Artificial Intelligence is challenging conventional ways of cancer screening

As medical science’s understanding of the evolution of cancer deepens, it amplifies the need for timely detection and, eventually, prevention. In 2020, Cancer Research UK released a strategic blueprint that mapped out the crucial early cancer detection and diagnosis domains. The roadmap highlighted the areas that require urgent intervention to expedite advancements in this crucial area.

When considering risk assessment, it’s essential to emphasize the crucial contribution of technology, specifically AI and machine learning. Through the utilization of advanced data and integrative analytical techniques, medical experts can amalgamate a wide array of information sources tailored to each individual. This valuable data might span from germline genome data and comprehensive medical records to the health data captured by smartphones, enabling a comprehensive evaluation of an individual’s cancer risk.

Artificial Intelligence has shown promising results in detecting breast cancer

According to a study published in The Lancet Oncology, Artificial Intelligence has shown a level of proficiency that matches the abilities of two human radiologists in evaluating breast cancer scans. The research sheds light on AI’s capability to reduce the time doctors require to analyze mammograms by half. Besides, initial tests have shown that AI might even detect more breast cancer than human detection rates.

During a comprehensive trial, a group of 80,000 women averaging 54 in age went through detailed scrutiny. The research was done between April 2021 and July 2022, during which the group underwent meticulous screening. With the help of a hybrid approach, the assessment used the power of AI detection alongside the expertise of multiple radiologists, resulting in the examination of 39,996 patients.

As for the remaining 40,024 patients, the conventional screening method was used, which involved two human radiologists. Interestingly, during the AI review, a total of 244 cases of breast cancer were successfully detected, while the radiologists managed to identify 203 cases within the group. It means the AI approach led to the detection of 41 additional cancer cases, out of which 19 were identified as invasive and 22 were categorized as in situ cancers.

Artificial Intelligence challenging conventional screening methods

The study raises a notable point regarding potential overdiagnosis due to the higher count of in situ cancers detected by AI compared to the traditional screening approach. The report highlights that AI-assisted screening identified 60 in situ cancers, while standard screening identified 38.

However, the analysis of false-positive screenings revealed a reassuring outcome: the AI-supported screening did not yield more false positives than the standard method. The average false-positive rate was consistent at 1.5% for both the AI-assisted and standard screening groups.

According to the researchers, the utilization of AI in the screening process has the potential to offer a significant benefit by alleviating the reading workload for radiologists. This reduction could amount to an impressive 44%, a particularly valuable effect, especially when the challenges of finding qualified people for these roles are rising.

Usually, the assessment of mammograms is a process that requires two radiologists. However, leveraging AI to screen can potentially reduce this requirement to only one radiologist. This could be a ray of hope in the wake of projections made by Radiology Business, which shows that a potential nationwide shortage of radiologists could occur in the US by 2024. This looming scarcity can be credited to factors including radiologist burnout and the influx of retirements within the field.

Deciding on potential patients for screening is crucial

So, how can these strategies be used to enhance breast cancer screening and early detection? The first and foremost step involves the critical task of determining the right person to screen. Traditionally, mammography screening has been guided by age and family history, which is derived from epidemiological data on cancer incidence.

However, recent studies that are focused on younger women have brought forward several factors that provide a higher degree of precision in predicting a person’s risk of developing cancer. These factors can be anything from domains, phenotypic attributes like breast density, as well as lifestyle elements such as smoking history and diet. This knowledge allows for creating more comprehensive risk models, empowering screening programs to be customized as per an individual’s profile.

Moreover, the issue still haunts the screening method. The prevailing gold-standard technique is mammography. However, promising results are coming forward from different modalities, including contrast-enhanced mammography and abbreviated magnetic resonance imaging. S costs continue to drop, and accessibility improves, the chances for these advanced imaging methods to take over X-ray mammography improve. Moreover, in-depth research is being carried out on pan or multi-cancer liquid biopsy screening tests, which have displayed encouraging potential, particularly for breast cancer.

Can Artificial Intelligence replace radiologists anytime soon?

Ductal carcinoma in situ (DCIS) is a crucial aspect of the screening process. As advancements in screening emerge, we will witness more detection of DCIS cases, with only a few progressing to breast cancer. That said, extensive research needs to be carried out to empower medical professionals with the knowledge required to counsel women on the specific risks linked with the biological nature of these lesions.

Artificial Intelligence has taken up a progressive role across diverse facets of early breast cancer detection. AI’s prowess has been observed in detecting cancers from medical imaging. However, the question still lingers about whether it can match or surpass radiologists’ capabilities. If it does, it holds the key to seamless integration into clinical pathways, augmenting existing methods and expediting the whole process. Besides, a different application of AI in risk assessment is starting to emerge, with studies showing that AI can predict future cancer development based on images, even in cases where no discerning lesions are currently found.

It goes without saying that fostering multidisciplinary collaboration that features a diverse range of experts, such as biologists, physicists, oncologists, chemists, epidemiologists, and software engineers, is very crucial for successfully addressing the several components of advancing breast cancer screening and early detection. The true victory lies in our ability to operate as a team, understanding that collective efforts are paramount for achieving success.