The overlooked contribution of trees outside forests to tree cover and woody biomass across Europe

Trees are an integral part in European landscapes, but only forest resources are systematically assessed by national inventories. The contribution of urban and agricultural trees to national-level carbon stocks remains largely unknown. Here we produced canopy cover, height and above-ground biomass maps from 3-meter resolution nanosatellite imagery across Europe. Our biomass estimates have a systematic bias of 7.6% (overestimation; R = 0.98) compared to national inventories of 30 countries, and our dataset is sufficiently highly resolved spatially to support the inclusion of tree biomass outside forests, which we quantify to 0.8 petagrams. Although this represents only 2% of the total tree biomass, large variations between countries are found (10% for UK) and trees in urban areas contribute substantially to national carbon stocks (8% for the Netherlands). The agreement with national inventory data, the scalability, and spatial details across landscapes, including trees outside forests, make our approach attractive for operational implementation to support national carbon stock inventory schemes.

More than one quarter of Africa's tree cover is found outside areas previously classified as forest

The consistent monitoring of trees both inside and outside of forests is key to sustainable land management. Current monitoring systems either ignore trees outside forests or are too expensive to be applied consistently across countries on a repeated basis. Here we use the PlanetScope nanosatellite constellation, which delivers global very high-resolution daily imagery, to map both forest and non-forest tree cover for continental Africa using images from a single year. Our prototype map of 2019 (RMSE = 9.57%, bias = -6.9%). demonstrates that a precise assessment of all tree-based ecosystems is possible at continental scale, and reveals that 29% of tree cover is found outside areas previously classified as tree cover in state-of-the-art maps, such as in croplands and grassland. Such accurate mapping of tree cover down to the level of individual trees and consistent among countries has the potential to redefine land use impacts in non-forest landscapes, move beyond the need for forest definitions, and build the basis for natural climate solutions and tree-related studies.

Towards more comprehensive nationwide familial aggregation studies in Denmark: The Danish Civil Registration System versus the lite Danish Multi-Generation Register

AIM

Linking information on family members in the Danish Civil Registration System (CRS) with information in Danish national registers provides unique possibilities for research on familial aggregation of diseases, health patterns, social factors and demography. However, the CRS is limited in the number of generations that it can identify. To allow more complete familial linkages, we introduce the lite Danish Multi-Generation Register (lite MGR) and the future full Danish MGR that is currently being developed.

METHODS

We generated the lite MGR by linking the current version of the CRS with historical versions stored by the Danish National Archives in the early 1970s, which contain familial links not saved in the current CRS. We describe and compare the completeness of familial links in the lite MGR and the current version of the CRS. We also describe planned procedures for generating the full MGR by linking the current CRS with scanned archived records from Parish Registers.

RESULTS

Among people born in Denmark in 1960 or later, the current CRS contains information on both parents. However, it has limited parental information for people born earlier. Among the 732,232 people born in Denmark during 1950-1959, 444,084 (60.65%) had information on both parents in the CRS. In the lite MGR, it was 560,594 (76.56%).

CONCLUSIONS

The lite MGR offers more complete information on familial relationships than the current CRS. The lite and full MGR will offer an infrastructure tying together existing research infrastructures, registers and biobanks, raising their joint research value to an unparalleled level.

Deep learning enables image-based tree counting, crown segmentation, and height prediction at national scale

Sustainable tree resource management is the key to mitigating climate warming, fostering a green economy, and protecting valuable habitats. Detailed knowledge about tree resources is a prerequisite for such management but is conventionally based on plot-scale data, which often neglects trees outside forests. Here, we present a deep learning-based framework that provides location, crown area, and height for individual overstory trees from aerial images at country scale. We apply the framework on data covering Denmark and show that large trees (stem diameter >10 cm) can be identified with a low bias (12.5%) and that trees outside forests contribute to 30% of the total tree cover, which is typically unrecognized in national inventories. The bias is high (46.6%) when our results are evaluated against all trees taller than 1.3 m, which involve undetectable small or understory trees. Furthermore, we demonstrate that only marginal effort is needed to transfer our framework to data from Finland, despite markedly dissimilar data sources. Our work lays the foundation for digitalized national databases, where large trees are spatially traceable and manageable.

Sub-continental-scale carbon stocks of individual trees in African drylands

The distribution of dryland trees and their density, cover, size, mass and carbon content are not well known at sub-continental to continental scales1-14. This information is important for ecological protection, carbon accounting, climate mitigation and restoration efforts of dryland ecosystems15-18. We assessed more than 9.9 billion trees derived from more than 300,000 satellite images, covering semi-arid sub-Saharan Africa north of the Equator. We attributed wood, foliage and root carbon to every tree in the 0-1,000 mm year-1 rainfall zone by coupling field data19, machine learning20-22, satellite data and high-performance computing. Average carbon stocks of individual trees ranged from 0.54 Mg C ha-1 and 63 kg C tree-1 in the arid zone to 3.7 Mg C ha-1 and 98 kg tree-1 in the sub-humid zone. Overall, we estimated the total carbon for our study area to be 0.84 (±19.8%) Pg C. Comparisons with 14 previous TRENDY numerical simulation studies23 for our area found that the density and carbon stocks of scattered trees have been underestimated by three models and overestimated by 11 models, respectively. This benchmarking can help understand the carbon cycle and address concerns about land degradation24-29. We make available a linked database of wood mass, foliage mass, root mass and carbon stock of each tree for scientists, policymakers, dryland-restoration practitioners and farmers, who can use it to estimate farmland tree carbon stocks from tablets or laptops.

The Liver Tumor Segmentation Benchmark (LiTS)

In this work, we report the set-up and results of the Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with the IEEE International Symposium on Biomedical Imaging (ISBI) 2017 and the International Conferences on Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2017 and 2018. The image dataset is diverse and contains primary and secondary tumors with varied sizes and appearances with various lesion-to-background levels (hyper-/hypo-dense), created in collaboration with seven hospitals and research institutions. Seventy-five submitted liver and liver tumor segmentation algorithms were trained on a set of 131 computed tomography (CT) volumes and were tested on 70 unseen test images acquired from different patients. We found that not a single algorithm performed best for both liver and liver tumors in the three events. The best liver segmentation algorithm achieved a Dice score of 0.963, whereas, for tumor segmentation, the best algorithms achieved Dices scores of 0.674 (ISBI 2017), 0.702 (MICCAI 2017), and 0.739 (MICCAI 2018). Retrospectively, we performed additional analysis on liver tumor detection and revealed that not all top-performing segmentation algorithms worked well for tumor detection. The best liver tumor detection method achieved a lesion-wise recall of 0.458 (ISBI 2017), 0.515 (MICCAI 2017), and 0.554 (MICCAI 2018), indicating the need for further research. LiTS remains an active benchmark and resource for research, e.g., contributing the liver-related segmentation tasks in http://medicaldecathlon.com/. In addition, both data and online evaluation are accessible via https://competitions.codalab.org/competitions/17094.

Nation-wide mapping of tree-level aboveground carbon stocks in Rwanda

Trees sustain livelihoods and mitigate climate change but a predominance of trees outside forests and limited resources make it difficult for many tropical countries to conduct automated nation-wide inventories. Here, we propose an approach to map the carbon stock of each individual overstory tree at the national scale of Rwanda using aerial imagery from 2008 and deep learning. We show that 72% of the mapped trees are located in farmlands and savannas and 17% in plantations, accounting for 48.6% of the national aboveground carbon stocks. Natural forests cover 11% of the total tree count and 51.4% of the national carbon stocks, with an overall carbon stock uncertainty of 16.9%. The mapping of all trees allows partitioning to any landscapes classification and is urgently needed for effective planning and monitoring of restoration activities as well as for optimization of carbon sequestration, biodiversity and economic benefits of trees.

Cross-Cohort Automatic Knee MRI Segmentation With Multi-Planar U-Nets

Background

Segmentation of medical image volumes is a time‐consuming manual task. Automatic tools are often tailored toward specific patient cohorts, and it is unclear how they behave in other clinical settings.

Purpose

To evaluate the performance of the open‐source Multi‐Planar U‐Net (MPUnet), the validated Knee Imaging Quantification (KIQ) framework, and a state‐of‐the‐art two‐dimensional (2D) U‐Net architecture on three clinical cohorts without extensive adaptation of the algorithms.

Study Type

Retrospective cohort study.

Subjects

A total of 253 subjects (146 females, 107 males, ages 57 ± 12 years) from three knee osteoarthritis (OA) studies (Center for Clinical and Basic Research [CCBR], Osteoarthritis Initiative [OAI], and Prevention of OA in Overweight Females [PROOF]) with varying demographics and OA severity (64/37/24/53/2 scans of Kellgren and Lawrence [KL] grades 0–4).

Field Strength/Sequence

0.18 T, 1.0 T/1.5 T, and 3 T sagittal three‐dimensional fast‐spin echo T1w and dual‐echo steady‐state sequences.

Assessment

All models were fit without tuning to knee magnetic resonance imaging (MRI) scans with manual segmentations from three clinical cohorts. All models were evaluated across KL grades.

Statistical Tests

Segmentation performance differences as measured by Dice coefficients were tested with paired, two‐sided Wilcoxon signed‐rank statistics with significance threshold α = 0.05.

Results

The MPUnet performed superior or equal to KIQ and 2D U‐Net on all compartments across three cohorts. Mean Dice overlap was significantly higher for MPUnet compared to KIQ and U‐Net on CCBR (0.83±0.04 vs. 0.81±0.06 and 0.82±0.05), significantly higher than KIQ and U‐Net OAI (0.86±0.03 vs. 0.84±0.04 and 0.85±0.03), and not significantly different from KIQ while significantly higher than 2D U‐Net on PROOF (0.78±0.07 vs. 0.77±0.07, P=0.10, and 0.73±0.07). The MPUnet performed significantly better on N=22 KL grade 3 CCBR scans with 0.78±0.06 vs. 0.75±0.08 for KIQ and 0.76±0.06 for 2D U‐Net.

Data Conclusion

The MPUnet matched or exceeded the performance of state‐of‐the‐art knee MRI segmentation models across cohorts of variable sequences and patient demographics. The MPUnet required no manual tuning making it both accurate and easy‐to‐use.

Level of Evidence

3

Technical Efficacy

Stage 2

Using machine learning for predicting intensive care unit resource use during the COVID-19 pandemic in Denmark

The COVID-19 pandemic has put massive strains on hospitals, and tools to guide hospital planners in resource allocation during the ebbs and flows of the pandemic are urgently needed. We investigate whether machine learning (ML) can be used for predictions of intensive care requirements a fixed number of days into the future. Retrospective design where health Records from 42,526 SARS-CoV-2 positive patients in Denmark was extracted. Random Forest (RF) models were trained to predict risk of ICU admission and use of mechanical ventilation after n days (n = 1, 2, …, 15). An extended analysis was provided for n = 5 and n = 10. Models predicted n-day risk of ICU admission with an area under the receiver operator characteristic curve (ROC-AUC) between 0.981 and 0.995, and n-day risk of use of ventilation with an ROC-AUC between 0.982 and 0.997. The corresponding n-day forecasting models predicted the needed ICU capacity with a coefficient of determination (R2) between 0.334 and 0.989 and use of ventilation with an R2 between 0.446 and 0.973. The forecasting models performed worst, when forecasting many days into the future (for large n). For n = 5, ICU capacity was predicted with ROC-AUC 0.990 and R2 0.928, and use of ventilator was predicted with ROC-AUC 0.994 and R2 0.854. Random Forest-based modelling can be used for accurate n-day forecasting predictions of ICU resource requirements, when n is not too large.

The International Workshop on Osteoarthritis Imaging Knee MRI Segmentation Challenge: A Multi-Institute Evaluation and Analysis Framework on a Standardized Dataset

PURPOSE

To organize a multi-institute knee MRI segmentation challenge for characterizing the semantic and clinical efficacy of automatic segmentation methods relevant for monitoring osteoarthritis progression.

MATERIALS AND METHODS

A dataset partition consisting of three-dimensional knee MRI from 88 retrospective patients at two time points (baseline and 1-year follow-up) with ground truth articular (femoral, tibial, and patellar) cartilage and meniscus segmentations was standardized. Challenge submissions and a majority-vote ensemble were evaluated against ground truth segmentations using Dice score, average symmetric surface distance, volumetric overlap error, and coefficient of variation on a holdout test set. Similarities in automated segmentations were measured using pairwise Dice coefficient correlations. Articular cartilage thickness was computed longitudinally and with scans. Correlation between thickness error and segmentation metrics was measured using the Pearson correlation coefficient. Two empirical upper bounds for ensemble performance were computed using combinations of model outputs that consolidated true positives and true negatives.

RESULTS

Six teams (T 1-T 6) submitted entries for the challenge. No differences were observed across any segmentation metrics for any tissues (P = .99) among the four top-performing networks (T 2, T 3, T 4, T 6). Dice coefficient correlations between network pairs were high (> 0.85). Per-scan thickness errors were negligible among networks T 1-T 4 (P = .99), and longitudinal changes showed minimal bias (< 0.03 mm). Low correlations (ρ < 0.41) were observed between segmentation metrics and thickness error. The majority-vote ensemble was comparable to top-performing networks (P = .99). Empirical upper-bound performances were similar for both combinations (P = .99).

CONCLUSION

Diverse networks learned to segment the knee similarly, where high segmentation accuracy did not correlate with cartilage thickness accuracy and voting ensembles did not exceed individual network performance.See also the commentary by Elhalawani and Mak in this issue.Keywords: Cartilage, Knee, MR-Imaging, Segmentation © RSNA, 2020Supplemental material is available for this article.

U-Sleep: resilient high-frequency sleep staging

Sleep disorders affect a large portion of the global population and are strong predictors of morbidity and all-cause mortality. Sleep staging segments a period of sleep into a sequence of phases providing the basis for most clinical decisions in sleep medicine. Manual sleep staging is difficult and time-consuming as experts must evaluate hours of polysomnography (PSG) recordings with electroencephalography (EEG) and electrooculography (EOG) data for each patient. Here, we present U-Sleep, a publicly available, ready-to-use deep-learning-based system for automated sleep staging ( sleep.ai.ku.dk ). U-Sleep is a fully convolutional neural network, which was trained and evaluated on PSG recordings from 15,660 participants of 16 clinical studies. It provides accurate segmentations across a wide range of patient cohorts and PSG protocols not considered when building the system. U-Sleep works for arbitrary combinations of typical EEG and EOG channels, and its special deep learning architecture can label sleep stages at shorter intervals than the typical 30 s periods used during training. We show that these labels can provide additional diagnostic information and lead to new ways of analyzing sleep. U-Sleep performs on par with state-of-the-art automatic sleep staging systems on multiple clinical datasets, even if the other systems were built specifically for the particular data. A comparison with consensus-scores from a previously unseen clinic shows that U-Sleep performs as accurately as the best of the human experts. U-Sleep can support the sleep staging workflow of medical experts, which decreases healthcare costs, and can provide highly accurate segmentations when human expertize is lacking.

Developing and validating COVID-19 adverse outcome risk prediction models from a bi-national European cohort of 5594 patients

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.

An unexpectedly large count of trees in the West African Sahara and Sahel

A large proportion of dryland trees and shrubs (hereafter referred to collectively as trees) grow in isolation, without canopy closure. These non-forest trees have a crucial role in biodiversity, and provide ecosystem services such as carbon storage, food resources and shelter for humans and animals^1,2. However, most public interest relating to trees is devoted to forests, and trees outside of forests are not well-documented³. Here we map the crown size of each tree more than 3 m² in size over a land area that spans 1.3 million km² in the West African Sahara, Sahel and sub-humid zone, using submetre-resolution satellite imagery and deep learning⁴. We detected over 1.8 billion individual trees (13.4 trees per hectare), with a median crown size of 12 m², along a rainfall gradient from 0 to 1,000 mm per year. The canopy cover increases from 0.1% (0.7 trees per hectare) in hyper-arid areas, through 1.6% (9.9 trees per hectare) in arid and 5.6% (30.1 trees per hectare) in semi-arid zones, to 13.3% (47 trees per hectare) in sub-humid areas. Although the overall canopy cover is low, the relatively high density of isolated trees challenges prevailing narratives about dryland desertification^5-7, and even the desert shows a surprisingly high tree density. Our assessment suggests a way to monitor trees outside of forests globally, and to explore their role in mitigating degradation, climate change and poverty.

Impact of device programming on the success of the first anti-tachycardia pacing therapy: An anonymized large-scale study

Background

Antitachycardia pacing (ATP) is an effective treatment for ventricular tachycardia (VT). We evaluated the efficacy of different ATP programs based on a large remote monitoring data set from patients with implantable cardioverter-defibrillators (ICDs).

Methods

A dataset from 18,679 ICD patients was used to evaluate the first delivered ATP treatment. We considered all device programs that were used for at least 50 patients, leaving us with 7 different programs and a total of 32,045 episodes. We used the two-proportions z-test (α = 0.01) to compare the probability of success and the probability for acceleration in each group with the corresponding values of the default setting.

Results

Overall, the first ATP treatment terminated in 78.4%–97.5% of episodes with slow VT and 81.5%–91.1% of episodes with fast VT. The default setting of the ATP programs with the number of sequences S = 3 was applied to treat 30.1% of the slow and 36.6% of the fast episodes. Reducing the maximum number of sequences to S = 2 decreased the success rate for slow VT (P < 0.0001, h = 0.38), while the setting S = 4 resulted in the highest success rate of 97.5% (P < 0.0001, h = 0.27).

Conclusion

While the default programs performed well, we found that increasing the number of sequences from 3 to 4 was a promising option to improve the overall ATP performance.

Predicting electrical storms by remote monitoring of implantable cardioverter-defibrillator patients using machine learning

Aims

Electrical storm (ES) is a serious arrhythmic syndrome that is characterized by recurrent episodes of ventricular arrhythmias. Electrical storm is associated with increased mortality and morbidity despite the use of implantable cardioverter-defibrillators (ICDs). Predicting ES could be essential; however, models for predicting this event have never been developed. The goal of this study was to construct and validate machine learning models to predict ES based on daily ICD remote monitoring summaries.

Methods and results

Daily ICD summaries from 19 935 patients were used to construct and evaluate two models [logistic regression (LR) and random forest (RF)] for predicting the short-term risk of ES. The models were evaluated on the parts of the data not used for model development. Random forest performed significantly better than LR (P < 0.01), achieving a test accuracy of 0.96 and an area under the curve (AUC) of 0.80 (vs. an accuracy of 0.96 and an AUC of 0.75). The percentage of ventricular pacing and the daytime activity were the most relevant variables in the RF model.

Conclusion

The use of large-scale machine learning showed that daily summaries of ICD measurements in the absence of clinical information can predict the short-term risk of ES.

Differential diagnosis of mild cognitive impairment and Alzheimer's disease using structural MRI cortical thickness, hippocampal shape, hippocampal texture, and volumetry

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.

•

The algorithm that won the CADDementia challenge is described and analyzed.

•

Evaluation on data from ADNI, AIBL and the CADDementia challenge.

•

Hippocampal texture is shown to be an important feature in the algorithm.

•

Structural MRI intensity variations may include so far unused information.

•

It is conjectured that additional features are needed in order to improve diagnostic performance.

Separating Timing, Movement Conditions and Individual Differences in the Analysis of Human Movement

A central task in the analysis of human movement behavior is to determine systematic patterns and differences across experimental conditions, participants and repetitions. This is possible because human movement is highly regular, being constrained by invariance principles. Movement timing and movement path, in particular, are linked through scaling laws. Separating variations of movement timing from the spatial variations of movements is a well-known challenge that is addressed in current approaches only through forms of preprocessing that bias analysis. Here we propose a novel nonlinear mixed-effects model for analyzing temporally continuous signals that contain systematic effects in both timing and path. Identifiability issues of path relative to timing are overcome by using maximum likelihood estimation in which the most likely separation of space and time is chosen given the variation found in data. The model is applied to analyze experimental data of human arm movements in which participants move a hand-held object to a target location while avoiding an obstacle. The model is used to classify movement data according to participant. Comparison to alternative approaches establishes nonlinear mixed-effects models as viable alternatives to conventional analysis frameworks. The model is then combined with a novel factor-analysis model that estimates the low-dimensional subspace within which movements vary when the task demands vary. Our framework enables us to visualize different dimensions of movement variation and to test hypotheses about the effect of obstacle placement and height on the movement path. We demonstrate that the approach can be used to uncover new properties of human movement.

When you move a cup to a new location on a table, the movement of lifting, transporting, and setting down the cup appears to be completely automatic. Although the hand could take continuously many different paths and move on any temporal trajectory, real movements are highly regular and reproducible. From repetition to repetition movements vary, and the pattern of variance reflects movement conditions and movement timing. If another person performs the same task, the movement will be similar. When we look more closely, however, there are systematic individual differences. Some people will overcompensate when avoiding an obstacle and some people will systematically move slower than others. When we want to understand human movement, all these aspects are important. We want to know which parts of a movement are common across people and we want to quantify the different types of variability. Thus, the models we use to analyze movement data should contain all the mentioned effects. In this work, we developed a framework for statistical analysis of movement data that respects these structures of movements. We showed how this framework modeled the individual characteristics of participants better than other state-of-the-art modeling approaches. We combined the timing-and-path-separating model with a novel factor analysis model for analyzing the effect of obstacles on spatial movement paths. This combination allowed for an unprecedented ability to quantify and display different sources of variation in the data. We analyzed data from a designed experiment of arm movements under various obstacle avoidance conditions. Using the proposed statistical models, we documented three findings: a linearly amplified deviation in mean path related to increase in obstacle height; a consistent asymmetric pattern of variation along the movement path related to obstacle placement; and the existence of obstacle-distance invariant focal points where mean trajectories intersect in the frontal and vertical planes.

Unsupervised Deep Learning Applied to Breast Density Segmentation and Mammographic Risk Scoring

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.

Early detection of Alzheimer's disease using MRI hippocampal texture

Cognitive impairment in patients with Alzheimer's disease (AD) is associated with reduction in hippocampal volume in magnetic resonance imaging (MRI). However, it is unknown whether hippocampal texture changes in persons with mild cognitive impairment (MCI) that does not have a change in hippocampal volume. We tested the hypothesis that hippocampal texture has association to early cognitive loss beyond that of volumetric changes. The texture marker was trained and evaluated using T1-weighted MRI scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, and subsequently applied to score independent data sets from the Australian Imaging, Biomarker & Lifestyle Flagship Study of Ageing (AIBL) and the Metropolit 1953 Danish Male Birth Cohort (Metropolit). Hippocampal texture was superior to volume reduction as predictor of MCI-to-AD conversion in ADNI (area under the receiver operating characteristic curve [AUC] 0.74 vs. 0.67; DeLong test, p = 0.005), and provided even better prognostic results in AIBL (AUC 0.83). Hippocampal texture, but not volume, correlated with Addenbrooke's cognitive examination score (Pearson correlation, r = -0.25, p < 0.001) in the Metropolit cohort. The hippocampal texture marker correlated with hippocampal glucose metabolism as indicated by fluorodeoxyglucose-positron emission tomography (Pearson correlation, r = -0.57, p < 0.001). Texture statistics remained significant after adjustment for volume in all cases, and the combination of texture and volume did not improve diagnostic or prognostic AUCs significantly. Our study highlights the presence of hippocampal texture abnormalities in MCI, and the possibility that texture may serve as a prognostic neuroimaging biomarker of early cognitive impairment.

Femoral cartilage segmentation in knee MRI scans using two stage voxel classification

Using more than one classification stage and exploiting class population imbalance allows for incorporating powerful classifiers in tasks requiring large scale training data, even if these classifiers scale badly with the number of training samples. This led us to propose a two-stage classifier for segmenting tibial cartilage in knee MRI scans combining nearest neighbor classification and support vector machines (SVMs). Here we apply it to femoral cartilage segmentation. We describe the similarities and differences between segmenting these two knee cartilages. For further speeding up batch SVM training, we propose loosening the stopping condition in the quadratic program solver before considering moving on to other approximation techniques such as online SVMs. The two-stage approach reached a higher accuracy in comparison to the one-stage state-of-the-art method. It also achieved better inter-scan segmentation reproducibility when compared to a radiologist as well as the current state-of-the-art method.

Bounding the Bias of Contrastive Divergence Learning

Optimization based on k-step contrastive divergence (CD) has become a common way to train restricted Boltzmann machines (RBMs). The k-step CD is a biased estimator of the log-likelihood gradient relying on Gibbs sampling. We derive a new upper bound for this bias. Its magnitude depends on k, the number of variables in the RBM, and the maximum change in energy that can be produced by changing a single variable. The last reflects the dependence on the absolute values of the RBM parameters. The magnitude of the bias is also affected by the distance in variation between the modeled distribution and the starting distribution of the Gibbs chain.

A dynamic neural field model of mesoscopic cortical activity captured with voltage-sensitive dye imaging

A neural field model is presented that captures the essential non-linear characteristics of activity dynamics across several millimeters of visual cortex in response to local flashed and moving stimuli. We account for physiological data obtained by voltage-sensitive dye (VSD) imaging which reports mesoscopic population activity at high spatio-temporal resolution. Stimulation included a single flashed square, a single flashed bar, the line-motion paradigm--for which psychophysical studies showed that flashing a square briefly before a bar produces sensation of illusory motion within the bar--and moving squares controls. We consider a two-layer neural field (NF) model describing an excitatory and an inhibitory layer of neurons as a coupled system of non-linear integro-differential equations. Under the assumption that the aggregated activity of both layers is reflected by VSD imaging, our phenomenological model quantitatively accounts for the observed spatio-temporal activity patterns. Moreover, the model generalizes to novel similar stimuli as it matches activity evoked by moving squares of different speeds. Our results indicate that feedback from higher brain areas is not required to produce motion patterns in the case of the illusory line-motion paradigm. Physiological interpretation of the model suggests that a considerable fraction of the VSD signal may be due to inhibitory activity, supporting the notion that balanced intra-layer cortical interactions between inhibitory and excitatory populations play a major role in shaping dynamic stimulus representations in the early visual cortex.

Maximum likelihood model selection for 1-norm soft margin SVMs with multiple parameters

Adapting the hyperparameters of support vector machines (SVMs) is a challenging model selection problem, especially when flexible kernels are to be adapted and data are scarce. We present a coherent framework for regularized model selection of 1-norm soft margin SVMs for binary classification. It is proposed to use gradient-ascent on a likelihood function of the hyperparameters. The likelihood function is based on logistic regression for robustly estimating the class conditional probabilities and can be computed efficiently. Overfitting is an important issue in SVM model selection and can be addressed in our framework by incorporating suitable prior distributions over the hyperparameters. We show empirically that gradient-based optimization of the likelihood function is able to adapt multiple kernel parameters and leads to better models than four concurrent state-of-the-art methods.

Toward registration of 3D ultrasound and CT images of the spine in clinical praxis: design and evaluation of a data acquisition protocol

Recent work has demonstrated the accuracy and operational viability of an algorithm proposed by the authors that successfully registers 3-D ultrasound data with CT or MRI data. The successful application of this method to intraoperative navigation, however, depends critically on the quality of the acquired ultrasound data. This gives rise to two questions concerning the usability of the algorithm in clinical praxis. First, how can one guarantee high-quality, user-independent ultrasound registration data with this procedure? Second, can this approach work reliably in clinical practice, namely within the operating theater? To address both of these questions, we present an ultrasound data acquisition protocol that leads the user through the data acquisition process and also provides the criteria to adjust the relevant ultrasound parameters. We also evaluated criteria for the visual inspection of the suitability of the ultrasound data for the registration process. Results for this evaluation show that these visual criteria can be used to decide preoperatively if an ultrasound registration will be successful in a patient. The intraoperative evaluation of the protocol showed that high-quality registrations can be achieved under realistic conditions. This protocol and the visual inspection criteria, together with the ultrasound registration algorithm, provide a surgical team with a means of performing precise, cost-effective navigation in patients for whom a navigated intervention was previously impossible. We evaluated the proposed procedure in clinical practice.

Evolutionary optimization of sequence kernels for detection of bacterial gene starts

Oligo kernels for biological sequence classification have a high discriminative power. A new parameterization for the K-mer oligo kernel is presented, where all oligomers of length K are weighted individually. The task specific choice of these parameters increases the classification performance and reveals information about discriminative features. For adapting the multiple kernel parameters based on cross-validation the covariance matrix adaptation evolution strategy is proposed. It is applied to optimize the trimer oligo kernels for the detection of bacterial gene starts. The resulting kernels lead to higher classification rates, and the adapted parameters reveal the importance of particular triplets for classification, for example of those occurring in the Shine-Dalgarno Sequence.

Second-Order SMO Improves SVM Online and Active Learning

Iterative learning algorithms that approximate the solution of support vector machines (SVMs) have two potential advantages. First, they allow online and active learning. Second, for large data sets, computing the exact SVM solution may be too time-consuming, and an efficient approximation can be preferable. The powerful LASVM iteratively approaches the exact SVM solution using sequential minimal optimization (SMO). It allows efficient online and active learning. Here, this algorithm is considerably improved in speed and accuracy by replacing the working set selection in the SMO steps. A second-order working set selection strategy, which greedily aims at maximizing the progress in each single step, is incorporated.

Gradient-based optimization of kernel-target alignment for sequence kernels applied to bacterial gene start detection

Biological data mining using kernel methods can be improved by a task-specific choice of the kernel function. Oligo kernels for genomic sequence analysis have proven to have a high discriminative power and to provide interpretable results. Oligo kernels that consider subsequences of different lengths can be combined and parameterized to increase their flexibility. For adapting these parameters efficiently, gradient-based optimization of the kernel-target alignment is proposed. The power of this new, general model selection procedure and the benefits of fitting kernels to problem classes are demonstrated by adapting oligo kernels for bacterial gene start detection.

Reducing the number of fitness evaluations in graph genetic programming using a canonical graph indexed database

In this paper we describe the genetic programming system GGP operating on graphs and introduce the notion of graph isomorphisms to explain how they influence the dynamics of GP. It is shown empirically how fitness databases can improve the performance of GP and how mapping graphs to a canonical form can increase these improvements by saving considerable evaluation time.

Covariance matrix adaptation for multi-objective optimization

The covariance matrix adaptation evolution strategy (CMA-ES) is one of the most powerful evolutionary algorithms for real-valued single-objective optimization. In this paper, we develop a variant of the CMA-ES for multi-objective optimization (MOO). We first introduce a single-objective, elitist CMA-ES using plus-selection and step size control based on a success rule. This algorithm is compared to the standard CMA-ES. The elitist CMA-ES turns out to be slightly faster on unimodal functions, but is more prone to getting stuck in sub-optimal local minima. In the new multi-objective CMAES (MO-CMA-ES) a population of individuals that adapt their search strategy as in the elitist CMA-ES is maintained. These are subject to multi-objective selection. The selection is based on non-dominated sorting using either the crowding-distance or the contributing hypervolume as second sorting criterion. Both the elitist single-objective CMA-ES and the MO-CMA-ES inherit important invariance properties, in particular invariance against rotation of the search space, from the original CMA-ES. The benefits of the new MO-CMA-ES in comparison to the well-known NSGA-II and to NSDE, a multi-objective differential evolution algorithm, are experimentally shown.

Gradient-Based Adaptation of General Gaussian Kernels

Gradient-based optimizing of gaussian kernel functions is considered. The gradient for the adaptation of scaling and rotation of the input space is computed to achieve invariance against linear transformations. This is done by using the exponential map as a parameterization of the kernel parameter manifold. By restricting the optimization to a constant trace subspace, the kernel size can be controlled. This is, for example, useful to prevent overfitting when minimizing radius-margin generalization performance measures. The concepts are demonstrated by training hard margin support vector machines on toy data.