Discovery of breast cancers within 1 year of a normal screening mammogram: how are they found?

A Review of Applications of Machine Learning in Mammography and Future Challenges

BACKGROUND

The aim of this study is to systematically review the literature to summarize the evidence surrounding the clinical utility of artificial intelligence (AI) in the field of mammography. Databases from PubMed, IEEE Xplore, and Scopus were searched for relevant literature. Studies evaluating AI models in the context of prediction and diagnosis of breast malignancies that also reported conventional performance metrics were deemed suitable for inclusion. From 90 unique citations, 21 studies were considered suitable for our examination. Data was not pooled due to heterogeneity in study evaluation methods.

SUMMARY

Three studies showed the applicability of AI in reducing workload. Six studies demonstrated that AI can aid in diagnosis, with up to 69% reduction in false positives and an increase in sensitivity ranging from 84 to 91%. Five studies show how AI models can independently mark and classify suspicious findings on conventional scans, with abilities comparable with radiologists. Seven studies examined AI predictive potential for breast cancer and risk score calculation. Key Messages: Despite limitations in the current evidence base and technical obstacles, this review suggests AI has marked potential for extensive use in mammography. Additional works, including large-scale prospective studies, are warranted to elucidate the clinical utility of AI.

Differential Burden of Rare and Common Variants on Tumor Characteristics, Survival, and Mode of Detection in Breast Cancer

Genetic variants that increase breast cancer risk can be rare or common. This study tests whether the genetic risk stratification of breast cancer by rare and common variants in established loci can discriminate tumors with different biology, patient survival, and mode of detection. Multinomial logistic regression tested associations between genetic risk load [protein-truncating variant (PTV) carriership in 31 breast cancer predisposition genes-or polygenic risk score (PRS) using 162 single-nucleotide polymorphisms], tumor characteristics, and mode of detection (OR). Ten-year breast cancer-specific survival (HR) was estimated using Cox regression models. In this unselected cohort of 5,099 patients with breast cancer diagnosed in Sweden between 2001 and 2008, PTV carriers (n = 597) were younger and associated with more aggressive tumor phenotypes (ER-negative, large size, high grade, high proliferation, luminal B, and basal-like subtype) and worse outcome (HR, 1.65; 1.16-2.36) than noncarriers. After excluding 92 BRCA1/2 carriers, PTV carriership remained associated with high grade and worse survival (HR, 1.76; 1.21-2.56). In 5,007 BRCA1/2 noncarriers, higher PRS was associated with less aggressive tumor characteristics (ER-positive, PR-positive, small size, low grade, low proliferation, and luminal A subtype). Among patients with low mammographic density (<25%), non-BRCA1/2 PTV carriers were more often interval than screen-detected breast cancer (OR, 1.89; 1.12-3.21) than noncarriers. In contrast, higher PRS was associated with lower risk of interval compared with screen-detected cancer (OR, 0.77; 0.64-0.93) in women with low mammographic density. These findings suggest that rare and common breast cancer susceptibility loci are differentially associated with tumor characteristics, survival, and mode of detection.Significance: These findings offer the potential to improve screening practices for breast cancer by providing a deeper understanding of how risk variants affect disease progression and mode of detection. Cancer Res; 78(21); 6329-38. ©2018 AACR.

Could screening participation bias symptom interpretation? An interview study on women's interpretations of and responses to cancer symptoms between mammography screening rounds

OBJECTIVES

To explore how women with negative mammography screening results, but who were later diagnosed with interval breast cancer, reacted when they observed breast symptoms that could indicate malignancy in-between screening rounds.

DESIGN

Semistructured individual interviews with women who have been diagnosed with breast cancer during mammography screening intervals.

SETTING

Two breast diagnostic units covering two counties in Norway.

PARTICIPANTS

26 women diagnosed with interval breast cancer.

RESULTS

Women with a screening negative result react in two ways when experiencing a possible symptom of breast cancer. Among 24 women with a self-detected palpable lesion, 14 sought medical advice immediately. Their argument was to dispose of potential cancer as soon as possible. Ten women delayed seeking medical advice, explaining their delay as a result of practical difficulties such as holidays, uncertainty about the symptom, and previous experiences of healthcare services' ability to handle diffuse symptoms. Also, a recent negative mammography scan led some women to assume that the palpable lesion was benign and wait for the next screening round.

CONCLUSIONS

Participating in mammography screening may contribute to a postponed reaction to breast cancer symptoms, although most women acted rapidly when detecting a palpable breast lesion. Furthermore, screening participation does not necessarily increase awareness of breast cancer symptoms.

Clinical and self breast examination remain important in the era of modern screening

BACKGROUND

Breast cancer screening recommendations are in flux. We reviewed the methods of detecting newly diagnosed breast neoplasms at our institution.

METHODS

A retrospective review of patients stratified by age was performed to compare mammography with self- (SBE) and clinical (CBE) breast examination methods of cancer detection from 2005 to 2009.

RESULTS

We identified 782 patients. Patients aged <50 years were more likely to present with palpable disease (P < 0.001). Overall, 75% of patients had a mammogram within 24 months. There was a higher incidence of Tis tumors and lower incidence of T1 tumors if patients had mammography performed within 12 months versus 13-24 months (P < 0.01); tumor size, hormonal status, and lymph node (LN) status were comparable between these two groups. Patients diagnosed by SBE/CBE who had mammography performed within 12 months versus 13-24 months did not differ statistically according to tumor characteristics. In the screened cohort (mammography within 24 months), the majority of patients (64%) were diagnosed by mammography. Cancers detected by SBE/CBE were larger tumors (2.4 vs. 1.3 cm), higher grade, more frequently ER- (29 vs. 16%), triple-negative (21 vs. 10%), and lymph node-positive (39 vs. 18%; all P ≤ 0.01). There were no statistically significant differences in tumor size, T stage, or hormonal status in patients who had analog versus digital mammography.

CONCLUSIONS

Whereas the majority of patients had image-detected breast cancer, a significant number of image-screened patients presented with palpable disease, which were more aggressive cancers. Until imaging techniques are refined, SBE and CBE remain important for breast cancer diagnosis.

Incidence, detection, and tumour stage of breast cancer in a cohort of Italian women with negative screening mammography report recommending early (short-interval) rescreen

BACKGROUND

Although poorly described in the literature, the practice of early (short-interval) rescreen after a negative screening mammogram is controversial due to its financial and psychological burden and because it is of no proven benefit.

METHODS

The present study targeted an Italian 2-yearly screening programme (Emilia-Romagna Region, 1997-2002). An electronic dataset of 647,876 eligible negative mammography records from 376,257 women aged 50-69 years was record-linked with the regional breast cancer registry. The statistical analysis addressed the following research questions: (1) the prevalence of recommendation for early (<24 months) rescreen (RES) among negative mammography reports; (2) factors associated with the likelihood of a women receiving RES; and (3) whether women receiving RES and women receiving standard negative reports differed in terms of proportional incidence of interval breast cancer, recall rate at the next rescreen, detection rate of breast cancer at the next rescreen and the odds of having late-stage breast cancer during the interscreening interval and at the next rescreen.

RESULTS

RES was used in eight out of 13 screening centres, where it was found in 4171 out of 313,320 negative reports (average rate 1.33%; range 0.05%-4.33%). Reports with RES were more likely for women aged 50-59 years versus older women (odds ratio (OR) 1.33; 95% CI 1.25-1.42), for the first versus subsequent screening rounds (OR 1.91; 95% CI 1.79-2.04) and with a centre-specific recall rate below the average of 6.2% (OR 1.41; 95% CI 1.32-1.50). RES predicted a 3.51-fold (95% CI 0.94-9.29) greater proportional incidence of first-year interval cancers, a 1.90-fold (95% CI 1.62-2.22) greater recall rate at the next screen, a 1.72-fold (95% CI 1.01-2.74) greater detection rate of cancer at the next screen and a non-significantly decreased risk of late disease stage (OR 0.59; 95% CI 0.23-1.53).

CONCLUSION

The prevalence of RES was in line with the maximum standard level established by the Italian national guidelines. RES identified a subset of women with greater incidence of interval cancers and greater prevalence of cancers detected at the next screen.

Pathways to breast cancer diagnosis and treatment: exploring the social relations of diagnostic delay

In epidemiological studies, inequitable access to breast cancer care aligns with such variables as income, age, education, ethnicity and residential location. These variables correspond to structural patterns of advantage and disadvantage, which in turn may constrain or facilitate timely access to care. The purpose of this study was to understand the complexities of women's pathways to diagnosis. Thirty-five women from diverse backgrounds and who had clinically detectable breast symptoms at time of diagnosis participated in semi-structured interviews. All were receiving or completing treatment when they participated. Data were analyzed using the strategies of induction, comparison and abduction. All of the participants described a variety of activities involved in seeking care for breast cancer. The findings illustrate how the social relations of health care, rather than simply patient or provider delay, may contribute barriers to timely diagnosis. We illustrate how women's widely differing social and material contexts offer opportunities and barriers to access.