40,000 women die due to breast cancer in the U.S. every year. Usually, if cancers are found early, the cancer is often curable. Mammograms are the best test available up till now, but the tests are not 100% accurate and can sometimes get false-positive results that can lead to unnecessary biopsies and surgeries.
One common cause of false positives is the seemingly high-risk lesions that look like a cause for concern on mammograms and usually contain abnormal cells when tested via a needle biopsy. This is when the patient has to make the decision to get the surgery done. Usually, the lesions tend to be benign. 1000’s women, every year, go through an unnecessary process that is not only expensive but painful as well. If the result of a mammogram showcases a lesion, a needle biopsy usually confirms whether it is cancerous or not. On average, 70% of the lesions are benign, 20% are malignant, and 10% are high-risk lesions. As artificial intelligence gets more popular especially in the healthcare sector, different innovative ways have come up in order to use AI to increase efficiency. A new initiative to detect breast cancer is a good example of AI being used for the greater good.
Artificial Intelligence is simply a computer that is capable of handling a task that typically requires human intelligence. To learn more, click here.
Using AI to assist in Detecting Breast Cancer
Supplemental screening (Screening for Breast Cancer varies depending on which model is used. The highest calculated risk should be used to determine screening recommendations) in women with dense breast tissue has resulted in increased sensitivity of cancer detection. Research from the Denise Tissue and Early Breast Neoplasm Screening (DENSE) Trial also supported supplemental screening with MRI.
Study lead author Erik Verburg, MSc, stated in a press release, “The DENSE trial showed that additional MRI screening for women with extremely dense breasts was beneficial. On the other hand, the DENSE trial confirmed that the vast majority of screened women do not have any suspicious findings on MRI.”
MRIs show the normal anatomical and physiological variation that usually require radiological review, in case it does, researchers created an AI method to minimize the radiologist’s workload. The feasibility of an automated triaging method based on deep learning was an inspiring idea for Verburg, that he and his colleagues set out to determine. The study used breast MRI data from the DENSE trial so that the deep learning algorithm can differentiate between breasts with and without lesions. The model was created with the dataset coming from seven hospitals and tested on data from eight hospitals.
A press release of the study stated, “More than 4,500 MRI datasets of extremely dense breasts were included. Of the 9,162 breasts, 838 had at least one lesion, of which 77 were malignant, and 8,324 had no lesions,”.
The deep learning model concluded that 90.7 percent of the MRIs with lesions were non-normal and marked them for radiological review. The model showcased 40 percent of lesion-free MRIs without missing any cancers. Verburg stated that through this study, in the near future one can use AI with decent malignant cancer risk-free. Though the results of the study were promising, the AI still has to achieve 60% more accuracy to be able to be used live.
The experts believe the AI-based triaging system could be used to immensely decrease the workload of the radiologists’. Verburg stated that this kind of approach will assist the radiologists to improve their overall reading time He also mentioned the fact that more time could become available for radiologists to focus on more complex breast MRI examinations. The experts involved, plan to train this deep learning model in other datasets and deploy it in the subsequent screening rounds of the DENSE trials. In conclusion, after examining the AI model’s potential usage in U.S healthcare. One can expect that a deep learning model which is trained with a sufficient dataset can produce a diagnosis, as accurate as experienced radiologists.
Want to learn more about AI in health?
Learn about the many ways AI is used in healthcare
The Rise of AI in Psychiatry
"Most comprehensive labelling tool in the market. Datasaur has saved us countless hours in building our own solution. My team lead never wants to go back to spreadsheets!"
"Operating in an industry where we have to be privacy- and security-conscientious with our data, Datasaur was the only acceptable solution for us. We recommend them for both feature set and support responsiveness."
"...information labeling tasks has been reduced by 80% which has allowed us to optimize our workflow much more, allowing us to focus on other areas that are also priorities for us..."