سامانه همایش‌ها TSTA | ثبت‌نام و ارسال مقاله کنفرانس‌ها و کنگره‌های ملی

صفحه اصلی / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

An Attention-Guided Convolutional Neural Network for Predicting Neoadjuvant Chemotherapy Response in Breast Cancer Patients

نویسندگان :

Parisa Donyaei¹ Javad Haddadnia²

1- دانشگاه حکیم سبزواری 2- دانشگاه حکیم سبزواری

کلمات کلیدی :

Breast Cancer،Pathological Complete Response،Neoadjuvant Chemotherapy (NAC)،Attention Mechanism،Convolutional Neural Network (CNN)،Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI)،Early Detection of Breast Cancer،Malignant Lesion Diagnosis

چکیده :

Breast cancer remains a leading global health concern among women. Due to factors like tumor adhesion or large volume making some patients initially ineligible for surgery, neoadjuvant chemotherapy (NAC) is commonly used to reduce tumor burden and metastasis before surgical intervention. However, accurately forecasting an individual’s response to NAC remains a major clinical challenge. It critically impacts personalized treatment planning, patient prognosis, and survival rates. It is also essential for minimizing unnecessary treatment-related toxicity. In this study, we propose an Attention-Guided Convolutional Neural Network (AG-CNN) that utilizes Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) data to predict pathological complete response (PCR). Our dual-branch architecture combines global and local feature extraction with attention mechanisms that enhance focus on diagnostically relevant regions, thereby improving predictive accuracy. The model achieved a prediction accuracy of 90.76%, a sensitivity of 96.66%, and a specificity of 89.95%. The model outperformed baseline methods lacking attention mechanisms, demonstrating the significant advantage of integrating attention in clinical outcome prediction. These findings underscore the significant clinical utility of attention-guided learning frameworks in enhancing the precision of treatment response predictions, ultimately supporting more informed, personalized therapeutic decisions and improving patient outcomes in breast cancer care.