Multi input-Multi output 3D CNN for dementia severity assessment with incomplete multimodal data.

Alzheimer's Disease is the most common cause of dementia, whose progression spans in different stages, from very mild cognitive impairment to mild and severe conditions. In clinical trials, Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) are mostly used for the early diagnosis of neurodegenerative disorders since they provide volumetric and metabolic function information of the brain, respectively. In recent years, Deep Learning (DL) has been employed in medical imaging with promising results. Moreover, the use of the deep neural networks, especially Convolutional Neural Networks (CNNs), has also enabled the development of DL-based solutions in domains characterized by the need of leveraging information coming from multiple data sources, raising the Multimodal Deep Learning (MDL). In this paper, we conduct a systematic analysis of MDL approaches for dementia severity assessment exploiting MRI and PET scans. We propose a Multi Input-Multi Output 3D CNN whose training iterations change according to the characteristic of the input as it is able to handle incomplete acquisitions, in which one image modality is missed. Experiments performed on OASIS-3 dataset show the satisfactory results of the implemented network, which outperforms approaches exploiting both single image modality and different MDL fusion techniques.

Tree ring segmentation using UNEt TRansformer neural network on stained microsections for quantitative wood anatomy.

Forests are critical in the terrestrial carbon cycle, and the knowledge of their response to ongoing climate change will be crucial for determining future carbon fluxes and climate trajectories. In areas with contrasting seasons, trees form discrete annual rings that can be assigned to calendar years, allowing to extract valuable information about how trees respond to the environment. The anatomical structure of wood provides highly-resolved information about the reaction and adaptation of trees to climate. Quantitative wood anatomy helps to retrieve this information by measuring wood at the cellular level using high-resolution images of wood micro-sections. However, whereas large advances have been made in identifying cellular structures, obtaining meaningful cellular information is still hampered by the correct annual tree ring delimitation on the images. This is a time-consuming task that requires experienced operators to manually delimit ring boundaries. Classic methods of automatic segmentation based on pixel values are being replaced by new approaches using neural networks which are capable of distinguishing structures, even when demarcations require a high level of expertise. Although neural networks have been used for tree ring segmentation on macroscopic images of wood, the complexity of cell patterns in stained microsections of broadleaved species requires adaptive models to accurately accomplish this task. We present an automatic tree ring boundary delineation using neural networks on stained cross-sectional microsection images from beech cores. We trained a UNETR, a combined neural network of UNET and the attention mechanisms of Visual Transformers, to automatically segment annual ring boundaries. Its accuracy was evaluated considering discrepancies with manual segmentation and the consequences of disparity for the goals of quantitative wood anatomy analyses. In most cases (91.8%), automatic segmentation matched or improved manual segmentation, and the rate of vessels assignment to annual rings was similar between the two categories, even when manual segmentation was considered better. The application of convolutional neural networks-based models outperforms human operator segmentations when confronting ring boundary delimitation using specific parameters for quantitative wood anatomy analysis. Current advances on segmentation models may reduce the cost of massive and accurate data collection for quantitative wood anatomy.

Mapping Tree Canopy in Urban Environments Using Point Clouds from Airborne Laser Scanning and Street Level Imagery.

Resilient cities incorporate a social, ecological, and technological systems perspective through their trees, both in urban and peri-urban forests and linear street trees, and help promote and understand the concept of ecosystem resilience. Urban tree inventories usually involve the collection of field data on the location, genus, species, crown shape and volume, diameter, height, and health status of these trees. In this work, we have developed a multi-stage methodology to update urban tree inventories in a fully automatic way, and we have applied it in the city of Pamplona (Spain). We have compared and combined two of the most common data sources for updating urban tree inventories: Airborne Laser Scanning (ALS) point clouds combined with aerial orthophotographs, and street-level imagery from Google Street View (GSV). Depending on the data source, different methodologies were used to identify the trees. In the first stage, the use of individual tree detection techniques in ALS point clouds was compared with the detection of objects (trees) on street level images using computer vision (CV) techniques. In both cases, a high success rate or recall (number of true positive with respect to all detectable trees) was obtained, where between 85.07% and 86.42% of the trees were well-identified, although many false positives (FPs) or trees that did not exist or that had been confused with other objects were always identified. In order to reduce these errors or FPs, a second stage was designed, where FP debugging was performed through two methodologies: (a) based on the automatic checking of all possible trees with street level images, and (b) through a machine learning binary classification model trained with spectral data from orthophotographs. After this second stage, the recall decreased to about 75% (between 71.43 and 78.18 depending on the procedure used) but most of the false positives were eliminated. The results obtained with both data sources were robust and accurate. We can conclude that the results obtained with the different methodologies are very similar, where the main difference resides in the access to the starting information. While the use of street-level images only allows for the detection of trees growing in trafficable streets and is a source of information that is usually paid for, the use of ALS and aerial orthophotographs allows for the location of trees anywhere in the city, including public and private parks and gardens, and in many countries, these data are freely available.

Automatic detection of pneumonia in chest X-ray images using textural features.

Fast and accurate diagnosis is critical for the triage and management of pneumonia, particularly in the current scenario of a COVID-19 pandemic, where this pathology is a major symptom of the infection. With the objective of providing tools for that purpose, this study assesses the potential of three textural image characterisation methods: radiomics, fractal dimension and the recently developed superpixel-based histon, as biomarkers to be used for training Artificial Intelligence (AI) models in order to detect pneumonia in chest X-ray images. Models generated from three different AI algorithms have been studied: K-Nearest Neighbors, Support Vector Machine and Random Forest. Two open-access image datasets were used in this study. In the first one, a dataset composed of paediatric chest X-ray, the best performing generated models achieved an 83.3% accuracy with 89% sensitivity for radiomics, 89.9% accuracy with 93.6% sensitivity for fractal dimension and 91.3% accuracy with 90.5% sensitivity for superpixels based histon. Second, a dataset derived from an image repository developed primarily as a tool for studying COVID-19 was used. For this dataset, the best performing generated models resulted in a 95.3% accuracy with 99.2% sensitivity for radiomics, 99% accuracy with 100% sensitivity for fractal dimension and 99% accuracy with 98.6% sensitivity for superpixel-based histons. The results confirm the validity of the tested methods as reliable and easy-to-implement automatic diagnostic tools for pneumonia.

Satellite data and machine learning reveal the incidence of late frost defoliations on Iberian beech forests.

Climate warming is driving an advance of leaf unfolding date in temperate deciduous forests, promoting longer growing seasons and higher carbon gains. However, an earlier leaf phenology also increases the risk of late frost defoliation (LFD) events. Compiling the spatiotemporal patterns of defoliations caused by spring frost events is critical to unveil whether the balance between the current advance in leaf unfolding dates and the frequency of LFD occurrence is changing and represents a threaten for the future viability and persistence of deciduous forests. We combined satellite imagery with machine learning techniques to reconstruct the spatiotemporal patterns of LFD events for the 2003-2018 period in the Iberian range of European beech (Fagus sylvatica), at the drier distribution edge of the species. We used MODIS Vegetation Index Products to generate a Normalized Difference Vegetation Index (NDVI) time series for each 250 × 250 m pixel in a total area of 1,013 km2 (16,218 pixels). A semi-supervised approach was used to train a machine learning model, in which a binary classifier called Support Vector Machine with Global Alignment Kernel was used to differentiate between late frost and non-late frost pixels. We verified the obtained estimates with photointerpretation and existing beech tree-ring chronologies to iteratively improve the model. Then, we used the model output to identify topographical and climatic factors that determined the spatial incidence of LFD. During the study period, LFD was a low recurrence phenomenon that occurred every 15.2 yr on average and showed high spatiotemporal heterogeneity. Most LFD events were condensed in 5 yr and clustered in western forests (86.5% in one-fifth of the pixels) located at high elevation with lower than average precipitation. Elevation and longitude were the major LFD risk factors, followed by annual precipitation. The synergistic effects of increasing drought intensity and rising temperature combined with more frequent late frost events may determine the future performance and distribution of beech forests. This interaction might be critical at the beech drier range edge, where the concentration of LFD at high elevations could constrain beech altitudinal shifts and/or favor species with higher resistance to late frosts.

Sample-Size Determination Methodologies for Machine Learning in Medical Imaging Research: A Systematic Review.

PURPOSE: The required training sample size for a particular machine learning (ML) model applied to medical imaging data is often unknown. The purpose of this study was to provide a descriptive review of current sample-size determination methodologies in ML applied to medical imaging and to propose recommendations for future work in the field. METHODS: We conducted a systematic literature search of articles using Medline and Embase with keywords including "machine learning," "image," and "sample size." The search included articles published between 1946 and 2018. Data regarding the ML task, sample size, and train-test pipeline were collected. RESULTS: A total of 167 articles were identified, of which 22 were included for qualitative analysis. There were only 4 studies that discussed sample-size determination methodologies, and 18 that tested the effect of sample size on model performance as part of an exploratory analysis. The observed methods could be categorized as pre hoc model-based approaches, which relied on features of the algorithm, or post hoc curve-fitting approaches requiring empirical testing to model and extrapolate algorithm performance as a function of sample size. Between studies, we observed great variability in performance testing procedures used for curve-fitting, model assessment methods, and reporting of confidence in sample sizes. CONCLUSIONS: Our study highlights the scarcity of research in training set size determination methodologies applied to ML in medical imaging, emphasizes the need to standardize current reporting practices, and guides future work in development and streamlining of pre hoc and post hoc sample size approaches.

Supervoxels-Based Histon as a New Alzheimer's Disease Imaging Biomarker.

Alzheimer's disease (AD) represents the prevalent type of dementia in the elderly, and is characterized by the presence of neurofibrillary tangles and amyloid plaques that eventually leads to the loss of neurons, resulting in atrophy in specific brain areas. Although the process of degeneration can be visualized through various modalities of medical imaging and has proved to be a valuable biomarker, the accurate diagnosis of Alzheimer's disease remains a challenge, especially in its early stages. In this paper, we propose a novel classification method for Alzheimer's disease/cognitive normal discrimination in structural magnetic resonance images (MRI), based on the extension of the concept of histons to volumetric images. The proposed method exploits the relationship between grey matter, white matter and cerebrospinal fluid degeneration by means of a segmentation using supervoxels. The calculated histons are then processed for a reduction in dimensionality using principal components analysis (PCA) and the resulting vector is used to train an support vector machine (SVM) classifier. Experimental results using the OASIS-1 database have proven to be a significant improvement compared to a baseline classification made using the pipeline provided by Clinica software.

Fusion of High Resolution Multispectral Imagery in Vulnerable Coastal and Land Ecosystems.

Ecosystems provide a wide variety of useful resources that enhance human welfare, but these resources are declining due to climate change and anthropogenic pressure. In this work, three vulnerable ecosystems, including shrublands, coastal areas with dunes systems and areas of shallow water, are studied. As far as these resources' reduction is concerned, remote sensing and image processing techniques could contribute to the management of these natural resources in a practical and cost-effective way, although some improvements are needed for obtaining a higher quality of the information available. An important quality improvement is the fusion at the pixel level. Hence, the objective of this work is to assess which pansharpening technique provides the best fused image for the different types of ecosystems. After a preliminary evaluation of twelve classic and novel fusion algorithms, a total of four pansharpening algorithms was analyzed using six quality indices. The quality assessment was implemented not only for the whole set of multispectral bands, but also for the subset of spectral bands covered by the wavelength range of the panchromatic image and outside of it. A better quality result is observed in the fused image using only the bands covered by the panchromatic band range. It is important to highlight the use of these techniques not only in land and urban areas, but a novel analysis in areas of shallow water ecosystems. Although the algorithms do not show a high difference in land and coastal areas, coastal ecosystems require simpler algorithms, such as fast intensity hue saturation, whereas more heterogeneous ecosystems need advanced algorithms, as weighted wavelet 'à trous' through fractal dimension maps for shrublands and mixed ecosystems. Moreover, quality map analysis was carried out in order to study the fusion result in each band at the local level. Finally, to demonstrate the performance of these pansharpening techniques, advanced Object-Based (OBIA) support vector machine classification was applied, and a thematic map for the shrubland ecosystem was obtained, which corroborates wavelet 'à trous' through fractal dimension maps as the best fusion algorithm for this ecosystem.