AI for Breast Cancer Diagnosis in Mammography

Focusing on artificial intelligence’s impact on the healthcare systems, there are several studies about artificial intelligence performing at parr with seemingly radiologist levels in the evaluation of digital mammography that would improve breast cancer screening accuracy and efficiency. Mammography is usually performed to check for the early signs of breast cancer when cancer is still treatable. Breast cancer is the most common cancer among women; it is also the second leading cause of cancer death in women after lung cancer. The most common risk factor for developing breast cancer is being diagnosed over the age of 55. And to be able to reduce breast cancer-related deaths, screening for breast cancer with mammography is launched in many countries over the worldthe past years. Thanks to the screening and development of treatments, breast cancer mortality has reduced but remained the primary reason for female cancer death. 

Breast cancer screening with mammography is done differently in different countries. However, predominantly, like in the US, breast cancer screening is institution-based. Women will be referred to or go by themselves to their gynecologist or physician who is more often than not, present at a breast imaging center and is affiliated with a hospital. The rest of the process will be organized  depending on the institution or hospital.

On the other hand, current screening programs are taking too much work due to the large number of women screened per detected cancer and the use of double reading to countercheck results which also leads to additional cost. According to the Journal of the National Cancer Institute, despite this practice, up to 25% of visible cancers seen on the mammograph are not detected at a screening. Due to the increasing scarcity of radiologists in other countries, alternative strategies are necessary to allow the continuation and development of current screening programs. Furthermore, it is extremely important to prevent lesions visible in digital mammography from being overlooked. 

A study published by the Journal of National Cancer Institute, describes computer-aided detection systems that have been developed to automatically detect breast lesions in mammograms. The implementation of digital mammography to diagnose breast cancer has propelled the refinement of automated detection techniques for breast cancer. However, up to this date, there are still no studies about traditional computer-aided systems that can directly improve the performance of breast screening or cost-effectiveness due to low specificity.  Artificial intelligence technology, on the contrary, is rapidly developing and growing due to achieving the perfect algorithms based on intensive and deep learning convolutional neural networks. These advancements are successfully automating difficult tasks. In medical imaging, comprehensive research is also rapidly closing the space between humans and computers. This has the potential to continuously improve the benefit to harm ratio of breast cancer screening programs.

According to the, Deep Learning in Mammography: Diagnostic Accuracy of a Multipurpose Image Analysis Software in the Detection of Breast Cancer study by Becker et al, and the Large scale deep learning for computer-aided detection of mammographic lesions study by Kooi et al., in the past few years, several deep learning-based algorithms have been developed for the analysis of mammograms, and some of these algorithms have shown favorable results when compared to radiologist but in a very finite and homologous scenario. 

In conclusion, AI holds an incredible promise for the development of medical science as it’s vast and unlimited. However, like other emerging technologies that we already have, the use of artificial intelligence for the automation of breast cancer detection in mammography requires a lot of research and evaluation for clinical effectiveness before it can be adopted completely. As we have already learned from experience,adopting new technologies too quickly could harm us and could be a costly mistake.