A variety of technological innovations have made modern-day cancer research exciting, as findings of various studies have provided hope for next-generation treatment methods that could save countless lives. Machine learning methods of data analysis are one of many trending oncology research topics, and studies have found numerous benefits of using machine learning as part of the cervical cancer screening process. Let’s take a deep dive into the current research on this promising new application of technology in cancer treatment.
What Is Machine Learning?
Machine learning is the science of using software programs that improve automatically through data analysis and experience. A machine learning algorithm does not need to be programmed by a human. Rather, the algorithm builds a model based on a sample set of data and uses this data to make relevant predictions or decisions.
Medicine and several other fields have used machine learning in a variety of applications. Medical researchers and clinicians can use machine learning data to more accurately study diseases, treat patients, and develop new medicines. Machine learning has also been used to identify gene sequences and quantify molecular variables related to diseases.
As machine learning and other forms of artificial intelligence continue to develop, oncologists are hopeful that these technologies could soon drastically improve outcomes for patients with cervical and other forms of cancer.
The Importance of Screening in Cervical Cancer Prevention
Even though cervical cancer can be prevented, over 13,000 women in the United States are diagnosed with the disease each year, and 30% of these individuals lose their lives. There are two main ways to prevent cervical cancer, which can be highly effective when implemented properly:
HPV Vacccinations—Human papillomavirus (HPV) is the cause of approximately 91% of cervical cancer diagnoses. There are over 100 strains of HPV, but over 70 percent of HPV-related cervical cancer cases are related to HPV-16 and HPV-18. Thus, vaccinations that prevent people from contracting HPV are an enormously powerful tool in the fight against cervical cancer.
Regular Screenings—All women should have regular reproductive health screenings during their reproductive years for several different reasons. Regular screenings can help doctors quickly identify abnormalities and begin treatment as soon as possible. American Cancer Society data estimates that between 60 and 80 percent of American women who have been diagnosed with invasive cervical cancer did not have a Pap test in the five years before their diagnosis.
How Cervical Cancer Screening Can Improve Through Machine Learning
Cervical cancer screening faces two primary challenges globally: the prevalence of false positives that can cause overtesting and increased medical costs, and the lack of vaccinations, testing, and women’s healthcare in parts of the developing world. A recent article in the Journal of the National Cancer Institute (JNCI) posits that machine learning could be used to address both of these obstacles.
According to the article, researchers used a cloud-based slide imaging platform to focus on two specific biomarkers that are associated with cervical cancer. The findings showed that the use of this screening technology showed significantly reduced false positives and that follow-up exams for abnormal cells in the cervix were reduced by almost one-third. Additionally, the cloud-based nature of the technology means it could theoretically be used remotely to analyze cervical cancer screenings from anywhere on Earth.
Methodology of Deep-Learning Study on Cervical Cancer Screening Applications
The JNCI study was based on the findings of a study by Nicolas Wentzensen et al. The team of researchers used deep learning methods in the cervical cancer screening process. Researchers used a technology called CYTOREADER, which is a cloud-based system that implements whole-slide scanners for imaging purposes.
The team trained deep-learning algorithms for automated evaluation of dual-stain positive cells by using tiles from whole slides, which contained either individual or small amounts of epithelial cells.
Algorithms were used to identify the number of dual-stain positive cells on each imaging slide by detecting how many tiles were above a designated likelihood threshold. Slides were labeled positive if the number of DS-positive cells was above a specific cutoff. Researchers then manually evaluated the tiles to confirm the number of DS-positive cells.
Future Applications of Machine Learning in Cervical Cancer Screening
The next step in applying this screening method on a global scale would be developing and distributing technology, such as a smartphone, tablet, or another smart device, that could be used to share these images. Areas of the world that have been limited in their ability to prevent cervical cancer because of a lack of screening infrastructure stand to benefit enormously from this technology if they have the tools they need to use it. Additionally, machine learning in cancer screenings has the potential to revolutionize cervical cancer treatment on a global scale, even in nations that already have advanced healthcare systems.