Image by Tyler Olson on Shutterstock
According to WHO, in 2020, breast cancer was the most common cancer diagnosed. However, it still ranks second as a leading cause of cancer deaths among women, trailing behind lung cancer.
A foremost cause for breast cancer deaths is late or limited detection rates. There is a gaping inefficiency in reaching life-saving treatment as basic screening tests are not available for all.
Through these constraints, technology and the digital transformation in healthcare are striving to create more cost-effective options and bridge gaps to reduce breast cancer-related deaths.
Exploring the breast cancer epidemic
While it may seem that the number of breast cancer cases is increasing, the statistics, to an extent, reflect the improvements in the screening of breast cancer cases.
While a notable feat, many new generation risk factors should also be considered among this surge in breast cancer cases.
For one, factors such as having a sedentary lifestyle or consuming a diet low in nutrition increase overall cancer risk. Another is the use of hormones, either for contraception or during menopause, contributing to breast cancer risk.
In addition to these, the CDC has listed factors such as age, genetic mutations, a family or personal history of breast cancer, and radiation exposure among uncontrollable risk factors increasing chances of developing breast cancer.
However, while the risk of developing breast cancer today is probably at an all-time high, we live in a generation where the prevention of breast cancer fatalities is also at its optimum.
Technological advancement in breast cancer diagnosis
Today, the most frequently used imaging methods to detect breast cancer are mammograms, ultrasounds, and breast MRI.
While these have been the mainstay for breast cancer diagnosis, technology has encouraged significant fine-tuning of existent modalities. An example is the 3D breast mammogram, digital breast tomosynthesis (DBT).
An evolution from the basic digital mammogram (DM), now 3D breast segments can be visualized, enabling physicians to have more detailed images of the complex breast tissue.
While incremental costs for DBT tend to be lower than DM than earlier cost analyses, DMT is more expensive than DM when recall costs are factored in.
Thermography, electrical impedance imaging spectroscopy, and diffuse optical imaging are more recent emerging techniques for breast cancer screening. These experiments with the different properties of both the normal and cancerous cells, in turn, assist with identifying tumor cells among breast tissue.
For example, thermography creates a heatmap image that AI can analyze for temperature variations caused by active cancer cells. In addition, an infrared camera that can be easily transported to rural areas is much more cost-effective than heavy machines.
A large portion of these technological advances has to go through rigorous testing to eliminate the struggles of misdiagnosis.
This is where AI technology steps in to reduce the chances of both imaging and human error. However, research shows thermography accurately detects breast cancer only 43% of the time, so it shouldn’t completely replace digital X-ray mammography.
How can AI be used to diagnose breast cancer accurately?
The key toward reducing the global cancer burden is to reduce errors in its early detection. And this naturally reduces the financial burden accumulated during ongoing treatment.
Artificial intelligence technology aims to improve the accuracy with which cancer is diagnosed. This, in turn, will optimize and minimize the treatment course required to tackle cancer following its detection.
AI technology is coupled with existing imaging modalities like digital X-ray machines to increase the accuracy with which breast cancer is detected. This makes having imaging systems such as mammograms or ultrasounds available at all screening facilities a must.
However, this can often hinder early diagnosis in rural settings, primarily due to financial constraints.
Coupling AI mobile devices and alternative technologies, such as detecting thermal stamps of breast cancer cells, have shown some promise toward building cost-effective tools for breast cancer screening.
Other issues can be caused by relying on only mammograms for AI algorithms. For example, women with dense breasts require an MRI scan for breast cancer screening. This means that building AI systems for breast cancer screening requires data sets to be pulled in through all avenues of cancer screening.
Additionally, these systems have to identify the variations in cancer cells. Data collected over time through cancer screening enables machine learning of AI systems to estimate with increasing accuracy malignant cells over benign ones.
AI systems increase their accuracy when the volume of data is large. Evaluating retrospective data then becomes crucial for future diagnostic algorithms. For example, while previous cancer screening results can have a radiologist bias for cancer, negative outcomes that resulted in cancer can be studied to improve the detection rate.
AI shaping the future for breast cancer care
Technological advances are not meant to make physicians obsolete. Instead, these systems are meant to advocate for the efficiency of healthcare delivery.
Sometimes, cancer prognosis is bleak due to delayed diagnosis and inappropriate treatment strategies. However, with early and relatively accurate detection, clinicians can now have a more holistic idea of approaching cancer management.
Additionally, AI technology fuels research into breast cancer management. Following recovery, approximately 30% of patients have a risk of relapse. Being able to detect patients at this high risk of relapse can also govern the initial methods of treatment that are employed.
In breast cancer care, AI can enable more effective communication between patients and their care providers. In addition, AI systems should also observe workflows in effective cancer management. This can help refine how healthcare is delivered, personalizing breast cancer management in the long run.
Limitations AI screening systems encounter
The rapid advancement of technology does face significant roadblocks. A notable mention is data privacy.
A diagnosis of breast cancer is highly personal. Therefore, the easy communication between such information systems can be a breach of individual privacy.
Training AI systems becomes a crucial facet. Attempts can be made toward training systems in-house for detecting breast cancer. In addition, breast cancer detection and test results can be shared with research centers by building systems that remove personal details when forwarding information.
Data collected can also experience bias from those building and refining the systems. While the key aim of AI systems is to eliminate human error, human bias might, unfortunately, mitigate efficient results from the system.
Finally, it is important to remember that doctors are both professionally and legally bound to all the decisions they make. For machines, this is a gray area. Therefore, legal implications of both false positives and false negatives need to be explored while machines are in their developmental phases.
Cancer was a diagnosis equivalent to death a few decades ago. However, medical science has vastly improved the outcomes of patients who have been diagnosed with cancer. From the fine-tuning of imaging techniques to the continual optimization of treatment regimens, breast cancer prognosis has significantly improved.
Technology has managed to take this one step further.
It is bridging gaps that once made detection of breast cancer in remote locations impossible. Technological advances also limit the requirement of expensive devices to screen for cancer cells.
Cataloging data globally will play a vital role in refining how AI algorithms perform in the long run. For example, AI aims to detect cancer early, enabling early medical management and lowering breast cancer deaths.