Early Detection of Breast Cancer Using AI-Based ImagingTechniques

Abstract

There is one of the major causes of death from cancer among women in various parts of the world. Detection is early, but mammographic interpretation is prone to human error and subjectivity. Artificial intelligence (AI) is integrated into diagnostic imaging, promising to improve the efficiency and accuracy of early detection of breast cancer. The objectives of this study were to assess the diagnostic performance of an AI based imaging system in the early detection of breast cancer and to determine whether the imaging system is more effective than experienced radiologists. One thousand mammographic images involving 380 confirmed malignant and 620 benign cases were analyzed retrospectively. An AI model, based on a convolutional neural network (CNN), was developed and tested to classify lesions. The performance metrics of sensitivity, specificity, accuracy and area under the curve (AUC) were then compared with senior radiologists' readings with statistical tests such as McNemar test. The sensitivity of the AI model was 94.2%, specificity was 88.7% and overall diagnostic accuracy of 91.1% which was better than radiologists, whose metrics were 89.5%, 85.2% and 87.3% respectively. Finally, the AI model was trained to take the set to 0.957 versus the 0.902 of radiologist. Results showed statistically significant improvements (p = 0.001). Moreover, the AI system remained robust with accuracy across various breast densities, particularly for dense breast tissue.AI based imaging systems hold promise as the potential to aid early breast cancer detection by increasing diagnostic accuracy and accuracy at the cost of human error. These tools may be used as powerful clinical decision support systems, especially in resource limited settings. But wider clinical adoption will require further multi center studies and integration of explainable AI features.