In honor of Breast Cancer Awareness Month, let’s dive into three AI-powered solutions set to change the way we detect, diagnose, and predict the disease. One deep-learning model is expected to someday identify risk of breast cancer up to five years into the future.
Kinda makes Siri and Alexa’s weather reports, Netflix updates, or reminders to pick up tacos a little…well…let’s call it anticlimactic.
1. Improved Early Breast Cancer Detection
“Women with dense breasts are subject to the masking effect of mammographic density and its association with breast cancer risk. The objective of the DIMASOS 2 trial is to test whether combined mammography and ultrasound exams can improve early cancer detection and if this can feasibly and cost effectively be done in routine screening workflow,” says Sylvia H. Heywang-Köbrunner, MD, Head of Referenzzentrum Mammographie München (Reference Center Mammography Munich) and internationally recognized for her pioneering work in contrast-enhanced breast MRI and modern biopsy procedures.1
The DIMASOS 2 trial is distinguished by its unique densitas densityai™ software, developed by Densitas Inc, a leading provider of AI solutions for digital mammography.2 The software, currently destined for about 24 clinics throughout Germany, will provide breast density measurements at point of care in order to identify women who would benefit from supplemental breast cancer screening. With standardized and reproducible patient-specific risk estimates – which would then flag a need for supplemental breast cancer screening – we may be able to significantly improve earlier breast cancer detection.
The densitas densityai software delivers fully automated, standardized, and reproducible breast density assessments from standard DICOM clinical use mammograms. Results are generated by two distinct algorithms that decouple the breast density assessment into quantitative and qualitative scales in alignment with the ACR BI-RADS 4th and 5th edition density classification system.1
2. Improved Breast Cancer Risk Prediction
Researchers from the Massachusetts Institute of Technology (MIT) and Massachusetts General Hospital (MGH) believe they’ve developed an advanced AI-powered tool to predict a woman’s future risk of breast cancer, according to a new study published in Radiology, May 2019.3
The ultimate hope, according to study lead author Adam Yala, a PhD candidate at MIT, is to tailor breast cancer screenings to individual women.4 “There’s much more information in a mammogram than just the four categories of breast density,” Yala says. “By using the deep learning model, we learn subtle cues that are indicative of future cancer.”
To find these “subtle cues,” the research team developed a deep learning (DL) model that uses full-field mammograms and traditional risk factors. Results thus far strongly suggest that this DL model is more accurate than the Tyrer-Cusick model (version 8), a current clinical standard.3
Here are three key takeaway points reported in the Radiology article:3
- A DL mammography-based model identified women at high risk for breast cancer and placed 31 percent of all patients with future breast cancer in the top risk decile compared with only 18 percent by the Tyrer-Cuzick model (version 8).
- The hybrid DL model is equally accurate for white and African American women, whereas the Tyrer-Cuzick model AUC was 0.62 and 0.45 for women who were white and African American, respectively.
- After comparing the hybrid DL model with breast density, researchers found that patients with nondense breasts and model-assessed high risk had 3.9 times the cancer incidence of patients with dense breasts and model-assessed low risk.
3. More Accurate Breast Cancer Diagnoses
UCLA researchers are developing an AI system that may help pathologists read biopsies with greater accuracy – and thus result in better breast cancer detection and diagnoses. To test the system, described in a study published in JAMA Network Open, August, 2019, researchers compared system readings to independent diagnoses made by 87 practicing US pathologists. While the AI program nearly performed as well as human doctors in differentiating cancer from non-cancer cases, the AI program outperformed doctors when differentiating DCIS from atypia – considered the greatest challenge in breast cancer diagnosis.5
The study’s researchers feel positive that AI can provide more accurate readings consistently – because by drawing from a large data set, the system will recognize patterns in the samples that are associated with cancer but are difficult for humans to see.
Quicker, consistent accuracy makes an enormous difference, explains Joann Elmore, MD, the study’s senior author and a professor of medicine at the David Geffen School of Medicine at UCLA. “It is critical to get a correct diagnosis from the beginning so that we can guide patients to the most effective treatments.”
Despite all the OMG-ing and WOW-ing over how artificial intelligence will soon “personalize” a woman’s breast screening, the issue of breast health and breast cancer rests between two humans – a doctor and a patient. In fact, AI’s goal to “personalize” breast screening is about being able to apply huge amounts of data to one specific woman and then make sense of it all on an individual level. When it comes to taking that personalized data and pairing it with a personalized breast cancer treatment, that is still very much – and hopefully forever – the physician and patient’s choice.
Resources
