ABSTRACT
Esophageal cancer (EC) remains a significant health challenge globally, with increasing incidence and high mortality rates. Despite advances in treatment, there remains a need for improved diagnostic methods and understanding of disease progression. This study addresses the significant challenges in the automatic classification of EC, particularly in distinguishing its primary subtypes: adenocarcinoma and squamous cell carcinoma, using histopathology images. Traditional histopathological diagnosis, while being the gold standard, is subject to subjectivity and human error and imposes a substantial burden on pathologists. This study proposes a binary class classification system for detecting EC subtypes in response to these challenges. The system leverages deep learning techniques and tissue-level labels for enhanced accuracy. We utilized 59 high-resolution histopathological images from The Cancer Genome Atlas (TCGA) Esophageal Carcinoma dataset (TCGA-ESCA). These images were preprocessed, segmented into patches, and analyzed using a pre-trained ResNet101 model for feature extraction. For classification, we employed five machine learning classifiers: Support Vector Classifier (SVC), Logistic Regression (LR), Decision Tree (DT), AdaBoost (AD), Random Forest (RF), and a Feed-Forward Neural Network (FFNN). The classifiers were evaluated based on their prediction accuracy on the test dataset, yielding results of 0.88 (SVC and LR), 0.64 (DT and AD), 0.82 (RF), and 0.94 (FFNN). Notably, the FFNN classifier achieved the highest Area Under the Curve (AUC) score of 0.92, indicating its superior performance, followed closely by SVC and LR, with a score of 0.87. This suggested approach holds promising potential as a decision-support tool for pathologists, particularly in regions with limited resources and expertise. The timely and precise detection of EC subtypes through this system can substantially enhance the likelihood of successful treatment, ultimately leading to reduced mortality rates in patients with this aggressive cancer.
ABSTRACT
Breast Cancer (BC) is the most commonly diagnosed cancer and second leading cause of mortality among women. About 1 in 8 US women (about 13%) will develop invasive BC throughout their lifetime. Early detection of this life-threatening disease not only increases the survival rate but also reduces the treatment cost. Fortunately, advancements in radiographic imaging like "Mammograms", "Computed Tomography (CT)", "Magnetic Resonance Imaging (MRI)", "3D Mammography", and "Histopathological Imaging (HI)" have made it feasible to diagnose this life-taking disease at an early stage. However, the analysis of radiographic images and Histopathological images is done by experienced radiologists and pathologists, respectively. The process is not only costly but also error-prone. Over the last ten years, Computer Vision and Machine Learning (ML) have transformed the world in every way possible. Deep learning (DL), a subfield of ML has shown outstanding results in a variety of fields, particularly in the biomedical industry, because of its ability to handle large amounts of data. DL techniques automatically extract the features by analyzing the high dimensional and correlated data efficiently. The potential and ability of DL models have also been utilized and evaluated in the identification and prognosis of BC, utilizing radiographic and Histopathological images, and have performed admirably. However, AI has shown good claims in retrospective studies only. External validations are needed for translating these cutting-edge AI tools as a clinical decision maker. The main aim of this research work is to present the critical analysis of the research and findings already done to detect and classify BC using various imaging modalities including "Mammography", "Histopathology", "Ultrasound", "PET/CT", "MRI", and "Thermography". At first, a detailed review of the past research papers using Machine Learning, Deep Learning and Deep Reinforcement Learning for BC classification and detection is carried out. We also review the publicly available datasets for the above-mentioned imaging modalities to make future research more accessible. Finally, a critical discussion section has been included to elaborate open research difficulties and prospects for future study in this emerging area, demonstrating the limitations of Deep Learning approaches.
