The use of AI in objective density assessment, risk assessment, and quality evaluation were all topics involving software from Volpara Health highlighted in research presented at the 2023 European Congress of Radiology (ECR).
Highlights from six oral presentations and posters include:
- Pairing the Transpara™ AI CAD with Volpara’s volumetric density assessment improved screening efficacy for women with dense breast tissue compared to a traditional double read by radiologists
- Using Tyrer-Cuzick with volumetric density assessment from Volpara® Scorecard™ to identify and triage intermediate and high-risk patients for ultrasound resulted in a higher cancer detection rate than traditional biennial mammography
- Evaluating radiographer performance with Volpara® Analytics™ software helped improve quality regardless of experience level
Volpara’s robust algorithms have firmly established the company as a trusted research partner across many high-profile studies, which have already generated a number of publications. Such research includes personalized screening and optimization studies, such as the DENSE, TOMMY, and To-Be trials, and the PROCAS I/II and KARMA studies. As a prime example, the DENSE trial, led by Professor Carla van Gils from the University Medical Center Utrecht, Netherlands, relied on Volpara’s TruDensity® product to identify women with extremely dense breasts in their randomized controlled trial of supplemental MRI.
New research presented at ECR demonstrates the important role Volpara software plays in optimizing breast density, risk assessment and mammography quality.
“Increasing our understanding of breast cancer risk, detection and prevention is essential to ensuring every patient is given the right care at the right time,” said Teri Thomas, Volpara CEO. “The evidence presented at ECR comes at a moment when European screening programs are realizing AI gives them an opportunity to help more women with more personalized care, despite staffing challenges. I’m eager to see this research put into practice.”
AI for Density and Risk Assessment Leading to Triage to Supplemental Care
Artificial intelligence (AI) and mammographic extremely dense breasts in BreastScreen Norway: could AI-based screening be an alternative to screening with MRI?
Henrik Wethe Koch, et al.
Motivated by EUSOBI’s recommendations for the use of MRI to screen women with extremely dense breasts, this retrospective study investigated the use of AI as an alternative to improve screening efficacy. Using the AI system Transpara® by ScreenPoint Medical and Volpara’s automated density calculation Volpara®Density™, malignancy scores were assigned to women with extremely dense breast tissue, of which 67 had screen-detected cancers, 38 had interval cancers, and 2403 had negative screening examinations. Compared to a 63.8% sensitivity for independent double reading of mammograms by radiologists, Transpara had a sensitivity of 81.9% in detecting malignancies. These results indicate the potential of AI systems to benefit women with dense breasts.
Risk-based breast screening (RIBBS) in young women: stratification of population and cancer detection rate (CDR) from recruitment
Francesca Caumo, et al.
This study aimed to stratify patients for personalized breast screening of women aged 45 to 49 based on 2-view digital breast tomosynthesis (DBT) analyzed using Volpara® Scorecard™ software to measure volumetric breast density (VBD). Women with VBD>25% received supplemental ultrasound screening. The Tyrer-Cuzick risk model was used with VBD as an input to calculate lifetime risk. Of 10,270 study participants, 57.9% were deemed low-risk, 33.6% were at intermediate risk, and 8.5% were high-risk. Low-risk patients were assigned to biennial screening, while intermediate- and high-risk patients were assigned to annual screening. The cancer detection rate (CDR) was compared between the groups and a significant difference was found according to risk. For the low-risk patients, the CDR was 3.0/1000, while it was 9.2/1000 for the intermediate- and high-risk patients combined. This confirms a higher CDR in women identified according to Tyrer-Cuzick with Volpara®Density™, suggesting efficacy of personalized risk-based screening programs.
Automated breast ultrasound in comparison to 2D mammography, digital breast tomosynthesis, hand-held ultrasound in the detection of breast cancer: a cohort of 4500 examinations
Athina Vourtsi, et al.
This prospective study compared automated breast ultrasound (ABUS), digital breast tomosynthesis (DBT), and hand-held breast ultrasound (HHUS) as supplemental screening tools for women with dense breasts as assessed as Volpara® Density Grade™ (VDG®) “c” or “d” using Volpara® Scorecard™. Of the 4500 examinations included in the study, 88 cancers were identified—74 using mammography, 75 using ABUS, s and 78 using DBT. Of the 75 cancers detected using ABUS, eight were invasive carcinomas not seen on DBT. Furthermore, a high level of agreement was observed between ABUS and HHUS, indicating the success of both screening modalities in detecting invasive breast cancers in patients with dense breasts.
Dr. Athina Vourtsi of Athena Medical reported on positive results of supplemental screening with ultrasound for women identified with dense breasts by Volpara® Scorecard™ software. “Quantitative assessment of breast composition using Volpara Scorecard is a great tool to give a reliable density score, especially in breast tomosynthesis where the synthetic view may not display density in the same way as what is seen by scrolling through the slices,” said Dr. Vourtsi.
As Europe moves forward with new EUSOBI guidelines requiring density reporting to all women receiving mammography screening and recommending MRI for women aged 50 to 70 with extremely dense breasts, the application of AI in clinical practice is a growing trend. Workforce shortages across Europe are also pushing AI further into the spotlight.
Contrast-enhanced mammography (CEM) as an alternative to breast MRI for screening of women at increased risk for breast cancer: preliminary results
Gisella Gennaro, et al.
With the aim of comparing personalized screening performance of contrast-enhanced mammography (CEM) and breast MRI, this preliminary retrospective analysis looked at 106 patients at intermediate or high risk for breast cancer who each underwent both CEM and MRI. Breast cancer risk assessment was completed using the Tyrer-Cuzick risk model with volumetric breast density input from Volpara® Scorecard™. Studies were reviewed by six independent radiologists and 51 lesions were diagnosed by either CEM, MRI, or both—22 were malignant and 29 were benign. Performance differences between CEM and MRI were not statistically significant, though CEM AUC was significantly greater than for low-energy CEM alone (0.954 vs 0.796). The study concludes that CEM holds potential as an alternative screening modality for patients at increased risk of breast cancer.
AI for Quality Evaluation and Optimization of Positioning & Compression
Breast positioning indicators associated with mammograms repeated due to blur
Melissa Hill, et al.
This study investigated breast positioning associated with mammography views repeated due to image blur in comparison to the accepted technical image repeat views. Using Volpara software, compression and positioning quality indicators were derived for mammography studies with one or more views repeated due to blur. In 633 included studies, significant differences in compression force, pressure, contact area, breast thicknesses, exposure time, and posterior nipple length (PNL) were observed between blurred and accepted views. In MLO views concave pectoral shape was associated with a reduced odds (OR 0.77) of blur. Factors associated with reduced odds of blur in both CC and MLO views included target or high compression pressures relative to low (OR 0.1 to 0.36). Increased odds of blur were associated with Moderate PGMI scores relative to Perfect PGMI scores in CC images (OR 2.67), and obscured vs visible IMF in MLO images (OR 1.46). Results indicate that there are several modifiable positioning factors associated with image blur.
Breast positioning and compression in screening with tomosynthesis: use of automatic software to improve the performance of breast radiographers
Gisella Gennaro, et al.
This study used Volpara software to evaluate and monitor breast positioning and compression performance of five breast radiographers with experience ranging from six months to over 25 years. Baseline performance was established over a 4 month period and then shared with the radiographer team. The individual most frequent positioning and compression errors were assigned to each radiographer as improvement goals, with an assignment to actively use the automated software to achieve their goals. Individual progress was shared with radiographers monthly over a 6 month period and resulted in significant improvements in positioning as assessed using outputs of Volpara’s TruPGMI™ clinical function. The proportion of mammograms scored as Perfect and Good increased from between 9.2% and 16.7% relative to the baseline period. Results of this study showed that the use of Volpara® Analytics™ software can help radiographers improve their performance, regardless of their experience level.