 448 patients with nonmetastatic invasive ductal breast cancer (Sun Yat-sen Memorial Hospital, Sun Yat-sen University,
n
= 362, training cohort; and Sun Yat-sen University Cancer Center,
n
= 86, validation cohort). The tumoral and peritumoral volumes of interest (VOIs) were segmented and MRI features were extracted. The radiomic features were filtered via a random forest algorithm, and a supporting vector machine was used for modeling. The receiver operator characteristic curve and area under the curve (AUC) were calculated to assess the performance of the radiomics-based classifiers.
Results
For each MRI sequence, a total of 863 radiomic features were extracted and the top 30 features were selected for model construction. The radiomic classifiers of tumoral VOI and peritumoral VOI were both promising for predicting pCR, with AUCs of 0.96 and 0.97 in the training cohort and 0.89 and 0.78 in the validation cohort, respectively. The tumoral + peritumoral VOI radiomic model could further improve the predictive accuracy, with AUCs of 0.98 and 0.92 in the training and validation cohorts.
Conclusions
The tumoral and peritumoral multiparametric MRI radiomics model can promisingly predict pCR in breast cancer using MRI images before surgery. Our results highlighted the potential value of the tumoral and peritumoral radiomic model in cancer management. |
---|---|
ISSN: | 1068-9265 1534-4681 |
DOI: | 10.1245/s10434-022-12034-w |