45 Lung Cancer Tissues Computer Vision News This automation translates to faster and more accurate diagnoses, a crucial factor in improving patient outcomes. My research project at Ummon Healthtech centers around EGFR gene mutations in LUAD, one of the NSCLC subtypes. The primary goal is to unearth the underlying patterns associated with EGFR mutations, enhancing the interpretability of these complex processes and, in turn, revolutionizing patient care. To achieve this, the research follows a systematic methodology that combines the strengths of deep learning and machine learning: Generation of Prediction Scores: This step involves the extraction of patches from Whole Slide Images (WSI). These patches are then processed using an EfficientNetB7 neural network, which generates embedding vectors. The resulting prediction scores are crucial for the subsequent analysis. Machine Learning Prediction Analysis: Handcrafted features are extracted from the WSI images. These features undergo random forest feature selection, and various machine learning techniques, including classification and regression, are applied to predict labels. The results of this study are promising: in regression analysis, the Random Forest model achieved notable R-squared scores, indicating its efficacy in understanding EGFR mutations. In terms of classification, the Random Forest model outperformed other models, particularly when utilizing a subset of informative features such as GLCM, LBP, pixel intensity, and color moments. As we conclude this journey, it's important to note that this research is just the beginning. Future work in this domain may explore advanced clinical and shape features and leverage a combination of deep learning and machine learning techniques to build even more robust models for lung cancer diagnosis.
RkJQdWJsaXNoZXIy NTc3NzU=