Conv1D-LSTM: Autonomous Breast Cancer Detection Using a One-Dimensional Convolutional Neural Network With Long Short-Term Memory

Breast cancer is an increasingly serious problem in contemporary society, with millions of women and men worldwide affected by the disease. While traditional cancer detection strategies are at times effective, they typically require costly and time-intensive methods for implementation. The major drawback of using conventional methods for identifying breast cancer using the available data sets is that a single algorithm is not sufficient for accurate breast cancer diagnosis due to the heterogeneity of tumors, diverse data types, pattern complexity, feature engineering and dataset overfitting. The aim is to surpass the constraints of the conventional models and develop a hybrid model. The idea is to attain higher accuracy and lower computational time than existing models. This paper introduces a new method for detecting breast cancer using a one-dimensional convolutional neural network (1D CNN) and long short-term memory (LSTM). The model combines the strengths of both approaches, extracting sequential features from local data and modeling temporal dependencies and relationships. To detect and classify breast cancer, the 1D CNN and LSTM are used to automatically extract and analyze features from distinguishing features from a real dataset generated from mammography reports. The developed model has been assessed on on the extracted feature of the primary available dataset consisting of mammograms from over 760 patients. The developed model achieves 99% accuracy on the test data, demonstrating its potential to provide an automated approach to breast cancer detection. The work emphasizes a significant improvement in feature extraction, accuracy, and robustness. Additionally, the proposed model's versatility allows it to handle diverse data types, achieve better generalization and lower computational time. The model offers a high level of interpretability, which is crucial for medical professionals to understand and trust the decision-making process of the system. The developed hybrid model outperforms various other state-of-the-art techniques like ANN, CNN, CNN-Bi-LSTM-GRU-AM (Convolutional Neural Network-Bidirectional Long Short-Term Memory-Gated Recurrent Unit-Attention Mechanism), and CNN-GRU (Convolutional Neural Network- Gated Recurrent Unit) in terms of accuracy, feature extraction and computational time. This work emphasises the potential of 1D CNN augmented with LSTM to create an automated system for identifying breast cancer. Hence, provides a promising foundation for further development and practical usage of deep learning for automated cancer diagnosis.