Background: Neoadjuvant therapy is crucial for breast cancer patients seeking to retain the breast by reducing the tumor size for conserving surgery. This treatment aims to achieve a pathologic complete response (pCR), eradicating cancer cells entirely and reducing the risk of recurrence. The study’s objective is to devise a predictive approach for identifying patients achieving pCR through neoadjuvant therapy, based on radiomic features extracted from MR images by leveraging the InceptionV3 model with advanced validation techniques.
Methods: In our study, we gathered data from 364 unique Patient IDs sourced from the-SPY 2 MRI database with the goal of classifying pCR (pathological complete response). Our research introduced three key areas of novelty.Firstly, we explored the extraction of advanced features such as region centroid, Entropy, Sphericity, and more. These features provided deeper insights into the characteristics of the MRI data and enhanced the discriminative power of our classification model.Secondly, we applied these extracted features to the InceptionV3 (GoogleNet) model. To optimize the model’s performance, we experimented with different combinations of loss functions, optimizer functions, and activation functions. This thorough exploration allowed us to identify the most effective configuration for the given task.Lastly, our classification results were subjected 1 to validation using advanced techniques such as Matthews Correlation Coefficient, Cohen’s Kappa, and Jaccard Index. These evaluation metrics provided a robust assessment of the model’s performance and ensured the reliability of our findings.
Results: The successful combination of advanced feature extraction, utilization of the InceptionV3 model with tailored hyperparameters, and thorough validation using cutting-edge techniques significantly enhanced the accuracy and reliability of our pCR classification study. By adopting a collabo-rative approach that involved both radiologists and the computer-aided system, we achieved superior predictive performance for pCR, as evi-denced by the impressive values obtained for the area under the curve (AUC) at 0.98.
Conclusion: Overall, the combination of advanced feature extraction, lever-aging the InceptionV3 model with customized hyperparameters, and rigorous validation using state-of-the-art techniques contributed to the accuracy and credibility of our pCR classification study.