In a significant stride towards enhancing breast cancer detection, a team of researchers led by Omneya Attallah from the Department of Electronics and Communications Engineering at the Arab Academy for Science, Technology and Maritime Transport in Alexandria, Egypt, has developed a novel computer-aided diagnosis (CAD) framework. This system leverages thermal imaging and deep learning to improve the accuracy and efficiency of breast cancer detection. The study, published in the journal ‘Applied Sciences’ (translated from German as ‘Applied Sciences’), introduces a method that could revolutionize non-invasive diagnostic techniques, with potential implications for maritime sectors involved in medical technology and remote healthcare.
Breast cancer remains the most common malignancy among women worldwide, and while mammography is the current gold standard for screening, it has its limitations. These include radiation exposure and the need for skilled personnel to interpret results. Thermal imaging, or thermography, offers a non-invasive, radiation-free alternative. However, existing CAD systems often rely on complex convolutional neural network (CNN) structures and resource-intensive preprocessing, which can be cumbersome and less efficient.
Attallah and her team addressed these challenges by incorporating streamlined CNN designs, transfer learning strategies, and multi-architecture ensemble methods. “We aimed to mitigate the limitations of current CAD systems by using simpler CNN designs and more efficient feature extraction techniques,” Attallah explained. The researchers utilized MobileNet, EfficientNetB0, and ShuffleNet architectures to obtain features from various layers, assessing the impact of individual layers on classification performance. This multi-layer feature amalgamation approach significantly improved the system’s accuracy.
The study found that integrating features from different CNNs and layers markedly enhanced performance, achieving a maximum accuracy of 99.4%. Moreover, combining attributes from all three layers of the CNNs with non-negative matrix factorization (NNMF) achieved an impressive 99.9% accuracy with just 350 features. “The combination of features from all three layers of the CNNs, along with NNMF, allowed us to achieve high accuracy with a reduced number of features, making the system more efficient,” Attallah noted.
The implications of this research extend beyond the medical field. For maritime sectors involved in medical technology and remote healthcare, this CAD system offers a promising tool for early detection and diagnosis. Ships and offshore platforms, often with limited medical facilities, could benefit from non-invasive, radiation-free diagnostic tools that require minimal preprocessing and computational resources. This could be particularly valuable in emergency situations or during long voyages where immediate medical attention is not readily available.
Furthermore, the commercial potential of this technology is substantial. The maritime industry could explore partnerships with medical technology companies to integrate this CAD system into existing healthcare solutions. This could lead to the development of portable, user-friendly devices that can be easily deployed in various maritime settings. The system’s efficiency and accuracy could also reduce the need for specialized personnel, lowering operational costs and improving accessibility to quality healthcare.
In conclusion, the research led by Omneya Attallah represents a significant advancement in breast cancer detection using thermal imaging and deep learning. The study, published in ‘Applied Sciences’, demonstrates the efficacy of multi-layer feature amalgamation and dimensionality reduction techniques in enhancing non-invasive diagnostic methods. For the maritime industry, this technology presents exciting opportunities to improve healthcare services on ships and offshore platforms, ultimately contributing to better health outcomes for seafarers and personnel working in remote maritime environments.