Fine-Tuning AI for Breast Screening

AI has shown in research studies it can help radiologists interpret breast screening exams, but for routine clinical use many questions remain about the optimal AI parameters to catch the most cancers while generating the fewest callbacks. Fortunately, a massive new study out of Norway in Radiology: Artificial Intelligence provides some guidance. 

Recent research such as the MASAI trial has already demonstrated that AI can help reduce the number of screening mammograms radiologists have to review, and for many low-risk cases eliminate the need for double-reading, which is commonplace in Europe. 

  • But growing interest in breast screening AI is tempered by the field’s experience with computer-aided detection, which was introduced over 20 years ago but generated many false alarms that slowed radiologists down. 

Fast forward to 2024. The new generation of breast AI algorithms seems to have addressed CAD’s shortcomings, but it’s still not clear exactly how they can best be used. 

  • Researchers from Norway’s national breast screening program tested one mammography AI tool – Lunit’s Insight MMG – in a study with data obtained from 662k women screened with 2D mammography from 2004 to 2018. 

Researchers tested AI with a variety of specificity and sensitivity settings based on AI risk scores; in one scenario, 50% of the highest risk scores were classified as positive for cancer, while in another that threshold was set to 10%. The group found …

  • At the 50% cutoff, AI would correctly identify 99% of screen-detected cancers and 85% of interval cancers. 
  • At the 10% cutoff, AI would detect 92% of screen-detected cancers and 45% of interval cancers 
  • AI understandably performed better in identifying false-positive cases as negative at the 10% threshold than 50% (69% vs. 17%)
  • AI had a higher AUC than double-reading for screen-detected cancers (0.97 vs. 0.88)

How generalizable is the study? It’s worth noting that the research relied on AI of 2D mammography, which is prevalent in Europe (most mammography in the US employs DBT). In fact, Lunit is targeting the US with its recently cleared Insight DBT algorithm rather than Insight MMG. 

The Takeaway

As with MASAI, the new study offers an exciting look at AI’s potential for breast screening. Ultimately, it may turn out that there’s no single sensitivity and specificity threshold at which mammography AI should be set; instead, each breast imaging facility might choose the parameters they feel best suit the characteristics of their radiologists and patient population. 

USPSTF’s Mammography Letdown?

Last year’s relief that the USPSTF would lower its recommended starting age for breast screening to 40 gave way to frustration this week that the group did not go farther in its final decision on mammography recommendations. 

In a series of papers in JAMA journals this week, the USPSTF tackled a range of breast screening issues, from the age at which screening should start to whether modalities like ultrasound and MRI should be used to supplement conventional mammography.

That was the good news. The bad news is that breast screening advocates mostly got shut out on a variety of other issues, with the USPSTF … 

  • Advising that breast screening be conducted biennially (every two years), rather than annually as most women’s imaging advocates would prefer
  • Declining to raise the recommended upper limit for screening from 74 to 79
  • Declining to recommend supplemental screening with MRI or ultrasound for women with dense breast tissue, even as women express frustration with the lack of reimbursement for these exams

On the positive side, the USPSTF finally weighed in on DBT, stating that the 3D mammography technology is equivalent to digital mammography for breast screening. 

  • But in another disappointment, the group said it couldn’t find any studies stating that DBT was better than 2D digital mammography. 

Given the fierce battles that have been fought over screening guidelines in the last 15 years, what made the USPSTF change its mind on mammography’s starting age? 

  • One big factor is the 2% annual rise in breast cancer incidence in women in their 40s from 2015 to 2019; the higher mortality rates among Black women was another issue (see story below in The Wire).

The Takeaway

The USPSTF’s move to lower its recommended starting age for screening mammography is a welcome – if overdue – change for women, who for 15 years have borne the brunt of the group’s conservative approach to guideline formation. The question remains, is the USPSTF making the same mistake all over again when it comes to supplemental imaging and annual screening? And how long will women have to wait this time until it sees the light?

Breast Screening Goes Green

Earth Day will be celebrated on April 22, and the event is a good opportunity to step back and take a look at medical imaging’s (not insignificant) contribution to climate change. Fortunately, a new paper in Health Policy details how one imaging service – breast screening – can be made more environmentally friendly. 

Previous studies have documented that medical imaging is a substantial contributor to greenhouse gas emissions, given the massive energy consumption required to keep all that big iron humming. 

  • Researchers have recommended a variety of solutions to reduce radiology’s environmental footprint, from powering equipment down overnight to switching to alternative energy sources to power medical facilities. 

The new study gets even more specific, analyzing the greenhouse emissions inherent in cancer screening – in particular patient travel – and offering ways to make it more planet-friendly. 

  • Researchers reviewed cancer screening programs in the Italian region of Tuscany, quantifying the CO2 emissions for different screening services. 

Greenhouse gas emissions could be cut dramatically by switching from a provider-centric model that requires patients to travel to centralized screening facilities to one in which mobile vans were sent into the field. Using model calculations for mammography screening, they found that in one district alone …

  • Breast screening was the most polluting cancer screening service, mostly because it had the highest number of invitees (3.4k women) traveling for screening
  • Institution-based breast screening generated CO2 emissions of 35,870 kgCO2-eq/km annually
  • Mobile breast screening had emissions of 805 kgCO2-eq/km – just 2.2% of emissions from site-based screening

The study is unique in that it views sustainability and environmental pollution as a healthcare issue that’s fully within the purview of providers to address. 

The Takeaway

The new study outlines a holistic approach to healthcare services that – right now – many US providers might believe is outside the scope of their operations. But as Earth Day approaches, it’s worth at least considering how in years to come healthcare could be delivered within a broader context of social and environmental stewardship.

MASAI Gets Even Better at ECR 2024

One of the biggest radiology stories of 2023 was the release of impressive interim results from the MASAI study, a large-scale trial of AI for breast screening in Sweden. At ECR 2024, MASAI researchers put an emphatic cap on the conference by presenting final data indicating that AI could have an even bigger impact on mammography screening than we thought. 

If you remember, MASAI’s interim results were published in August in Lancet Oncology and showed that ScreenPoint Medical’s Transpara AI algorithm was able to reduce radiologist workload by 44% when used as part of the kind of double-reading screening program that’s common in Europe.

  • Another MASAI finding was that AI-aided screening had a 20% higher cancer detection rate than conventional double-reading with human radiologists, but the difference was not statistically significant. 

That’s all changed with the final MASAI results, presented at ECR on March 2 by senior author Kristina Lång, MD, of Lund University.

  • Lång presented data from 106k participants who were randomized to either screening with Transpara V. 1.7 or conventional double reading without AI.

Transpara triaged mammograms by giving them a risk score of 1-10, and only those classified as high risk received double reading; lower-risk mammograms got a single human reader. In the final analysis, AI-aided screening … 

  • Had a 28% higher cancer detection rate per 1k women (6.4 vs. 5.0), a difference that was statistically significant (p=0.002)
  • Detected more cancers 10-20 mm (122 vs. 79)
  • Detected more cancers of non-specific histologic type (204 vs. 155)
  • Detected 20 more non-luminal A invasive cancers and 12 more DCIS grade 3 lesions

The Takeaway

When combined with the Lancet Oncology data, the new MASAI results indicate that AI could enable breast radiologists to have their cake and eat it too: a lower workload with higher cancer detection rates. 

When Access to Screening Isn’t Enough

A new study published this week in JAMA Network Open indicates that – even when women have access to breast screening – other factors can limit mammography’s life-saving impact. Researchers found that women with more unmet social needs had lower breast screening rates and higher rates of advanced disease – even if they had access to a mammography center. 

Research into social determinants of health – the racial, demographic, and environmental factors that can affect the quality of a person’s health – have gained steam in the last several years.

In the new study, researchers noted that unmet social needs can include housing instability, social isolation, food insecurity, and transportation challenges, and these needs can occur even in high-income areas with access to screening mammography. 

  • They studied the issue in Miami-Dade County, Florida, where all women 200% below the poverty line have access to no-cost screening mammography at safety net hospitals – in theory removing cost as a barrier to breast screening.

Researchers studied 336 women who filled out a survey on social needs; of these, 62% self-identified as Hispanic, 19% as Black, and 19% as White, and 76% had screening mammograms. Researchers found a lower odds ratio for getting a mammogram due to …

  • An increasing number of unmet social needs (OR=0.74)
  • Increasing age at diagnosis (OR=0.92)

Patients were also more likely to present with late-stage disease if they …

  • Had two or more unmet social needs (33% vs. 18%)
  • Had problems with their home utilities (17% vs. 5%) or childcare access (12% vs. 3%)
  • Were presenting to a safety net hospital (31% vs. 18%)

The authors noted that although no single unmet social need was found to have a statistically significant impact on screening mammography rates, multiple needs piling up could “overwhelm” patients so they can’t find the time to schedule preventive health check-ups. 

The Takeaway

The new findings offer a more complex view of breast screening disparities beyond just access to mammograms. Public health authorities and hospitals providing women’s health services may need to offer screening of at-risk patients and a broader range of services in order to make sure that the life-saving benefits of mammography are enjoyed on a wider – and more equitable – scale.

A New Breast Imaging Option?

When it comes to mammography screening for women with dense breast tissue, radiologists have long looked for alternatives to established modalities like MRI and ultrasound. In a paper in Radiology: Imaging Cancer, researchers put a new twist on an older technology, positron emission mammography (PEM). 

Molecular imaging technologies like PEM have been investigated for years as potential adjuncts to conventional mammography due to the challenges X-ray imaging has with dense breast tissue. 

  • These technologies have carried different names – PEM, breast-specific gamma imaging, molecular breast imaging – but in the end all have fallen short due to the higher radiation dose they deliver compared to mammography. 

But Canadian startup Radialis has developed a new version of PEM with its Radialis PET Imager that drastically cuts radiation dose by targeting specific organs, enabling clinicians to use far lower doses of radiopharmaceuticals. The company received clearance for the system in 2022. 

  • Radialis touts its system as having high spatial resolution and a small field of view thanks to digital detectors with thousands of silicon sensors that can be placed next to the target organ; this makes it well-suited for imaging specific organs like the breast.

In the new paper, Canadian researchers tested the Radialis system as an adjunct to X-ray mammography in a pilot study of 25 women recently diagnosed with breast cancer. 

  • They wanted to see if PEM performed as well as breast MRI, but with fewer false positives and a radiation dose closer to screening mammography.  

Women underwent PEM at three FDG dose levels – 37, 74, or 185 MBq (for comparison, standard whole-body PET uses 370 MBq, a level that translates to a radiation exposure of 6.2-7.1 mSv). Researchers found …

  • PEM had sensitivity of 87% across all FDG dose levels (MRI was 100%)
    • The sample size was too small to detect statistically significant differences in sensitivity between dose levels
  • PEM had specificity of 95%
  • PEM detected 96% of known index malignant lesions (24 of 25), with the one miss occurring in a patient at the 37MBq level
  • PEM’s radiation dose ranged from 0.62-1.42 mSv, versus 0.44 mSv for a two-view screening digital mammogram

The Takeaway

The findings show that PEM with the Radialis system is a feasible adjunctive breast imaging modality at a radiation dose that’s mostly acceptable relative to X-ray-based mammography. But (as always) additional studies with larger patient populations are needed.

Breast Cancer in Younger Women Rises

Breast cancer rates have been rising in younger women – many of whom aren’t yet eligible for screening – and a new study in JAMA Network Open offers a perspective. 

Breast cancer mortality has dropped consistently over the last several decades, with a recent study in JAMA attributing the decline to the combination of screening and treatment. 

The problem is that even the most liberal breast screening guidelines recommend that average-risk women don’t start getting screened until age 40. 

  • This leaves younger women at risk of developing cancers that may present as more advanced disease.

The new study delves into this phenomenon, with researchers examining data from 218k women ages 20-49 who were diagnosed with invasive breast cancer from 2000-2019. Researchers found that cancer incidence …

  • Increased 0.79% annually across all women
  • Accelerated “dramatically” starting in 2016 
  • Rates per 100k women were similar for non-Hispanic Black and White women (71 & 70) across all age groups
  • But were sharply lower for Hispanic women (53)
  • Rates for Black women 20-29 and 30-39 were the highest among race and age cohorts (8 and 51)
  • Rates varied by hormone receptor status

The lower incidence rate for Hispanic women was an intriguing finding that researchers attributed to younger age at the birth of their first child, higher maternal parity, and longer periods of breastfeeding – all factors that may be changing with lower fertility rates.

  • The higher incidence rates for younger Black women are particularly problematic as these women also are more likely to present with advanced disease, which leads to higher mortality rates.

The Takeaway

The new study provides background to what’s become one of the more disturbing trends in public health. While incidence rates in younger women are still much lower than in older women, the rise raises the question of whether health interventions such as risk assessment and targeted screening – such as for younger Black women – are necessary.

Why Has Breast Cancer Mortality Fallen?

There’s no question that breast cancer mortality has fallen dramatically over the last several decades. The question is why. 

Proponents of cancer screening believe that early detection has played a major role by finding cancer and enabling treatment to start before it spreads. 

  • But that position is disputed by a vocal minority of skeptics who believe that better cancer treatments deserve most of the credit. 

A case in point was the Bretthauer et al study published in 2023, which claimed that there was no evidence to support screening’s beneficial impact on all-cause mortality. 

  • This despite a demonstrated long-term decline in mortality for the cancers targeted by the four major population-based screening programs: breast, cervical, prostate, and lung. 

A new study in JAMA offers clarity in the debate by placing a numeric value on the tools that have contributed to lower breast cancer mortality. Researchers led by Jennifer Caswell-Jin, MD, of Stanford University used simulation models based on CISNET data to analyze breast cancer mortality from 1975 to 2019, drawing the following conclusions:

  • Screening and treatment together produced a 58% decline in breast cancer mortality, from a death rate of 48/100,000 women to 27/100,000
  • 47% of the reduction was due to treatment of stage I to III cancer 
  • 29% was due to treatment for metastatic breast cancer 
  • 25% was associated with mammography screening 

The authors also discovered that the biggest improvement in breast cancer survival after metastatic recurrence (3.2 vs. 1.9 years) happened between 2000-2019. 

The Takeaway

The new results in Caswell-Jin et al should be seen as another victory for the screening community. In addition to setting a numeric figure for screening’s value, they also demonstrate the synergistic effect when screening and treatment work together to target breast cancer before it has a chance to spread. Efforts to separate the two are quixotic at best and dangerous to women at worst. 

AI’s Impact on Breast Screening

One of the most exciting radiology use cases for AI is in breast screening. At last week’s RSNA 2023 show, a paper highlighted the technology’s potential for helping breast imagers focus on cases more likely to have cancer.

Looking for cancers on screening mammography has been compared to finding a needle in a haystack, and as such it’s considered to be one of the areas where AI can best help. 

  • One of the earliest use cases was in identifying suspicious breast lesions during radiologist interpretation (remember computer-aided detection?), but more recently researchers have focused on using AI as a triage tool, by identifying cases most likely to be normal that could be removed from the radiologist’s urgent worklist. Studies have found that 30-40% of breast screening cases could be read by AI alone or triaged to a low-suspicion list.

But what impact would AI-based breast screening triage have on radiologist metrics such as recall rate? 

  • To answer this question, researchers from NYU Langone Health prospectively tested their homegrown AI algorithm for analyzing DBT screening cases.

The algorithm was trained to identify extremely low-risk cases that could be triaged from the worklist while more complex cases where the AI was uncertain were sent to radiologists, who knew in advance the cases they were reading were more complicated. In 11.7k screening mammograms, researchers examined recall rates over two periods, one before AI triage and one after, finding: 

  • The overall recall rate went from 13% before the triage period to 15% after 
  • Recall rates for complex cases went from 17% to 20%
  • Recall rates for extremely low-risk studies went from 6% to 5%
  • There were no statistically significant differences in any of the comparisons
  • No change in median self-reported perceived difficulty of reading from the triage lists compared to non-triage list, regardless of years of experience

In future work, the NYU Langone researchers will continue their study to look at AI’s impact on cancer detection rate, biopsy rate, positive predictive value, and other metrics.

The Takeaway

The NYU Langone study puts a US spin on research like MASAI from Sweden, in which AI was able to reduce radiologists’ breast screening workload by 44%. Given the differences in screening protocols between the US and Europe, it’s important to assess how AI affects workload between the regions.

Further work is needed in this ongoing study, but early results indicate that AI can triage complex cases without having an undue impact on recall rate or self-perceived difficulty in interpreting exams – a surrogate measure for burnout.

Uneven Success Against Breast Cancer

The decline in breast cancer mortality has been one of public health’s major success stories. But when you look at it from a global perspective, it’s the best of times and the worst of times. 

That’s because success in fighting breast cancer has been uneven around the world. While countries in North America, Western Europe, and Oceania have seen dramatic declines in breast cancer mortality and advanced-stage disease, other regions continue to be plagued by what really is becoming a survivable disease for most women. 

A new study in JAMA Oncology points out these disparities, documenting major differences in rates of advanced breast disease between countries in what researchers said was the most comprehensive review to date of global differences in breast cancer stage at diagnosis. 

  • Researchers conducted a meta-analysis of 133 studies covering 2.4M women across 81 nations over the past two decades, documenting differences in rates of advanced breast disease at diagnosis both over time and between countries. 

While most high-income nations have seen declines in rates of distant metastatic disease over the past 20 years, advanced-stage disease remains stubbornly common in lower middle-income countries. Researchers found: 

  • Rates of distant metastatic disease varied across countries by region, with sub-Saharan Africa the highest and North America the lowest (6-31% vs. 0-6%)
  • Lower socioeconomic status was tied to more advanced disease when women in the most disadvantaged group were compared to least disadvantaged (3-11% vs. 2-8%)
  • There were pronounced disparities even in high-resource countries with established screening programs, as rates of metastatic disease were twice as high in women of low socioeconomic status (SES) compared to high SES women, such as in the US (8% vs. 4%) 
  • Older women had a much higher prevalence of advanced disease across different countries compared to younger women (range of 4-34% vs. 2-16%), a phenomenon that could be because most screening programs stop at age 75
  • 40% of countries did not meet the Global Breast Cancer Initiative goal of having 60% or more of patients diagnosed at stage I or II

The Takeaway

The new findings indicate that it’s too soon to take a victory lap in the battle against breast cancer. While progress at higher socioeconomic levels in high-income countries has been impressive, breast cancer remains a scourge among more disadvantaged women and across wide regions of the world.

Get every issue of The Imaging Wire, delivered right to your inbox.

You might also like..

Select All

You're signed up!

It's great to have you as a reader. Check your inbox for a welcome email.

-- The Imaging Wire team

You're all set!