Search

Background: Night-shift work (NSW) has been suggested as a possible cause of breast cancer, and its association with mammographic density (MD), one of the strongest risk factors for breast cancer, has been scarcely addressed. This study examined NSW and MD in Spanish women.Methods: The study covered 2,752 women aged 45–68 years recruited in 2007–2008 in 7 population-based public breast cancer screening centers, which included 243 women who had performed NSW for at least one year. Occupational data and information on potential confounders were collected by personal interview. Two trained radiologist estimated the percentage of MD assisted by a validated semiautomatic computer tool (DM-scan). Multivariable mixed linear regression models with random screening center–specific intercepts were fitted using log-transformed percentage of MD as the dependent variable and adjusting by known confounding variables.Results: Having ever worked in NSW was not associated with MD [{e^\beta }$:0.96; 95% confidence interval (CI), 0.86–1.06]. However, the adjusted geometric mean of the percentage of MD in women with NSW for more than 15 years was 25% higher than that of those without NSW history (MD>15 years:20.7% vs. MDnever:16.5%;{e^\beta }$:1.25; 95% CI,1.01–1.54). This association was mainly observed in postmenopausal participants ({e^\beta }$:1.28; 95% CI, 1.00–1.64). Among NSW-exposed women, those with ≤2 night-shifts per week had higher MD than those with 5 to 7 nightshifts per week ({e^\beta }$:1.42; 95% CI, 1.10–1.84).Conclusions: Performing NSW was associated with higher MD only in women with more than 15 years of cumulated exposure. These findings warrant replication in futures studies.Impact: Our findings suggest that MD could play a role in the pathway between long-term NSW and breast cancer. Cancer Epidemiol Biomarkers Prev; 26(6); 905–13. ©2017 AACR.

Supplementary table S1. Association of current or recent (stopped NSW

Supplementary Figure 1. Graphic esqueme summarizing the proposed pathways of the association between night-shift work and mammographic density.

This paper describes an ensemble feature identification algorithm called SEQENS, and measures its capability to identify the relevant variables in a case-control study using a genetic expression microarray dataset. SEQENS uses Sequential Feature Search on multiple sample splitting to select variables showing stronger relation with the target, and a variable relevance ranking is finally produced. Although designed for feature identification, SEQENS could also serve as a basis for feature selection (classifier optimisation). Cliff, a ranking evaluation metric is also presented and used to assess the feature identification algorithms when a groundtruth of relevant variables is available. To test performance, three types of synthetic groundtruths emulating fictitious diseases are generated from ten randomly chosen variables following different target pattern distributions using the E-MTAB-3732 dataset. Several sample-to-dimensionality ratios ranging from 300 to 3,000 observations and 854 to 54,675 variables are explored. SEQENS is compared with other feature selection or identification state-of-the-art methods. On average, the proposed algorithm identifies better the relevant genes and exhibits a stronger stability. The algorithm is available to the community.

[EN] Background and Objective: Breast density assessed from digital mammograms is a biomarker for higher risk of developing breast cancer. Experienced radiologists assess breast density using the Breast Image and Data System (BI-RADS) categories. Supervised learning algorithms have been developed with this objective in mind, however, the performance of these algorithms depends on the quality of the ground truth information which is usually labeled by expert readers. These labels are noisy approximations of the ground truth, as there is often intra-and inter-reader variability among labels. Thus, it is crucial to provide a reliable method to obtain digital mammograms matching BI-RADS categories. This paper presents RegL (Labels Regularizer), a methodology that includes different image pre-processes to allow both a correct breast segmentation and the enhancement of image quality through an intensity adjustment, thus allowing the use of deep learning to classify the mammograms into BI-RADS categories. The Confusion Matrix (CM) CNN network used implements an architecture that models each radiologist's noisy label. The final methodology pipeline was determined after comparing the performance of image pre-processes combined with different DL architectures.Methods: A multi-center study composed of 1395 women whose mammograms were classified into the four BI-RADS categories by three experienced radiologists is presented. A total of 892 mammograms were used as the training corpus, 224 formed the validation corpus, and 279 the test corpus.Results: The combination of five networks implementing the RegL methodology achieved the best results among all the models in the test set. The ensemble model obtained an accuracy of (0.85) and a kappa index of 0.71. Conclusions: The proposed methodology has a similar performance to the experienced radiologists in the classification of digital mammograms into BI-RADS categories. This suggests that the pre-processing steps and modelling of each radiologist's label allows for a better estimation of the unknown ground truth labels., This work was partially funded by Generalitat Valenciana through IVACE (Valencian Institute of Business Competitiveness) distributed nominatively to Valencian technological innovation centres under project expedient IMAMCN/2021/1.

[EN] Introduction: Mammographic density (MD), the proportion of radiologically dense breast tissue, is a strong risk factor for breast cancer. Our objective is to investigate the influence of occupations and occupational exposure to physical, chemical, and microbiological agents on MD in Spanish premenopausal women. Methods: This is a cross-sectional study based on 1362 premenopausal workers, aged 39-50, who attended a gynecological screening in a breast radiodiagnosis unit of Madrid City Council. The work history was compiled through a personal interview. Exposure to occupational agents was evaluated using the Spanish job-exposure matrix MatEmESp. MD percentage was assessed using the validated semi-automated computer tool DM-Scan. The association between occupation, occupational exposures, and MD was quantified using multiple linear regression models, adjusted for age, educational level, body mass index, parity, previous breast biopsies, family history of breast cancer, energy intake, use of oral contraceptives, smoking, and alcohol consumption. Results: Although no occupation was statistically significantly associated with MD, a borderline significant inverse association was mainly observed in orchard, greenhouse, nursery, and garden workers (beta = -6.60; 95% confidence interval (95%CI) = -14.27; 1.07) and information and communication technology technicians (beta = -7.27; 95%CI = -15.37; 0.84). On the contrary, a positive association was found among technicians in art galleries, museums, and libraries (beta = 8.47; 95%CI = -0.65; 17.60). Women occupationally exposed to fungicides, herbicides, and insecticides tended to have lower MD. The percentage of density decreased by almost 2% for every 5 years spent in occupations exposed to the mentioned agents. Conclusions: Although our findings point to a lack of association with the occupations and exposures analyzed, this study supports a deeper exploration of the role of certain occupational agents in MD, such as pesticides., This work was supported by the Carlos III Institute of Health (AESI PI15CIII/00029 and AESI PI15CIII/00013). The article presents independent research. The views expressed are those of the authors and not necessarily those of the Carlos III Institute of Health.

[EN] Chest X-ray images are useful for early COVID-19 diagnosis with the advantage that X-ray devices are already available in health centers and images are obtained immediately. Some datasets containing X-ray images with cases (pneumonia or COVID-19) and controls have been made available to develop machine-learning-based methods to aid in diagnosing the disease. However, these datasets are mainly composed of different sources coming from pre-COVID-19 datasets and COVID-19 datasets. Particularly, we have detected a significant bias in some of the released datasets used to train and test diagnostic systems, which might imply that the results published are optimistic and may overestimate the actual predictive capacity of the techniques proposed. In this article, we analyze the existing bias in some commonly used datasets and propose a series of preliminary steps to carry out before the classic machine learning pipeline in order to detect possible biases, to avoid them if possible and to report results that are more representative of the actual predictive power of the methods under analysis., This work was supported by Generalitat Valenciana through the "Instituto Valenciano de Competitividad Empresarial-IVACE'' under Grant IMDEEA/2020/69.

[EN] Background: The role of fatty acids (FAs) on mammographic density (MD) is unclear, and available studies are based on self-reported dietary intake. Objectives: This study assessed the association between specific serum phospholipid fatty acids (PLFAs) and MD in premenopausal women. Methods: The cross-sectional study DDM-Madrid recruited 1392 Spanish premenopausal women, aged 39-50 y, who attended a screening in a breast radiodiagnosis unit of Madrid City Council. Women completed lifestyle questionnaires and FFQs. Percentage MD was estimated using a validated computer tool (DM-Scan), and serum PLFA percentages were measured by GC-MS. Multivariable linear regression models were used to quantify the association of FA tertiles with MD. Models were adjusted for age, education, BMI, waist circumference, parity, oral contraceptive use, previous breast biopsies, and energy intake, and they were corrected for multiple testing. Results: Women in the third tertile of SFAs showed significantly higher MD compared with those in the first tertile (beta(T3vsT1) = 7.53; 95% CI: 5.44, 9.61). Elevated relative concentrations of palmitoleic (beta(T3vsT1) = 3.12; 95% CI: 0.99, 5.25) and gondoic (beta(T3vsT1) = 2.67; 95% CI: 0.57, 4.77) MUFAs, as well as high relative concentrations of palmitelaidic (beta(T3vsT1) = 5.22; 95% CI: 3.15, 7.29) and elaidic (beta(T3vsT1) = 2.69; 95% CI: 0.59, 4.79) trans FAs, were also associated with higher MD. On the contrary, women with elevated relative concentrations of n-6 (omega-6) linoleic (beta(T3vsT1) = -5.49; 95% CI; -7.62, -3.35) and arachidonic (beta(T3vsT1) = -4.68; 95% CI: -6.79, -2.58) PUFAs showed lower MD. Regarding desaturation indices, an elevated palmitoleic to palmitic ratio and a low ratio of oleic to steric and arachidonic to dihomo. gamma-linolenic acids were associated with higher MD. Conclusions: Spanish premenopausal women with high relative concentrations of most SFAs and some MUFAs and trans FAs showed an increased MD, whereas those with high relative concentrations of some n-6 PUFAs presented lower density. These results, which should be confirmed in further studies, underscore the importance of analyzing serum FAs individually., This study was funded by the Carlos III Institute of Health (AESI PI15CIII/0029). The article presents independent research. The views expressed are those of the authors and not necessarily those of the Carlos III Institute of Health.

[EN] Background and Objective: Breast cancer is the most frequent cancer in women. The Spanish healthcare network established population-based screening programs in all Autonomous Communities, where mammograms of asymptomatic women are taken with early diagnosis purposes. Breast density assessed from digital mammograms is a biomarker known to be related to a higher risk to develop breast cancer. It is thus crucial to provide a reliable method to measure breast density from mammograms. Furthermore the complete automation of this segmentation process is becoming fundamental as the amount of mammograms increases every day. Important challenges are related with the differences in images from different devices and the lack of an objective gold standard. This paper presents a fully automated framework based on deep learning to estimate the breast density. The framework covers breast detection, pectoral muscle exclusion, and fibroglandular tissue segmentation. Methods: A multi-center study, composed of 1785 women whose "for presentation" mammograms were segmented by two experienced radiologists. A total of 4992 of the 6680 mammograms were used as training corpus and the remaining (1688) formed the test corpus. This paper presents a histogram normalization step that smoothed the difference between acquisition, a regression architecture that learned segmentation parameters as intrinsic image features and a loss function based on the DICE score. Results: The results obtained indicate that the level of concordance (DICE score) reached by the two radiologists (0.77) was also achieved by the automated framework when it was compared to the closest breast segmentation from the radiologists. For the acquired with the highest quality device, the DICE score per acquisition device reached 0.84, while the concordance between radiologists was 0.76. Conclusions: An automatic breast density estimator based on deep learning exhibits similar performance when compared with two experienced radiologists. It suggests that this system could be used to support radiologists to ease its work., This work was partially funded by Generalitat Valenciana through I+D IVACE (Valencian Institute of Business Competitiviness) and GVA (European Regional Development Fund) supports under the project IMAMCN/2019/1, and by Carlos III Institute of Health under the project DTS15/00080.

[EN] Objectives: Mammographic density (MD) is a strong risk factor for breast cancer. The present study evaluates the association between relative caloric intake and MD in Spanish women. Study design: We conducted a cross-sectional study in which 3517 women were recruited from seven breast cancer screening centers. MD was measured by an experienced radiologist using craniocaudal mammography and Boyd's semi-quantitative scale. Information was collected through an epidemiological survey. Predicted calories were calculated using linear regression models, including the basal metabolic rate and physical activity as explanatory variables. Overeating and caloric restriction were defined taking into account the 99% confidence interval of the predicted value. Odds ratios (OR) and 95% confidence intervals (95%CI) were estimated using center-specific mixed ordinal logistic regression models, adjusted for age, menopausal status, body mass index, parity, tobacco use, family history of breast cancer, previous biopsies, age at menarche and adherence to a Western diet. Main outcome measure: Mammographic density. Results: Those women with an excessive caloric intake ( > 40% above predicted) presented higher MD (OR = 1.41, 95%CI = 0.97-2.03; p = 0.070). For every 20% increase in relative caloric consumption the probability of having higher MD increased by 5% (OR = 1.05, 95%CI = 0.98-1.14; p = 0.178), not observing differences between the categories of explanatory variables. Caloric restriction was not associated with MD in our study. Conclusions: This is the first study exploring the association between MD and the effect of caloric deficit or excessive caloric consumption according to the energy requirements of each woman. Although caloric restriction does not seem to affect breast density, a caloric intake above predicted levels seems to increase this phenotype, This study was supported by the Research Grant FIS PI060386 from Spanish Public Health Research Fund (Fondo de Investigacion Sanitaria); the Carlos III Institute of Health (Institute de Salud Carlos III)PI15CIII/0029 and PI15CIII/00013; the EPY 1306/06Collaboration Agreement between Astra-Zeneca and the Carlos III Institute of Health; and a grant from the Spanish Federation of Breast Cancer Patients (FECMA EPY 1169/10).

[EN] Objectives: The association between breast cancer (BC) and thyroid disorders has been widely explored with unclear results. Mammographic density (MD) is one of the strongest risk factor for BC. This study explores the relationship between thyroid diseases and MD in Spanish women. Materials & methods: This cross-sectional study covered 2883 women aged 47-71 years participating in 7 BC screening programs in 2010. They allowed access to their mammograms, had anthropometrical-measures taken, and answered a telephonic epidemiological interview which included specific questions on thyroid diseases. Percentage of MD was assessed with a semiautomatic-computer tool (DM-scan) by two trained radiologists. We calculated the geometric mean of MD percentages (mean MD). Multivariable mixed linear regression models with random screening-center-specific intercepts were fitted, using log-transformed percentage of MD as dependent variable and adjusting for age, body mass index, menopausal status and other confounders. e(beta) represents the relative increase of mean MD. Results: 13.9% of the participants reported personal history of thyroid disease. MD was not associated to hyperthyroidism (e(beta): 1.05, 95%CI: 0.82-1.36), hypothyroidism (e(beta):1.02, 95%CI: 0.75-1.38), thyroid nodules (e(beta):1.01, 95%CI: 0.85-1.19) or thyroid cancer (e(beta):1.03, 95%CI: 0.56-1.92). However, women with goiter had lower MD (mean MDno-goiter: 13.4% vs mean MDgoiter: 10.6%; e(beta):0.79, 95%CI: 0.64-0.98) and those with Hashimoto thyroiditis had higher MD (mean MDno-thyroiditis: 13.3% vs mean MDthyroidits: 25.8%; e(beta) 95%CI: 1.00-3.77). Conclusion: Functional thyroid disorders were not related to MD. However, MD was lower in women with goiter and higher in those reporting Hashimoto's thyroiditis. These relationships should be confirmed in future studies. (C) 2017 Elsevier Ltd. All rights reserved., This work was supported by the Spanish Ministry of Economy and Competitiveness - Carlos III Institute of Health (ISCIII) (FI14CIII/00013, FIS PI060386 & PS09/0790); the Spanish Federation of Breast Cancer Patients (FECMA 485 EPY 1170-10); the Gent per Gent Fund (EDEMAC Project); and the EPY1306/06 collaboration agreement between Astra-Zeneca and the ISCIII and partially funded by the European Regional Development Fund. None of these funding sources had any role in the study design, in the collection, analysis or interpretation of data. Other members of DDM-Spain/Var-DDM: María Casals, Inmaculada Martínez, María Ederra and Jesús Vioque.

Introduction: Myeloid malignancies are clonal disorders of hematopoietic stem cells and include acute myeloid leukemia (AML), myelodysplastic syndrome (MDS) and myeloproliferative neoplasm (MPN). Common biological markers have been described in the molecular pathogenesis, including gene mutations in splicing factors, epigenetic modifiers, transcription factors, signal pathways and tumor suppressors. These mechanisms have been associated with MDS and MPN progression to AML. Objectives: The main objective of this study is to identify differences in the mutational landscape of myeloid malignancies and describe mutation frequencies of genes and functional pathways in each neoplasm, as well as determine their clinical impact. Methods: This study involved a retrospective analysis of 430 patients with AML (209), MDS (106) and Philadelphia negative MPN (86) diagnosed in the Hospital Universitario 12 de Octubre (Spain). They were analyzed by a next generation sequencing (NGS)- panel for myeloid malignancies. The panel include 32 genes: CALR, ASXL1, EZH2, PHF6, DNMT3A 2, TET2, IDH1, IDH2, KDM6A, KMT2A, SF1, SF3A1, SF3B1, SRSF2, U2AF1, ZRSR2, PRPF40B, EPOR, FLT3, JAK2, KIT, SH2B3, MPL, CBL, HRAS, NRAS, KRAS, ETV6, RUNX1, VHL, TP53, PTEN. In addition, there were included 29 patients diagnosed with benign pathology that were referred to rule out MPN or congenital polyglobulia. Results: In the analyzed cohort we obtained a larger number of mutations in the more aggressive malignancies, AML and MDS. Mutations in epigenetic modifiers and signal pathways were the most frequent detected (31% and 24% respectively). The epigenetic modifiers were notably affected in AML (78%) and MDS (60.4%), whereas signal pathways were mutated more frequently in MPN (70.9%). Transcription factors, tumor suppressors and splicing factors mutations were more detected in AML and MDS (40%, 32%, 44% and 22%, 13%, 32% respectively). The mutation landscape obtained by genes was: Signal pathways: FLT3, NRAS, KIT, KRAS y SH2B3 were specially detected in AML (25%, 11%, 6%, 5% and 4% respectively). JAK2, CALR and MPL in MPN (38%, 15% and 6% respectively). Transcription factors: RUNX1, ETV6, PHF6, CEBPA and WT1 mutations were regularly observed in AML (21%, 6%, 6%, 6% and 5% respectively), and GATA1 in SMD (3.8%). Tumor suppressors: TP53 was particularly affected in AML (21%) and MDS (11%). Epigenetic modifiers: TET2 was notably mutated in MDS (32%), whereas ASXL1, DNMT3A, IDH2, IDH1 and EZH2 were in AML (21%, 21%, 17% 16% and 8% respectively). Splicing factors: SF3B1 was more frequently detected in MDS (18%) than AML (7%), whereas ZRSR2 presented a similar frequency in both pathologies (around 8%). U2AF1 was most commonly mutated in MPN (9%). SRSF2 was specially mutated in AML (23%). SF3A1 was altered in around 1%, similar in all three malignancies. With regard to survival studies, the presence of mutations in splicing factors (primarily in U2AF1) and its absence in signal pathways conferred an adverse outcome for overall survival (OS) in MPN. In MDS, gene mutations in tumor suppressors (especially TP53), U2AF1 splicing factor and EZH2 epigenetic modifier were associated with poor outcome. In our series of AML, gene mutations in tumor suppressors and TP53 were related to unfavorable prognosis in OS. Conclusion: The largest number of mutations and affected genes observed in AML suggest that leukemic transformation of MDS and MPN is conditioned by acquisition of new mutations. We observed different frequencies of mutations between AML, MDS and MPN that could guide the diagnostic and identify new targets of treatment. Further, some mutations have demonstrated differential prognostic impact. An extension of this study and the design of an algorithm with mutation data to elucidate a more accurate molecular prognosis will be presented at the meeting. This work has been financed thanks to the grant PI16/01225, PI 19/01518 and PI19/00730 from the Instituto de Salud Carlos III (Ministerio de Economia, Industria y Competititvidad) and cofinanced by the European Development Fund. Figure 1. Mutations detected (%) in AML, MDS and MPN classified by function. Table 1. Median overall survival of patients with MPN, MDS and AML according to gene state (mutated or not). Figure Disclosures No relevant conflicts of interest to declare.

[EN] Background The breast dense tissue percentage on digital mammograms is one of the most commonly used markers for breast cancer risk estimation. Geometric features of dense tissue over the breast and the presence of texture structures contained in sliding windows that scan the mammograms may improve the predictive ability when combined with the breast dense tissue percentage. Methods A case/control study nested within a screening program covering 1563 women with craniocaudal and mediolateral-oblique mammograms (755 controls and the contralateral breast mammograms at the closest screening visit before cancer diagnostic for 808 cases) aging 45 to 70 from Comunitat Valenciana (Spain) was used to extract geometric and texture features. The dense tissue segmentation was performed using DMScan and validated by two experienced radiologists. A model based on Random Forests was trained several times varying the set of variables. A training dataset of 1172 patients was evaluated with a 10-stratified-fold cross-validation scheme. The area under the Receiver Operating Characteristic curve (AUC) was the metric for the predictive ability. The results were assessed by only considering the output after applying the model to the test set, which was composed of the remaining 391 patients. Results The AUC score obtained by the dense tissue percentage (0.55) was compared to a machine learning-based classifier results. The classifier, apart from the percentage of dense tissue of both views, firstly included global geometric features such as the distance of dense tissue to the pectoral muscle, dense tissue eccentricity or the dense tissue perimeter, obtaining an accuracy of 0.56. By the inclusion of a global feature based on local histograms of oriented gradients, the accuracy of the classifier was significantly improved (0.61). The number of well-classified patients was improved up to 236 when it was 208. Conclusion Relative geometric features of dense tissue over the breast and histograms of standardized local texture features based on sliding windows scanning the whole breast improve risk prediction beyond the dense tissue percentage adjusted by geometrical variables. Other classifiers could improve the results obtained by the conventional Random Forests used in this study., This work was partially funded by Generalitat Valenciana through I+D IVACE (Valencian Institute of Business Competitiviness) and GVA (European Regional Development Fund) supports under the project IMAMCN/2018/1, and by Carlos III Institute of Health under the project DTS15/00080

[EN] Objectives The association between occupational exposures and mammographic density (MD), a marker of breast cancer risk, has not been previously explored. Our objective was to investigate the influence of occupational exposure to chemical, physical and microbiological agents on MD in adult women. Methods This is a population-based cross-sectional study based on 1476 female workers aged 45-65 years from seven Spanish breast cancer screening programmes. Occupational history was surveyed by trained staff. Exposure to occupational agents was assessed using the Spanish job-exposure matrix MatEmESp. Percentage of MD was measured by two radiologists using a semiautomatic computer tool. The association was estimated using mixed log-linear regression models adjusting for age, education, body mass index, menopausal status, parity, smoking, alcohol intake, type of mammography, family history of breast cancer and hormonal therapy use, and including screening centre and professional reader as random effects terms. Results Although no association was found with most of the agents, women occupationally exposed to perchloroethylene (e(beta)=1.51; 95% CI 1.04 to 2.19), ionising radiation (e(beta)=1.23; 95% CI 0.99 to 1.52) and mould spores (e(beta)=1.44; 95% CI 1.01 to 2.04) tended to have higher MD. The percentage of density increased 12% for every 5 years exposure to perchloroethylene or mould spores, 11% for every 5 years exposure to aliphatic/alicyclic hydrocarbon solvents and 3% for each 5 years exposure to ionising radiation. Conclusions Exposure to perchloroethylene, ionising radiation, mould spores or aliphatic/alicyclic hydrocarbon solvents in occupational settings could be associated with higher MD. Further studies are needed to clarify the accuracy and the reasons for these findings., This study was funded by the Spanish Ministry of Economy and Competitiveness - Carlos II Institute of Health (ISCI) (AESI PI15CI/00013); the Scientific Foundation of the Spanish Association Against Cancer (Fundacion Cientifica de la Asociacion Espanola Contra el Cancer (AEC) - EVP-1178/14); the Spanish Public Health Research Fund (FIS PI060386 & PS09/0790); the collaboration agreement between Astra-Zeneca and the ISCI (EPY1306/06) and a grant from the Spanish Federation of Breast Cancer Patients (FECMA 485 EPY 1170-10).

[EN] We explored the relationship between sleep patterns and sleep disorders and mammographic density (MD), a marker of breast cancer risk. Participants in the DDM-Spain/var-DDM study, which included 2878 middle-aged Spanish women, were interviewed via telephone and asked questions on sleep characteristics. Two radiologists assessed MD in their left craneo-caudal mammogram, assisted by a validated semiautomatic-computer tool (DM-scan). We used log-transformed percentage MD as the dependent variable and fitted mixed linear regression models, including known confounding variables. Our results showed that neither sleeping patterns nor sleep disorders were associated with MD. However, women with frequent changes in their bedtime due to anxiety or depression had higher MD (e¿:1.53;95%CI:1.04¿2.26)., This work was supported by grants from the Spanish Ministry of Economy and Competitiveness - Carlos III Institute of Health (ISCIII) (FI14CIII/00013, FIS PI060386 & PS09/0790), from the Spanish Federation of Breast Cancer Patients (FECMA 485 EPY 1170-10), Gent per Gent Fund (EDEMAC Project), the EPY1306/06 collaboration agreement between Astra-Zeneca and the ISCIII and partially funded by the European Regional Development Fund (FEDER)

A method that improves the interaction in Human-Machine Interfaces is proposed.The method combines the evidences of several sources of information.Models are repesented using Weighted Finite-State Transducers (WFSTs) and composed.Models are built separately but final composed WFST keeps full error-recovery power.A significant interaction user effort can be saved when using the proposed process. Although there are many tasks where output strings are automatically generated from a set of evidence, they are not perfect and human intervention is often required to correct the result. In this paper we present a generic Symbol Input Interaction Method for Human-Machine Interfaces that combines multi-source information: an input Hypothesis Model, an Error Model, a Constraint Model and a user interaction scheme.We use Weighted Finite-State Transducers (WFSTs) to represent the different sources of information available: the initial hypotheses, the possible errors, the constraints imposed by the task (interaction language) and the user input. The fusion of these models to find the most probable output string can be performed efficiently by using carefully selected transducer operations. The proposed system initially suggests an output based on the set of hypotheses, possible errors and Constraint Models. Then, if human intervention is needed, a multimodal approach, where the user input is combined with the aforementioned models, is applied to produce, with a minimum user effort, the desired output. This approach offers the practical advantages of a de-coupled model (e.g. input-system + parameterized rules + post-processor), keeping at the same time the error-recovery power of an integrated approach, where all the steps of the process are performed in the same formal machine (as in a typical HMM in speech recognition) to avoid that an error at a given step remains unrecoverable in the subsequent steps. After a presentation of the theoretical basis of the proposed multi-source information system, its application to two real world problems, as an example of the possibilities of this architecture, is addressed. The experimental results obtained demonstrate that a significant user effort can be saved when using the proposed procedure. A simple demonstration, to better understand and evaluate the proposed system, is available on the web https://demos.iti.upv.es/hi/.

Measurement of mammographic density (MD), one of the leading risk factors for breast cancer, still relies on subjective assessment. However, the consistency of MD measurement in full-digital mammograms has yet to be evaluated. We studied inter- and intra-rater agreement with respect to estimation of breast density in full-digital mammograms, and tested whether any of the women's characteristics might have some influence on them. After an initial training period, three experienced radiologists estimated MD using Boyd scale in a left breast cranio-caudal mammogram of 1,431 women, recruited at three Spanish screening centres. A subgroup of 50 randomly selected images was read twice to estimate short-term intra-rater agreement. In addition, a reading of 1,428 of the images, performed 2 years before by one rater, was used to estimate long-term intra-rater agreement. Pair-wise weighted kappas with 95% bootstrap confidence intervals were calculated. Dichotomous variables were defined to identify mammograms in which any rater disagreed with other raters or with his/her own assessment, respectively. The association between disagreement and women's characteristics was tested using multivariate mixed logistic models, including centre as a random-effects term, and taking into account repeated measures when required. All quadratic-weighted kappa values for inter- and intra-rater agreement were excellent (higher than 0.80). None of the studied women's features, i.e. body mass index, brassiere size, menopause, nulliparity, lactation or current hormonal therapy, was associated with higher risk of inter- or intra-rater disagreement. However, raters differed significantly more in images that were classified in the higher-density MD categories, and disagreement in intra-rater assessment was also lower in low-density mammograms. The reliability of MD assessment in full-field digital mammograms is comparable to that for original or digitised images. The reassuring lack of association between subjects' MD-related characteristics and agreement suggests that bias from this source is unlikely.

The task of breast density quantification is becoming increasingly relevant due to its association with breast cancer risk. In this work, a semi-automated and a fully automated tools to assess breast density from full-field digitized mammograms are presented. The first tool is based on a supervised interactive thresholding procedure for segmenting dense from fatty tissue and is used with a twofold goal: for assessing mammographic density(MD) in a more objective and accurate way than via visual-based methods and for labeling the mammograms that are later employed to train the fully automated tool. Although most automated methods rely on supervised approaches based on a global labeling of the mammogram, the proposed method relies on pixel-level labeling, allowing better tissue classification and density measurement on a continuous scale. The fully automated method presented combines a classification scheme based on local features and thresholding operations that improve the performance of the classifier. A dataset of 655 mammograms was used to test the concordance of both approaches in measuring MD. Three expert radiologists measured MD in each of the mammograms using the semi-automated tool (DM-Scan). It was then measured by the fully automated system and the correlation between both methods was computed. The relation between MD and breast cancer was then analyzed using a case-control dataset consisting of 230 mammograms. The Intraclass Correlation Coefficient (ICC) was used to compute reliability among raters and between techniques. The results obtained showed an average ICC = 0.922 among raters when using the semi-automated tool, whilst the average correlation between the semi-automated and automated measures was ICC = 0.838. In the case-control study, the results obtained showed Odds Ratios (OR) of 1.38 and 1.50 per 10% increase in MD when using the semi-automated and fully automated approaches respectively. It can therefore be concluded that the automated and semi-automated MD assessments present a good correlation. Both the methods also found an association between MD and breast cancer risk, which warrants the proposed tools for breast cancer risk prediction and clinical decision making. A full version of the DM-Scan is freely available. (C) 2014 Elsevier Ireland Ltd. All rights reserved., This work was supported by research grants from Gent per Gent Fund (EDEMAC Project); Spain's Health Research Fund (Fondo de Investigacion Santiaria) (PI060386 & FIS PS09/00790); Spanish MICINN grants TIN2009-14205-C04-02 and Consolider-Ingenio 2010: MIPRCV (CSD2007-00018); Spanish Federation of Breast Cancer Patients (Federacion Espanola de Cancer de Mama) (FECMA 485 EPY 1170-10). The English revision of this paper was funded by the Universitat Politecnica de Valencia, Spain.

We developed a semi-automated tool to assess mammographic density (MD), a phenotype risk marker for breast cancer (BC), in full-field digital images and evaluated its performance testing its reproducibility, comparing our MD estimates with those obtained by visual inspection and using Cumulus, verifying their association with factors that influence MD, and studying the association between MD measures and subsequent BC risk. Three radiologists assessed MD using DM-Scan, the new tool, on 655 processed images (craniocaudal view) obtained in two screening centers. Reproducibility was explored computing pair-wise concordance correlation coefficients (CCC). The agreement between DM-Scan estimates and visual assessment (semi- uantitative scale, 6 categories) was quantified computing weighted kappa statistics (quadratic weights). DM-Scan and Cumulus readings were compared using CCC. Variation of DM-Scan measures by age, body mass index (BMI) and other MD modifiers was tested in regression mixed models with mammographic device as a random-effect term. The association between DM-Scan measures and subsequent BC was estimated in a case control study. All BC cases in screening attendants (2007 2010) at a center with full-field digital mammography were matched by age and screening year with healthy controls (127 pairs). DM-Scan was used to blindly assess MD in available mammograms (112 cases/119 controls). Unconditional logistic models were fitted, including age, menopausal status and BMI as confounders. DM-Scan estimates were very reliable (pairwise CCC: 0.921, 0.928 and 0.916). They showed a reasonable agreement with visual MD assessment (weighted kappa ranging 0.79-0.81). DM-Scan and Cumulus measures were highly concordant (CCC ranging 0.80-0.84), but ours tended to be higher (4%-5% on average). As expected, DM-Scan estimates varied with age, BMI, parity and family history of BC. Finally, DM-Scan measures were significantly associated with BC (p-trend=0.005). Taking MD=29%=3.10 (95% CI=1.35-7.14). Our results confirm that DM-Scan is a reliable tool to assess MD in full-field digital mammograms., This work was supported by research grants from Spain's Health Research Fund (Fondo de INvestigacion Santiaria) (PI060386 & PI09/1230); Gent per Gent Fund (EDEMAC Project) and the Spanish Federation of Breast Cancer Patients (Federacion Espanola de Cancer de Mama) (FECMA 485 EPY 1170-10).

In this paper, a generic Symbol Input Correction Method for Human-Machine Interfaces, especially useful for embedded devices where the input subsystem is often size-constrained, combining a Language Model, the Input Hypothesis information and an Error Model is proposed. The approach can be seen as a flexible and efficient way to perform Stochastic Error-Correcting Language Modeling. We use Weighted Finite-State Transducers (WFSTs) to represent the Language Model, the complete set of symbol-input Hypotheses interpreted as a sequence of vectors of symbol probabilities, and an Error Model. This approach is different to other methods since it does not involve an explicit parsing process and also because it combines the practical advantages of a de-coupled (input-system + post-processor) model with the error-recovery power of a integrated approach where the capture and the interpretation are performed in the same element (e.g. in a HMM). The symbol-input subsystem can be a physical, "soft" (touch screen-based) or reduced (as in a mobile phone) keyboard, a speech or gesture-based recognizer, an Off-line or On-line OCR system or any other Human-Machine Interface consisting of a sequence of symbols conveying information belonging to a language where the segmentation of the input is known (although some segmentation errors can be recovered by the Error Model).

In an automatic handwritten form processing system it is often necessary to use the lexical or linguistic restrictions present in the field contents in order to obtain acceptable recognition rates. Since each field is known to hold a given kind of information (name, address...), a language model can be defined for it. But, often, in a typical form there are fields linked by known relations, like “Street” and “Postal Code” or “Country” and “City”. We have used Weighted Finite-State Transducers (WFSTs) to combine Stochastic Error-Correcting Language Models from different interdependent fields in real handwritten forms and measured the improvements obtained.

In an OCR post-processing task, a language model is used to find the best transformation of the OCR hypothesis into a string compatible with the language. The cost of this transformation is used as a confidence value to reject the strings that are less likely to be correct, and the error rate of the accepted strings should be strictly controlled by the user. In this work, the expected error rate distribution of an unknown language model is estimated from a training set composed of known language models. This means that after building a new language model, the user should be able to automatically "fix" the expected error rate at an acceptable level instead of having to deal with an arbitrary threshold.

Para diagnosticar un cáncer se realiza, entre otras pruebas, algún test de imagen, como puede ser una radiografía, ecografía o resonancia magnética. Mediante estos tests pueden detectarse zonas con alta sospecha tumoral, cuyo diagnóstico debe confirmase finalmente mediante la realización de una biopsia. Este tipo de imágenes, sin embargo, no son fáciles de interpretar, lo que provoca que el profesional encargado de analizarlas, a pesar de su experiencia, no sea capaz de detectar en ellas un porcentaje importante de tumores (falsos negativos). Una posibilidad para mejorar el diagnóstico y disminuir el número de falsos negativos consiste en utilizar sistemas de diagnóstico asistido por ordenador o computer-aided diagnosis (CAD). Un sistema de CAD analiza la imagen médica y trata de detectar zonas sospechosas de contener alguna anomalía. Estas zonas son marcadas sobre la propia imagen con un doble objetivo: llamar la atención del profesional encargado de analizarla hacia la zona sospechosa y aportar una segunda opinión respecto al diagnóstico. En esta tesis se presentan y evaluan diversas técnicas de visión por computador y reconocimiento de formas orientadas a la detección de tumores en imágenes médicas, con el objetivo de diseñar sistemas de CAD que permitan un mejor diagnóstico. El trabajo se ha centrado en el diagnóstico de cáncer de próstata a partir de imágenes de ecografía, y en el diagnóstico de cáncer de mama a partir de imágenes de radiografía. Se han evaluado diversos métodos de extracción de características basados en la intensidad, frecuencia, texturas o en gradientes. En la etapa de clasificación se ha utilizado un clasificador no paramétrico basado en distancias (k-vecinos más cercanos) y otro paramétrico basado en modelos de Markov. A lo largo del trabajo se evidencian las distintas problemáticas que surgen en este tipode tareas y se proponen soluciones a cada una de ellas. El diagnóstico de cáncer de próstata asistido por ordenador es una tarea extrema, Llobet Azpitarte, R. (2006). Aportaciones al diagnóstico de cáncer asistido por ordenador [Tesis doctoral no publicada]. Universitat Politècnica de València. doi:10.4995/Thesis/10251/1862.

Background Prostate cancer is one of the most frequent cancers in men and is a major cause of mortality in developed countries. Detection of prostate carcinoma at an early stage is crucial for successful treatment. Material and methods A method for the analysis of transrectal ultrasound images aimed at computer-aided diagnosis of prostate cancer is tested in this paper. First, two classifiers based on k -nearest neighbors and Hidden Markov models are compared. Second, the diagnostic capacity of our system is tested by means of a set of experiments where humans with varying degrees of experience classified a set of ultrasound images with and without the aid of the computer-aided system. The corpus used in this study was specifically acquired for this purpose. It consists of 4944 ultrasound images corresponding to 303 patients, and is publicly available for non-commercial use upon request. Results The best classification results achieve an area under the receiver operating characteristic curve of 61.6%. However, the diagnostic capacity of an expert urologist using the computer-aided system improves only slightly compared with his/her capacity without the aid of the system. Conclusions Despite the difficulty of this task, the obtained results indicate that discrimination between cancerous and non-cancerous tissue is possible to a certain degree. The computer-aided system helps an inexperienced user to make a better diagnosis, however it must be able to perform better in order to be useful in a real-world clinical context.

Prostate cancer is one of the most frequent cancer in men and a major cause of mortality in developed countries. Detection of the prostate carcinoma at an early stage is crucial for a succesfull treatment. In this paper, a method for analysis of transrectal ultrasonography images aimed at computer-aided diagnosis of prostate cancer is presented. Althogh the task is extremely difficult due to a problem of imperfect supervision, we have obtained promising results indicating that valid information for the diagnostic is present in the images. Two classifiers based on k-Nearest Neighbours and Hidden Markov Models are compared.

A feature extraction method for images is presented. It is based on the classical concept of distance transformation (DT) from which we develop a generalization: the continuous distance transformation (CDT). Whereas the DT can only be applied to binary images, the CDT can be applied to both binary and gray-scale or color pictures. Furthermore, we define a number of new metrics and dissimilarity measures based on the CDT. Comparative experimental results are also presented for the new measures, using the NIST handwritten characters database. Several experiments using the k-nearest neighbors classification rule are reported, with results that significantly improve the recognition rate of other measures like the Euclidean and some DT-based distances.

In this paper, a process of expansion of the training set by synthetic generation of handwritten uppercase letters via deformations of natural images is tested in combination with an approximate k-Nearest Neighbor (k-NN) classifier. It has been previously shown [11] [10] that approximate nearest neighbors search in large databases can be successfully used in an OCR task, and that significant performance improvements can be consistently obtained by simply increasing the size of the training set. In this work, extensive experiments adding distorted characters to the training set are performed, and the results are compared to directly adding new natural samples to the set of prototypes.