Google's AI System Outperforms Medical Experts in Diagnosing Breast Cancer
Sat, April 10, 2021

Google's AI System Outperforms Medical Experts in Diagnosing Breast Cancer

Breast cancer can occur in both men and women, although it's far more common in the latter / Photo by: Katarzyna Białasiewicz via 123RF

 

Breast cancer can occur in both men and women, although it's far more common in the latter. The World Health Organization reported that breast cancer is the most common cancer in women both in developed and less developed nations, impacting 2.1 million women around the world every year. It is estimated that in 2018, more than 627,000 women died of breast cancer, which is approximately 15% of all cancer deaths among women. While the illness has been affecting millions of women across the world, women in less developed countries are more prone, with almost 50% of breast cancer cases and 58% of deaths in the world occurring in those nations. 

As of January 2019, there were more than 3.1 million American women with a history of breast cancer. These included women currently being treated and those who have finished treatment. However, a woman’s risk of breast cancer doubles if she has a first-degree relative (mother, sister, daughter) who has been diagnosed with breast cancer. Additionally, about 85% of breast cancer cases occur in women who have no family history of the illness. This usually happens due to genetic mutations as a result of the aging process and life in general rather than inherited mutations.

Thus, early diagnosis is extremely important to patients and their families so they can start with the treatment immediately. Various methods are used to diagnose this illness such as breast cancer screening tools including mammography, clinical breast exam, and breast self-exam. However, these methods, particularly mammography, are not accurate. The American Cancer Society reported that radiologists miss about 20% of breast cancers in mammograms. On the other hand, many women get a false positive, showing them having cancer even though they don’t have it, causing them undue concern.

According to MIT Technology Review, an online magazine that aims to bring about better-informed and more conscious decisions about technology through authoritative, influential, and trustworthy journalism, screening tests of breast cancers have high rates of error. Reports showed that about 1 in 5 screenings are false negative or failing to find breast cancer even when it’s present. More than 50% of women who receive annual mammograms also get at least one false alarm over a 10-year period. 

New AI System in Diagnosing Breast Cancer

Artificial intelligence has played a huge role in improving diagnosis. Recently, researchers from Google Health and Imperial College London introduced an AI system that can identify breast cancer with greater accuracy than human experts. They hope that this tool can prove to be a breakthrough in the fight against cancer.

According to Nature, a leading multidisciplinary science journal, the researchers used a large scale of datasets to train and validate the AI algorithm. The datasets were composed of 25,856 mammograms from women in the UK and 3,097 women in the US. The AI system was then used to identify the presence of breast cancer in mammograms of women. The participants had either a normal follow-up imaging results at least 365 days later or biopsy-proven breast cancer. The images were reviewed by medical professionals. The difference this time is that the tool had no patient history to inform its diagnosis.

The findings of the study published in Nature showed that the AI system was able to decrease errors in both false positive and false negative diagnoses. It reduced false negatives by 9.4% and false positives by 5.7% for US patients. Meanwhile, it was a reduction of 2.7% for false negatives and 1.2% for false positives in the case of UK patients. 

"This is promising early research, which suggests that in the future it may be possible to make screening more accurate and efficient, which means less waiting and worrying for patients and better outcomes,” said Sara Hiom, the director of cancer intelligence and early diagnosis at the Cancer Research UK (CRUK) Imperial Center. 

The researchers also conducted a separate experiment to test the AI system’s ability to generalize. By training the model using only mammograms from UK patients and evaluating its performance on US patients, the team discovered that it reduced false negatives and positives by 8.1% and 3.5%, respectively, overall, outperforming human radiologists. 

"Our team is really proud of these research findings, which suggest that we are on our way to developing a tool that can help clinicians spot breast cancer with greater accuracy,” said Dominic King, the UK lead at Google Health. 

Limitations

The AI system can cause a significant change in better diagnosis of breast cancer. According to BBC News, an operational business division of the British Broadcasting Corporation responsible for the gathering and broadcasting of news and current affairs, it will save radiologists time since reading mammograms is a time-consuming work. 

"This went far beyond my expectations. It will have a significant impact on improving the quality of reporting and also free up radiologists to do even more important things,” stated professor Ara Darzi, the report co-author and director of CRUK. 

However, the researchers stated that clinical trials will be needed to further assess the utility of this tool in medical practice. This is because the actual diagnosing of breast cancer is much more complicated and potentially more diverse than the type of controlled research environment reported in this study. It should also be noted that the demographics of the population in this research are not well defined. 

Nonetheless, the new AI system shows that it is possible to overcome one of the biggest challenges in the healthcare system.

Recently, researchers from Google Health and Imperial College London introduced an AI system that can identify breast cancer with greater accuracy than human experts / Photo by: Uladzik Kryhin via 123RF