Deep Learning Models for Automated Assessment of Breast Density Using Multiple Mammographic Image Types.

Recently, convolutional neural network (CNN) models have been proposed to automate the assessment of breast density, breast cancer detection or risk stratification using single image modality. However, analysis of breast density using multiple mammographic types using clinical data has not been reported in the literature. In this study, we investigate pre-trained EfficientNetB0 deep learning (DL) models for automated assessment of breast density using multiple mammographic types with and without clinical information to improve reliability and versatility of reporting. 120,000 for-processing and for-presentation full-field digital mammograms (FFDM), digital breast tomosynthesis (DBT), and synthesized 2D images from 5032 women were retrospectively analyzed. Each participant underwent up to 3 screening examinations and completed a questionnaire at each screening encounter. Pre-trained EfficientNetB0 DL models with or without clinical history were optimized. The DL models were evaluated using BI-RADS (fatty, scattered fibroglandular densities, heterogeneously dense, or extremely dense) versus binary (non-dense or dense) density classification. Pre-trained EfficientNetB0 model performances were compared using inter-observer and commercial software (Volpara) variabilities. Results show that the average Fleiss' Kappa score between-observers ranged from 0.31-0.50 and 0.55-0.69 for the BI-RADS and binary classifications, respectively, showing higher uncertainty among experts. Volpara-observer agreement was 0.33 and 0.54 for BI-RADS and binary classifications, respectively, showing fair to moderate agreement. However, our proposed pre-trained EfficientNetB0 DL models-observer agreement was 0.61-0.66 and 0.70-0.75 for BI-RADS and binary classifications, respectively, showing moderate to substantial agreement. Overall results show that the best breast density estimation was achieved using for-presentation FFDM and DBT images without added clinical information. Pre-trained EfficientNetB0 model can automatically assess breast density from any images modality type, with the best results obtained from for-presentation FFDM and DBT, which are the most common image archived in clinical practice.

Blood-Based Biomarker Panel for Personalized Lung Cancer Risk Assessment.

PURPOSE: To investigate whether a panel of circulating protein biomarkers would improve risk assessment for lung cancer screening in combination with a risk model on the basis of participant characteristics. METHODS: A blinded validation study was performed using prostate lung colorectal ovarian (PLCO) Cancer Screening Trial data and biospecimens to evaluate the performance of a four-marker protein panel (4MP) consisting of the precursor form of surfactant protein B, cancer antigen 125, carcinoembryonic antigen, and cytokeratin-19 fragment in combination with a lung cancer risk prediction model (PLCOm2012) compared with current US Preventive Services Task Force (USPSTF) screening criteria. The 4MP was assayed in 1,299 sera collected preceding lung cancer diagnosis and 8,709 noncase sera. RESULTS: The 4MP alone yielded an area under the receiver operating characteristic curve of 0.79 (95% CI, 0.77 to 0.82) for case sera collected within 1-year preceding diagnosis and 0.74 (95% CI, 0.72 to 0.76) among the entire specimen set. The combined 4MP + PLCOm2012 model yielded an area under the receiver operating characteristic curve of 0.85 (95% CI, 0.82 to 0.88) for case sera collected within 1 year preceding diagnosis. The benefit of the 4MP in the combined model resulted from improvement in sensitivity at high specificity. Compared with the USPSTF2021 criteria, the combined 4MP + PLCOm2012 model exhibited statistically significant improvements in sensitivity and specificity. Among PLCO participants with ≥ 10 smoking pack-years, the 4MP + PLCOm2012 model would have identified for annual screening 9.2% more lung cancer cases and would have reduced referral by 13.7% among noncases compared with USPSTF2021 criteria. CONCLUSION: A blood-based biomarker panel in combination with PLCOm2012 significantly improves lung cancer risk assessment for lung cancer screening.

Assessment of Lung Cancer Risk on the Basis of a Biomarker Panel of Circulating Proteins.

Importance: There is an urgent need to improve lung cancer risk assessment because current screening criteria miss a large proportion of cases. Objective: To investigate whether a lung cancer risk prediction model based on a panel of selected circulating protein biomarkers can outperform a traditional risk prediction model and current US screening criteria. Design, Setting, and Participants: Prediagnostic samples from 108 ever-smoking patients with lung cancer diagnosed within 1 year after blood collection and samples from 216 smoking-matched controls from the Carotene and Retinol Efficacy Trial (CARET) cohort were used to develop a biomarker risk score based on 4 proteins (cancer antigen 125 [CA125], carcinoembryonic antigen [CEA], cytokeratin-19 fragment [CYFRA 21-1], and the precursor form of surfactant protein B [Pro-SFTPB]). The biomarker score was subsequently validated blindly using absolute risk estimates among 63 ever-smoking patients with lung cancer diagnosed within 1 year after blood collection and 90 matched controls from 2 large European population-based cohorts, the European Prospective Investigation into Cancer and Nutrition (EPIC) and the Northern Sweden Health and Disease Study (NSHDS). Main Outcomes and Measures: Model validity in discriminating between future lung cancer cases and controls. Discrimination estimates were weighted to reflect the background populations of EPIC and NSHDS validation studies (area under the receiver-operating characteristics curve [AUC], sensitivity, and specificity). Results: In the validation study of 63 ever-smoking patients with lung cancer and 90 matched controls (mean [SD] age, 57.7 [8.7] years; 68.6% men) from EPIC and NSHDS, an integrated risk prediction model that combined smoking exposure with the biomarker score yielded an AUC of 0.83 (95% CI, 0.76-0.90) compared with 0.73 (95% CI, 0.64-0.82) for a model based on smoking exposure alone (P = .003 for difference in AUC). At an overall specificity of 0.83, based on the US Preventive Services Task Force screening criteria, the sensitivity of the integrated risk prediction (biomarker) model was 0.63 compared with 0.43 for the smoking model. Conversely, at an overall sensitivity of 0.42, based on the US Preventive Services Task Force screening criteria, the integrated risk prediction model yielded a specificity of 0.95 compared with 0.86 for the smoking model. Conclusions and Relevance: This study provided a proof of principle in showing that a panel of circulating protein biomarkers may improve lung cancer risk assessment and may be used to define eligibility for computed tomography screening.

Detection and localization of surgically resectable cancers with a multi-analyte blood test.

Earlier detection is key to reducing cancer deaths. Here, we describe a blood test that can detect eight common cancer types through assessment of the levels of circulating proteins and mutations in cell-free DNA. We applied this test, called CancerSEEK, to 1005 patients with nonmetastatic, clinically detected cancers of the ovary, liver, stomach, pancreas, esophagus, colorectum, lung, or breast. CancerSEEK tests were positive in a median of 70% of the eight cancer types. The sensitivities ranged from 69 to 98% for the detection of five cancer types (ovary, liver, stomach, pancreas, and esophagus) for which there are no screening tests available for average-risk individuals. The specificity of CancerSEEK was greater than 99%: only 7 of 812 healthy controls scored positive. In addition, CancerSEEK localized the cancer to a small number of anatomic sites in a median of 83% of the patients.