Histopathological Image Analysis for Cancer Detection Using Inception-V3 and Gradient Boosting

The current paper presents an improved system for classifying cancer in histopathological images utilizing the integration of Inception-v3 and Gradient Boosting. The main goal is to increase the reliability of the computer assisted cancer diagnosis system and to be a helpful tool for pathologists in practice. Inception-v3 Densenet is used for feature extraction because they are deep convolutional neural networks able to learn patterns and textures of histopathological images. These features are then taken to a Gradient Boosting classifier, which resolves the problem of overfitting and thus optimizes the classifier. On a histopathological image dataset the proposed model is tested and its accuracy, precision, recall and F1-score are found to be at 96.5%, 97.3%, 95.6% and 96.4%, respectively. The model provides a significantly better result in diagnostic performance comparing to more standard deep learning architectures, such as ResNet-50 and VGG-16. Moreover, the computational efficiency indicated in the latter approach will enable quick inference in clinical applications, making this approach highly beneficial for real-time application. The work also includes techniques in the explainability area, such as Grad-CAM and Layer-wise Relevance Propagation (LRP), which contribute to the model's decision-making. The graph made is useful for illustrating the ideas mentioned above The proposed hybrid model is reliable, efficient and can easily be implemented in order to enhance cancer detection hence making the diagnostic process accurate and more efficient to support timely decisions in the health care profession.