Mammographic risk scoring has commonly been automated by extracting a set of handcrafted features from mammograms, and relating the responses directly or indirectly to breast cancer risk. We present a method that learns a feature hierarchy from unlabeled data. When the learned features are used as the input to a simple classifier, two different tasks can be addressed: i) breast density segmentation, and ii) scoring of mammographic texture. The proposed model learns features at multiple scales. To control the models capacity a novel sparsity regularizer is introduced that incorporates both lifetime and population sparsity. We evaluated our method on three different clinical datasets. Our state-of-the-art results show that the learned breast density scores have a very strong positive relationship with manual ones, and that the learned texture scores are predictive of breast cancer. The model is easy to apply and generalizes to many other segmentation and scoring problems.

Abstract. Clinical studies including thousands of magnetic resonance imaging (MRI) scans offer potential for pathogenesis research in osteoarthritis. However, comprehensive quantification of all bone, cartilage, and meniscus compartments is challenging. We propose a segmentation framework for fully automatic segmentation of knee MRI. The framework combines multiatlas rigid registration with voxel classification and was trained on manual segmentations with varying configurations of bones, cartilages, and menisci. The validation included high-and low-field knee MRI cohorts from the Center for Clinical and Basic Research, the osteoarthritis initiative (QAI), and the segmentation of knee images10 (SKI10) challenge. In total, 1907 knee MRIs were segmented during the evaluation. No segmentations were excluded. Our resulting OAI cartilage volume scores are available upon request. The precision and accuracy performances matched manual reader re-segmentation well. The cartilage volume scan-rescan precision was 4.9% (RMS CV). The Dice volume overlaps in the medial/lateral tibial/femoral cartilage compartments were 0.80 to 0.87. The correlations with volumes from independent methods were between 0.90 and 0.96 on the OAI scans. Thus, the framework demonstrated precision and accuracy comparable to manual segmentations. Finally, our method placed second for cartilage segmentation in the SKI10 challenge. The comprehensive validation suggested that automatic segmentation is appropriate for cohorts with thousands of scans.

This paper presents a brain T1-weighted structural magnetic resonance imaging (MRI) biomarker that combines several individual MRI biomarkers (cortical thickness measurements, volumetric measurements, hippocampal shape, and hippocampal texture). The method was developed, trained, and evaluated using two publicly available reference datasets: a standardized dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the imaging arm of the Australian Imaging Biomarkers and Lifestyle flagship study of ageing (AIBL). In addition, the method was evaluated by participation in the Computer-Aided Diagnosis of Dementia (CADDementia) challenge. Cross-validation using ADNI and AIBL data resulted in a multi-class classification accuracy of 62.7% for the discrimination of healthy normal controls (NC), subjects with mild cognitive impairment (MCI), and patients with Alzheimer's disease (AD). This performance generalized to the CADDementia challenge where the method, trained using the ADNI and AIBL data, achieved a classification accuracy 63.0%. The obtained classification accuracy resulted in a first place in the challenge, and the method was significantly better (McNemar's test) than the bottom 24 methods out of the total of 29 methods contributed by 15 different teams in the challenge. The method was further investigated with learning curve and feature selection experiments using ADNI and AIBL data. The learning curve experiments suggested that neither more training data nor a more complex classifier would have improved the obtained results. The feature selection experiment showed that both common and uncommon individual MRI biomarkers contributed to the performance; hippocampal volume, ventricular volume, hippocampal texture, and parietal lobe thickness were the most important. This study highlights the need for both subtle, localized measurements and global measurements in order to discriminate NC, MCI, and AD simultaneously based on a single structural MRI scan. It is likely that additional non-structural MRI features are needed to further improve the obtained performance, especially to improve the discrimination between NC and MCI.

Patients with severe COVID-19 have overwhelmed healthcare systems worldwide. We hypothesized that machine learning (ML) models could be used to predict risks at different stages of management and thereby provide insights into drivers and prognostic markers of disease progression and death. From a cohort of approx. 2.6 million citizens in Denmark, SARS-CoV-2 PCR tests were performed on subjects suspected for COVID-19 disease; 3944 cases had at least one positive test and were subjected to further analysis. SARS-CoV-2 positive cases from the United Kingdom Biobank was used for external validation. The ML models predicted the risk of death (Receiver Operation Characteristics—Area Under the Curve, ROC-AUC) of 0.906 at diagnosis, 0.818, at hospital admission and 0.721 at Intensive Care Unit (ICU) admission. Similar metrics were achieved for predicted risks of hospital and ICU admission and use of mechanical ventilation. Common risk factors, included age, body mass index and hypertension, although the top risk features shifted towards markers of shock and organ dysfunction in ICU patients. The external validation indicated fair predictive performance for mortality prediction, but suboptimal performance for predicting ICU admission. ML may be used to identify drivers of progression to more severe disease and for prognostication patients in patients with COVID-19. We provide access to an online risk calculator based on these findings.

Background: Developments in artificial intelligence (AI) systems to assist radiologists in reading mammograms could improve breast cancer screening efficiency.Purpose: To investigate whether an AI system could detect normal, moderate-risk, and suspicious mammograms in a screening sample to safely reduce radiologist workload and evaluate across Breast Imaging Reporting and Data System (BI-RADS) densities. Materials and Methods:This retrospective simulation study analyzed mammographic examination data consecutively collected from January 2014 to December 2015 in the Danish Capital Region breast cancer screening program. All mammograms were scored from 0 to 10, representing the risk of malignancy, using an AI tool. During simulation, normal mammograms (score , 5) would be excluded from radiologist reading and suspicious mammograms (score . recall threshold [RT]) would be recalled. Two radiologists read the remaining mammograms. The RT was fitted using another independent cohort (same institution) by matching to the radiologist sensitivity. This protocol was further applied to each BI-RADS density. Screening outcomes were measured using the sensitivity, specificity, workload, and false-positive rate. The AI-based screening was tested for noninferiority sensitivity compared with radiologist screening using the Farrington-Manning test. Specificities were compared using the McNemar test. Results:The study sample comprised 114 421 screenings for breast cancer in 114 421 women, resulting in 791 screen-detected, 327 interval, and 1473 long-term cancers and 2107 false-positive screenings. The mean age of the women was 59 years 6 6 (SD). The AI-based screening sensitivity was 69.7% (779 of 1118; 95% CI: 66.9, 72.4) and was noninferior (P = .02) to the radiologist screening sensitivity of 70.8% (791 of 1118; 95% CI: 68.0, 73.5). The AI-based screening specificity was 98.6% (111 725 of 113 303; 95% CI: 98.5, 98.7), which was higher (P , .001) than the radiologist specificity of 98.1% (111 196 of 113 303; 95% CI: 98.1, 98.2). The radiologist workload was reduced by 62.6% (71 585 of 114 421), and 25.1% (529 of 2107) of false-positive screenings were avoided. Screening results were consistent across BI-RADS densities, although not significantly so for sensitivity. Conclusion:Artificial intelligence (AI)-based screening could detect normal, moderate-risk, and suspicious mammograms in a breast cancer screening program, which may reduce the radiologist workload. AI-based screening performed consistently across breast densities.

No abstract

Background Screening mammography works better in fatty than in dense breast tissue. Computerized assessment of parenchymal texture is a non-subjective method to obtain a refined description of breast tissue, potentially valuable in addition to breast density scoring for the identification of women in need of supplementary imaging. We studied the sensitivity of screening mammography by a combination of radiologist-assessed Breast Imaging Reporting and Data System (BI-RADS) density score and computer-assessed parenchymal texture marker, mammography texture resemblance (MTR), in a population-based screening program. Methods Breast density was coded according to the fourth edition of the BI-RADS density code, and MTR marker was divided into quartiles from 1 to 4. Screening data were followed up for the identification of screen-detected and interval cancers. We calculated sensitivity and specificity with 95% confidence intervals (CI) by BI-RADS density score, MTR marker, and combination hereof. Results Density and texture were strongly correlated, but the combination led to the identification of subgroups with different sensitivity. Sensitivity was high, about 80%, in women with BI-RADS density score 1 and MTR markers 1 or 2. Sensitivity was low, 67%, in women with BI-RADS density score 2 and MTR marker 4. For women with BI-RADS density scores 3 and 4, the already low sensitivity was further decreased for women with MTR marker 4. Specificity was 97–99% in all subgroups. Conclusion Our study showed that women with low density constituted a heterogenous group. Classifying women for extra imaging based on density only might be a too crude approach. Screening sensitivity was systematically high in women with fatty and homogenous breast tissue.

PurposeThe currently recommended double reading of all screening mammography examinations is an economic burden for screening programs. The sensitivity of screening is higher for women with low breast density than for women with high density. One may therefore ask whether single reading could replace double reading at least for women with low density. We addressed this question using data from a screening program where the radiologists coded their readings independently.MethodsData include all screening mammography examinations in the Capital Region of Denmark from 1 November 2012 to 31 December 2013. Outcome of screening was assessed by linkage to the Danish Pathology Register. We calculated sensitivity, specificity, number of interval cancers, and false positive-tests per 1000 screened women by both single reader and consensus BI-RADS density code.ResultsIn total 54,808 women were included. The overall sensitivity of double reading was 72%, specificity was 97.6%, 3 women per 1000 screened experienced an interval cancer, and 24 a false-positive test. Across all BI-RADS density codes, single reading consistently decreased sensitivity as compared with consensus reading. The same was true for specificity, apart from results across BI-RADS density codes set by reader 2.ConclusionsSingle reading decreased sensitivity as compared with double reading across all BI-RADS density codes. This included results based on consensus BI-RADS density codes. This means that replacement of double with single reading would have negative consequences for the screened women, even if density could be assessed automatically calibrated to the usual consensus level.