Breast Cancer Classification through Meta-Learning Ensemble Technique Using Convolution Neural Networks

Deep learning radiomics based on multimodal imaging for distinguishing benign and malignant breast tumours

Objectives

This study aimed to develop a deep learning radiomic model using multimodal imaging to differentiate benign and malignant breast tumours.

Methods

Multimodality imaging data, including ultrasonography (US), mammography (MG), and magnetic resonance imaging (MRI), from 322 patients (112 with benign breast tumours and 210 with malignant breast tumours) with histopathologically confirmed breast tumours were retrospectively collected between December 2018 and May 2023. Based on multimodal imaging, the experiment was divided into three parts: traditional radiomics, deep learning radiomics, and feature fusion. We tested the performance of seven classifiers, namely, SVM, KNN, random forest, extra trees, XGBoost, LightGBM, and LR, on different feature models. Through feature fusion using ensemble and stacking strategies, we obtained the optimal classification model for benign and malignant breast tumours.

Results

In terms of traditional radiomics, the ensemble fusion strategy achieved the highest accuracy, AUC, and specificity, with values of 0.892, 0.942 [0.886-0.996], and 0.956 [0.873-1.000], respectively. The early fusion strategy with US, MG, and MRI achieved the highest sensitivity of 0.952 [0.887-1.000]. In terms of deep learning radiomics, the stacking fusion strategy achieved the highest accuracy, AUC, and sensitivity, with values of 0.937, 0.947 [0.887-1.000], and 1.000 [0.999-1.000], respectively. The early fusion strategies of US+MRI and US+MG achieved the highest specificity of 0.954 [0.867-1.000]. In terms of feature fusion, the ensemble and stacking approaches of the late fusion strategy achieved the highest accuracy of 0.968. In addition, stacking achieved the highest AUC and specificity, which were 0.997 [0.990-1.000] and 1.000 [0.999-1.000], respectively. The traditional radiomic and depth features of US+MG + MR achieved the highest sensitivity of 1.000 [0.999-1.000] under the early fusion strategy.

Conclusion

This study demonstrated the potential of integrating deep learning and radiomic features with multimodal images. As a single modality, MRI based on radiomic features achieved greater accuracy than US or MG. The US and MG models achieved higher accuracy with transfer learning than the single-mode or radiomic models. The traditional radiomic and depth features of US+MG + MR achieved the highest sensitivity under the early fusion strategy, showed higher diagnostic performance, and provided more valuable information for differentiation between benign and malignant breast tumours.

Fine tuning deep learning models for breast tumor classification

This paper proposes an approach to enhance the differentiation task between benign and malignant Breast Tumors (BT) using histopathology images from the BreakHis dataset. The main stages involve preprocessing, which encompasses image resizing, data partitioning (training and testing sets), followed by data augmentation techniques. Both feature extraction and classification tasks are employed by a Custom CNN. The experimental results show that the proposed approach using the Custom CNN model exhibits better performance with an accuracy of 84% than applying the same approach using other pretrained models, including MobileNetV3, EfficientNetB0, Vgg16, and ResNet50V2, that present relatively lower accuracies, ranging from 74 to 82%; these four models are used as both feature extractors and classifiers. To increase the accuracy and other performance metrics, Grey Wolf Optimization (GWO), and Modified Gorilla Troops Optimization (MGTO) metaheuristic optimizers are applied to each model separately for hyperparameter tuning. In this case, the experimental results show that the Custom CNN model, refined with MGTO optimization, reaches an exceptional accuracy of 93.13% in just 10 iterations, outperforming the other state-of-the-art methods, and the other four used pretrained models based on the BreakHis dataset.

Nested genetic algorithm-based classifier selection and placement in multi-level ensemble framework for effective disease diagnosis

Effective disease diagnosis is a critical unmet need on a global scale. The intricacies of the numerous disease mechanisms and underlying symptoms make developing a model for early diagnosis and effective treatment extremely difficult. Machine learning (ML) can help to solve some of these issues. Recently, various ensemble-based ML models have benefited clinicians in early diagnosis. However, one of the most difficult challenges in multi-level ensemble approaches is the classifier selection and their placement in the ensemble framework as it improves the overall performance. Let m classifiers have to select from n classifiers there are ( n m ) ways. Again, these ( n m ) possibilities can be arranged in m ! ways. Finding the best m classifiers and their positions from total ( n m ) m ! ways is a challenging and hard problem. To address this challenge, a dynamic three-level ensemble framework is proposed. A nested Genetic Algorithm (GA) and ensemble-based fitness function are employed to optimize the classifier selection and their placement in a three-level ensemble framework. Our approach used eleven classifiers and chose seven classifiers by maximizing the fitness function. The proposed model experiments on 12 disease datasets. The proposed model outperformed in terms of accuracy, F1, and G-measure on the Chronic Kidney Disease (CKD) dataset is 0.987, 0.988, and 0.989, respectively. In terms of AUC on the Heart disease dataset (HDD) is 0.998 and in terms of recall on the Hypothyroid disease dataset (HyDD) is 0.988. In addition, the proposed model superiority is statically evaluated by Wilcoxon-Signed-Rank (WSR) test compared with other ensemble models, such as random forest (RF), bagging classifier (BC), XGBoost (XGB), and gradient boost classifier (GBC) with probability value p < 0.05 results shows all the traditional ensemble model differs with proposed model and also effective size evaluated with using the matched-pairs rank biserial correlation coefficient wc and statistical results shows effective size is large with RF and BC and effective size is medium with XGB and GBC. Proposed model has outperformed comparing with State-Of-The-Art (SOTA) ensemble and non-ensemble models. Further, the proposed model outperformed in terms of the ROC curve in the majority of the disease datasets. The results suggest the usage of the proposed model for disease diagnosis applications.