Early Diagnosis of Endometrial Cancer: An Ensemble-Based Deep Learning Approach
Bhawna Swarnkar, Parag Maheshwari, Nilay Khare, Manasi Gyanchandani · 2024
Early diagnosis and treatment are key to improving the prognosis of endometrial cancer. However, conventional machine learning approaches have limited capacity to simulate the complex links between histopathological images and their interpretations, making it challenging to achieve accurate results. The objective of this paper is to develop a machine learning algorithm for the accurate diagnosis of endometrial cancer using histopathology images. A machine learning-based ensemble method was utilized to classify histopathology images into four categories). Three pre-trained convolutional neural network architectures, namely VGG16, DenseNet121, and MobileNetV2, were employed for feature extraction. The last convolutional layer of each CNN is used for high level feature extraction from the input and single feature vector is the output of these concatenated layers. The concatenated feature vector was then fed into a softmax layer for classification. The dataset consisted of 3302 histopathology images classified into four categories. The proposed algorithm achieved an accuracy of 97.17%, F-score of 96.79%, sensitivity of 97.17%, and precision of 97.25%. These results demonstrate its potential for facilitating early detection and diagnosis of endometrial cancer. The methodology of concatenating multiple pre-trained CNNs and utilizing their extracted features provides a robust and accurate approach for endometrial cancer classification