Comparison of AI-integrated pathways with human-AI interaction in population mammographic screening for breast cancer.
Ontology highlight
ABSTRACT:
SUBMITTER: Frazer HML
PROVIDER: S-EPMC11364867 | biostudies-literature | 2024 Aug
REPOSITORIES: biostudies-literature
Publications
Comparison of AI-integrated pathways with human-AI interaction in population mammographic screening for breast cancer.
Frazer Helen M L HML Peña-Solorzano Carlos A CA Kwok Chun Fung CF Elliott Michael S MS Chen Yuanhong Y Wang Chong C Lippey Jocelyn F JF Hopper John L JL Brotchie Peter P Carneiro Gustavo G McCarthy Davis J DJ
Nature communications 20240830 1
Artificial intelligence (AI) readers of mammograms compare favourably to individual radiologists in detecting breast cancer. However, AI readers cannot perform at the level of multi-reader systems used by screening programs in countries such as Australia, Sweden, and the UK. Therefore, implementation demands human-AI collaboration. Here, we use a large, high-quality retrospective mammography dataset from Victoria, Australia to conduct detailed simulations of five potential AI-integrated screenin ...[more]
Similar Datasets
Project description:BackgroundBecause of the mixed reports from smaller studies, we examined associations of race with mammographic breast density and evaluated racial differences in the determinants of breast density.MethodsParticipants included 37 839 women (23 166 non-Hispanic white and 14 673 African American) receiving screening mammograms at the Joanne Knight Breast Health Center at Washington University School of Medicine from June 2010 to December 2015. Mammographic breast density was assessed using the Breast Imaging Reporting and Data System (5th edition). To determine the association of race and participant characteristics with mammographic breast density, we used multivariable polytomous logistic regression models (reference group: almost entirely fatty).ResultsAfrican American women had increased odds of extremely dense (adjusted odds ratio = 1.31, 95% confidence interval = 1.13 to 1.52) and reduced odds of heterogeneously dense breasts (adjusted odds ratio = 0.91, 95% confidence interval = 0.84 to 0.99) compared with non-Hispanic white women. Altogether, race, parity and age at first birth, current age, current body mass index (BMI), BMI at age 18 years, menarche, family history of breast cancer, oral contraceptive use, alcohol use, and menopausal status explained 33% of the variation in mammographic breast density. Among African American and non-Hispanic white women, these factors explained nearly 28.6% and 33.6% of the variation in mammographic density, respectively. Current BMI provided the greatest explanation of breast density (26.2% overall, 22.2% in African American, and 26.2% in non-Hispanic white women).ConclusionsThe determinants of mammographic breast density were generally similar between African American women and non-Hispanic white women. After adjustments for confounders, African Americans had higher likelihood of extremely dense breasts but lower likelihood of heterogeneously dense breasts. The greatest explanation of breast density was provided by BMI, regardless of race.
| S-EPMC7192029 | biostudies-literature
Project description:ObjectivesThis study surveyed the views of breast screening readers in the UK on how to incorporate Artificial Intelligence (AI) technology into breast screening mammography.MethodsAn online questionnaire was circulated to the UK breast screening readers. Questions included their degree of approval of four AI implementation scenarios: AI as triage, AI as a companion reader/reader aid, AI replacing one of the initial two readers, and AI replacing all readers. They were also asked to rank five AI representation options (discrete opinion; mammographic scoring; percentage score with 100% indicating malignancy; region of suspicion; heat map) and indicate which evidence they considered necessary to support the implementation of AI into their practice among six options offered.ResultsThe survey had 87 nationally accredited respondents across the UK; 73 completed the survey in full. Respondents approved of AI replacing one of the initial two human readers and objected to AI replacing all human readers. Participants were divided on AI as triage and AI as a reader companion. A region of suspicion superimposed on the image was the preferred AI representation option. Most screen readers considered national guidelines (77%), studies using a nationally representative dataset (65%) and independent prospective studies (60%) as essential evidence. Participants' free-text comments highlighted concerns and the need for additional validation.ConclusionsOverall, screen readers supported the introduction of AI as a partial replacement of human readers and preferred a graphical indication of the suspected tumour area, with further evidence and national guidelines considered crucial prior to implementation.
| S-EPMC9733732 | biostudies-literature
Project description:Screening mammography can distort estimated effects in breast cancer risk models due to associations with other risk factors. Mammography information was available in the Nurses' Health Study from 1988, and 1,815 incident breast cancers were accrued through 2000 among 55,625 women with risk factor data. Logistic models were fit for screening mammography, and inverse probability weighting was used to adjust parameters in an established breast cancer risk model. Approximately 80% of women in each 2-year follow-up period had screening mammograms, which were positively associated with history of benign breast disease, family history of breast cancer, hormone therapy, alcohol use, physical activity, multivitamins, and calcium supplements, and negatively associated with postmenopause, current smoking, and body mass index. Markers of medical attention, including hypertension, high cholesterol, and osteoarthritis, were positively associated, while cardiovascular disease was negative. Inverse probability weighting led to small changes in effects of benign breast disease, family history, and hormone therapy. An apparent reduced risk associated with current smoking in unadjusted models was eliminated after weighting. Thus, several risk factors for breast cancer and cancer diagnosis are associated with mammographic screening. Adjustment for screening had some impact on breast cancer prediction in this cohort, especially for hormone therapy and smoking.
| S-EPMC2800262 | biostudies-literature
Project description:BackgroundHigh participation in mammographic screening is essential for its effectiveness to detect breast cancers early and thereby, improve breast cancer outcomes. Breast density is a strong predictor of breast cancer risk and significantly reduces the sensitivity of mammography to detect the disease. There are increasing mandates for routine breast density notification within mammographic screening programs. It is unknown if breast density notification impacts the likelihood of women returning to screening when next due (i.e. rescreening rates). This study investigates the association between breast density notification and rescreening rates using individual-level data from BreastScreen Western Australia (WA), a population-based mammographic screening program.MethodsWe examined 981,705 screening events from 311,656 women aged 40+ who attended BreastScreen WA between 2008 and 2017. Mixed effect logistic regression was used to investigate the association between rescreening and breast density notification status.ResultsResults were stratified by age (younger, targeted, older) and screening round (first, second, third+). Targeted women screening for the first time were more likely to return to screening if notified as having dense breasts (Percentunadjusted notified vs. not-notified: 57.8% vs. 56.1%; Padjusted = 0.016). Younger women were less likely to rescreen if notified, regardless of screening round (all P < 0.001). There was no association between notification and rescreening in older women (all P > 0.72).ConclusionsBreast density notification does not deter women in the targeted age range from rescreening but could potentially deter younger women from rescreening. These results suggest that all breast density notification messaging should include information regarding the importance of regular mammographic screening to manage breast cancer risk, particularly for younger women. These results will directly inform BreastScreen programs in Australia as well as other population-based screening providers outside Australia who notify women about breast density or are considering implementing breast density notification.
| S-EPMC8760641 | biostudies-literature
Project description:Overdiagnosis of breast cancer, i.e. the detection of slow-growing tumors that would never have caused symptoms or death, became more prevalent with the implementation of population-based screening. Only rough estimates have been made of the proportion of patients that are overdiagnosed and identification of those patients is difficult. Therefore, the aim of this study is to evaluate whether tumor biology can help identify patients with screen-detected tumors at such a low risk of recurrence that they are likely to be overdiagnosed. Furthermore, we wish to evaluate the impact of the transition from film-screen mammography (FSM) to the more sensitive full-field digital mammography (FFDM) on the biology of the tumors detected by each screening-modality. All Dutch breast cancer patients enrolled in the MINDACT trial (EORTC-10041) accrued 2007-2011, who participated in the national screening program (biennial screening ages 50-75) were included (n = 1,165). We calculated the proportions of high-, low- and among those the ultralow-risk tumors according to the 70-gene signature for patients with screen-detected (n = 775) and interval (n = 390) cancers for FSM and FFDM. Screen-detected cancers had significantly more often a low-risk tumor biology (68 %) of which 54 % even an ultralow-risk compared to interval cancers (53 % low-, of which 45 % ultralow-risk (p = 0.001) with an OR of 2.33 (p < 0.0001; 95 % CI 1.73-3.15). FFDM detected significantly more high-risk tumors (35 %) compared to FSM (27 %) (p = 0.011). Aside from favorable clinico-pathological factors, screen-detected cancers were also more likely to have a biologically low-risk or even ultralow-risk tumor. Especially for patients with screen-detected cancers the use of tools, such as the 70-gene signature, to differentiate breast cancers by risk of recurrence may minimize overtreatment. The recent transition in screening-modalities led to an increase in the detection of biologically high-risk cancers using FFDM.
| S-EPMC3924026 | biostudies-literature
Project description:PurposeTo evaluate the stand-alone diagnostic performances of AI-CAD and outcomes of AI-CAD detected abnormalities when applied to the mammographic interpretation workflow.MethodsFrom January 2016 to December 2017, 6499 screening mammograms of 5228 women were collected from a single screening facility. Historic reads of three radiologists were used as radiologist interpretation. A commercially-available AI-CAD was used for analysis. One radiologist not involved in interpretation had retrospectively reviewed the abnormality features and assessed the significance (negligible vs. need recall) of the AI-CAD marks. Ground truth in terms of cancer, benign or absence of abnormality was confirmed according to histopathologic diagnosis or negative results on the next-round screen.ResultsOf the 6499 mammograms, 6282 (96.7%) were in the negative, 189 (2.9%) were in the benign, and 28 (0.4%) were in the cancer group. AI-CAD detected 5 (17.9%, 5 of 28) of the 9 cancers that were intially interpreted as negative. Of the 648 AI-CAD recalls, 89.0% (577 of 648) were marks seen on examinations in the negative group, and 267 (41.2%) of the AI-CAD marks were considered to be negligible. Stand-alone AI-CAD has significantly higher recall rates (10.0% vs. 3.4%, P < 0.001) with comparable sensitivity and cancer detection rates (P = 0.086 and 0.102, respectively) when compared to the radiologists' interpretation.ConclusionAI-CAD detected 17.9% additional cancers on screening mammography that were initially overlooked by the radiologists. In spite of the additional cancer detection, AI-CAD had significantly higher recall rates in the clinical workflow, in which 89.0% of AI-CAD marks are on negative mammograms.
| S-EPMC10362167 | biostudies-literature
Project description:Artificial Intelligence is making rapid and remarkable progress in the development of more sophisticated and powerful systems. However, the acknowledgement of several problems with modern machine learning approaches has prompted a shift in AI benchmarking away from task-oriented testing (such as Chess and Go) towards ability-oriented testing, in which AI systems are tested on their capacity to solve certain kinds of novel problems. The Animal-AI Environment is one such benchmark which aims to apply the ability-oriented testing used in comparative psychology to AI systems. Here, we present the first direct human-AI comparison in the Animal-AI Environment, using children aged 6-10 (n = 52). We found that children of all ages were significantly better than a sample of 30 AIs across most of the tests we examined, as well as performing significantly better than the two top-scoring AIs, "ironbar" and "Trrrrr," from the Animal-AI Olympics Competition 2019. While children and AIs performed similarly on basic navigational tasks, AIs performed significantly worse in more complex cognitive tests, including detour tasks, spatial elimination tasks, and object permanence tasks, indicating that AIs lack several cognitive abilities that children aged 6-10 possess. Both children and AIs performed poorly on tool-use tasks, suggesting that these tests are challenging for both biological and non-biological machines.
| S-EPMC9172850 | biostudies-literature
Project description:BackgroundArtificial intelligence (AI) systems are proposed as a replacement of the first reader in double reading within mammography screening. We aimed to assess cancer detection accuracy of an AI system in a Danish screening population.MethodsWe retrieved a consecutive screening cohort from the Region of Southern Denmark including all participating women between Aug 4, 2014, and August 15, 2018. Screening mammograms were processed by a commercial AI system and detection accuracy was evaluated in two scenarios, Standalone AI and AI-integrated screening replacing first reader, with first reader and double reading with arbitration (combined reading) as comparators, respectively. Two AI-score cut-off points were applied by matching at mean first reader sensitivity (AIsens) and specificity (AIspec). Reference standard was histopathology-proven breast cancer or cancer-free follow-up within 24 months. Coprimary endpoints were sensitivity and specificity, and secondary endpoints were positive predictive value (PPV), negative predictive value (NPV), recall rate, and arbitration rate. Accuracy estimates were calculated using McNemar's test or exact binomial test.ResultsOut of 272,008 screening mammograms from 158,732 women, 257,671 (94.7%) with adequate image data were included in the final analyses. Sensitivity and specificity were 63.7% (95% CI 61.6%-65.8%) and 97.8% (97.7-97.8%) for first reader, and 73.9% (72.0-75.8%) and 97.9% (97.9-98.0%) for combined reading, respectively. Standalone AIsens showed a lower specificity (-1.3%) and PPV (-6.1%), and a higher recall rate (+ 1.3%) compared to first reader (p < 0.0001 for all), while Standalone AIspec had a lower sensitivity (-5.1%; p < 0.0001), PPV (-1.3%; p = 0.01) and NPV (-0.04%; p = 0.0002). Compared to combined reading, Integrated AIsens achieved higher sensitivity (+ 2.3%; p = 0.0004), but lower specificity (-0.6%) and PPV (-3.9%) as well as higher recall rate (+ 0.6%) and arbitration rate (+ 2.2%; p < 0.0001 for all). Integrated AIspec showed no significant difference in any outcome measures apart from a slightly higher arbitration rate (p < 0.0001). Subgroup analyses showed higher detection of interval cancers by Standalone AI and Integrated AI at both thresholds (p < 0.0001 for all) with a varying composition of detected cancers across multiple subgroups of tumour characteristics.ConclusionsReplacing first reader in double reading with an AI could be feasible but choosing an appropriate AI threshold is crucial to maintaining cancer detection accuracy and workload.
| S-EPMC10731688 | biostudies-literature
Project description:BackgroundArtificial intelligence (AI) has been proposed to reduce false-positive screens, increase cancer detection rates (CDRs), and address resourcing challenges faced by breast screening programs. We compared the accuracy of AI versus radiologists in real-world population breast cancer screening, and estimated potential impacts on CDR, recall and workload for simulated AI-radiologist reading.MethodsExternal validation of a commercially-available AI algorithm in a retrospective cohort of 108,970 consecutive mammograms from a population-based screening program, with ascertained outcomes (including interval cancers by registry linkage). Area under the ROC curve (AUC), sensitivity and specificity for AI were compared with radiologists who interpreted the screens in practice. CDR and recall were estimated for simulated AI-radiologist reading (with arbitration) and compared with program metrics.FindingsThe AUC for AI was 0.83 compared with 0.93 for radiologists. At a prospective threshold, sensitivity for AI (0.67; 95% CI: 0.64-0.70) was comparable to radiologists (0.68; 95% CI: 0.66-0.71) with lower specificity (0.81 [95% CI: 0.81-0.81] versus 0.97 [95% CI: 0.97-0.97]). Recall rate for AI-radiologist reading (3.14%) was significantly lower than for the BSWA program (3.38%) (-0.25%; 95% CI: -0.31 to -0.18; P < 0.001). CDR was also lower (6.37 versus 6.97 per 1000) (-0.61; 95% CI: -0.77 to -0.44; P < 0.001); however, AI detected interval cancers that were not found by radiologists (0.72 per 1000; 95% CI: 0.57-0.90). AI-radiologist reading increased arbitration but decreased overall screen-reading volume by 41.4% (95% CI: 41.2-41.6).InterpretationReplacement of one radiologist by AI (with arbitration) resulted in lower recall and overall screen-reading volume. There was a small reduction in CDR for AI-radiologist reading. AI detected interval cases that were not identified by radiologists, suggesting potentially higher CDR if radiologists were unblinded to AI findings. These results indicate AI's potential role as a screen-reader of mammograms, but prospective trials are required to determine whether CDR could improve if AI detection was actioned in double-reading with arbitration.FundingNational Breast Cancer Foundation (NBCF), National Health and Medical Research Council (NHMRC).
| S-EPMC9996220 | biostudies-literature
Project description:Attention in the 2000s on the importance of mammographic density led us to study screening sensitivity, breast cancer incidence, and associations with risk factors by mammographic density in Danish breast cancer screening programs. Here, we summarise our approaches and findings. Dichotomized density codes: fatty, equal to BI-RADS density code 1 and part of 2, and other mixed/dense data from the 1990s-were available from two counties, and BI-RADS density codes from one region were available from 2012/13. Density data were linked with data on vital status, incident breast cancer, and potential risk factors. We calculated screening sensitivity by combining data on screen-detected and interval cancers. We used cohorts to study high density as a predictor of breast cancer risk; cross-sectional data to study the association between life style factors and density, adjusting for age and body mass index (BMI); and time trends to study the prevalence of high density across birth cohorts. Sensitivity decreased with increasing density from 78% in women with BI-RADS 1 to 47% in those with BI-RADS 4. For women with mixed/dense compared with those with fatty breasts, the rate ratio of incident breast cancer was 2.45 (95% CI 2.14-2.81). The percentage of women with mixed/dense breasts decreased with age, but at a higher rate the later the women were born. Among users of postmenopausal hormone therapy, the percentage of women with mixed/dense breasts was higher than in non-users, but the patterns across birth cohorts were similar. The occurrence of mixed/dense breast at screening age decreased by a z-score unit of BMI at age 13-odds ratio (OR) 0.56 (95% CI 0.53-0.58)-and so did breast cancer risk and hazard ratio (HR) 0.92 (95% CI 0.84-1.00), but it changed to HR 1.01 (95% CI 0.93-1.11) when controlled for density. Age and BMI adjusted associations between life style factors and density were largely close to unity; physical activity OR 1.06 (95% CI 0.93-1.21); alcohol consumption OR 1.01 (95% CI 0.81-1.27); air pollution OR 0.96 (95% 0.93-1.01) per 20 μg/m3; and traffic noise OR 0.94 (95% CI 0.86-1.03) per 10 dB. Weak negative associations were seen for diabetes OR 0.61 (95% CI 0.40-0.92) and cigarette smoking OR 0.86 (95% CI 0.75-0.99), and a positive association was found with hormone therapy OR 1.24 (95% 1.14-1.35). Our data indicate that breast tissue in middle-aged women is highly dependent on childhood body constitution while adult life-style plays a modest role, underlying the need for a long-term perspective in primary prevention of breast cancer.
| S-EPMC6912479 | biostudies-literature