CryoSegNet: accurate cryo-EM protein particle picking by integrating the foundational AI image segmentation model and attention-gated U-Net.

Picking protein particles in cryo-electron microscopy (cryo-EM) micrographs is a crucial step in the cryo-EM-based structure determination. However, existing methods trained on a limited amount of cryo-EM data still cannot accurately pick protein particles from noisy cryo-EM images. The general foundational artificial intelligence-based image segmentation model such as Meta's Segment Anything Model (SAM) cannot segment protein particles well because their training data do not include cryo-EM images. Here, we present a novel approach (CryoSegNet) of integrating an attention-gated U-shape network (U-Net) specially designed and trained for cryo-EM particle picking and the SAM. The U-Net is first trained on a large cryo-EM image dataset and then used to generate input from original cryo-EM images for SAM to make particle pickings. CryoSegNet shows both high precision and recall in segmenting protein particles from cryo-EM micrographs, irrespective of protein type, shape and size. On several independent datasets of various protein types, CryoSegNet outperforms two top machine learning particle pickers crYOLO and Topaz as well as SAM itself. The average resolution of density maps reconstructed from the particles picked by CryoSegNet is 3.33 Å, 7% better than 3.58 Å of Topaz and 14% better than 3.87 Å of crYOLO. It is publicly available at https://github.com/jianlin-cheng/CryoSegNet.

CRCS: An automatic image processing pipeline for hormone level analysis of Cushing's disease.

Due to the abnormal secretion of adreno-cortico-tropic-hormone (ACTH) by tumors, Cushing's disease leads to hypercortisonemia, a precursor to a series of metabolic disorders and serious complications. Cushing's disease has high recurrence rate, short recurrence time and undiscovered recurrence reason after surgical resection. Qualitative or quantitative automatic image analysis of histology images can potentially in providing insights into Cushing's disease, but still no software has been available to the best of our knowledge. In this study, we propose a quantitative image analysis-based pipeline CRCS, which aims to explore the relationship between the expression level of ACTH in normal cell tissues adjacent to tumor cells and the postoperative prognosis of patients. CRCS mainly consists of image-level clustering, cluster-level multi-modal image registration, patch-level image classification and pixel-level image segmentation on the whole slide imaging (WSI). On both image registration and classification tasks, our method CRCS achieves state-of-the-art performance compared to recently published methods on our collected benchmark dataset. In addition, CRCS achieves an accuracy of 0.83 for postoperative prognosis of 12 cases. CRCS demonstrates great potential for instrumenting automatic diagnosis and treatment for Cushing's disease.

MFNet: Meta-learning based on frequency-space mix for MRI segmentation in nasopharyngeal carcinoma.

Deep learning techniques have been applied to medical image segmentation and demonstrated expert-level performance. Due to the poor generalization abilities of the models in the deployment in different centres, common solutions, such as transfer learning and domain adaptation techniques, have been proposed to mitigate this issue. However, these solutions necessitate retraining the models with target domain data and annotations, which limits their deployment in clinical settings in unseen domains. We evaluated the performance of domain generalization methods on the task of MRI segmentation of nasopharyngeal carcinoma (NPC) by collecting a new dataset of 321 patients with manually annotated MRIs from two hospitals. We transformed the modalities of MRI, including T1WI, T2WI and CE-T1WI, from the spatial domain to the frequency domain using Fourier transform. To address the bottleneck of domain generalization in MRI segmentation of NPC, we propose a meta-learning approach based on frequency domain feature mixing. We evaluated the performance of MFNet against existing techniques for generalizing NPC segmentation in terms of Dice and MIoU. Our method evidently outperforms the baseline in handling the generalization of NPC segmentation. The MF-Net clearly demonstrates its effectiveness for generalizing NPC MRI segmentation to unseen domains (Dice = 67.59%, MIoU = 75.74% T1W1). MFNet enhances the model's generalization capabilities by incorporating mixed-feature meta-learning. Our approach offers a novel perspective to tackle the domain generalization problem in the field of medical imaging by effectively exploiting the unique characteristics of medical images.

Computational framework for the generation of one-dimensional vascular models accounting for uncertainty in networks extracted from medical images.

One-dimensional (1D) cardiovascular models offer a non-invasive method to answer medical questions, including predictions of wave-reflection, shear stress, functional flow reserve, vascular resistance and compliance. This model type can predict patient-specific outcomes by solving 1D fluid dynamics equations in geometric networks extracted from medical images. However, the inherent uncertainty in in vivo imaging introduces variability in network size and vessel dimensions, affecting haemodynamic predictions. Understanding the influence of variation in image-derived properties is essential to assess the fidelity of model predictions. Numerous programs exist to render three-dimensional surfaces and construct vessel centrelines. Still, there is no exact way to generate vascular trees from the centrelines while accounting for uncertainty in data. This study introduces an innovative framework employing statistical change point analysis to generate labelled trees that encode vessel dimensions and their associated uncertainty from medical images. To test this framework, we explore the impact of uncertainty in 1D haemodynamic predictions in a systemic and pulmonary arterial network. Simulations explore haemodynamic variations resulting from changes in vessel dimensions and segmentation; the latter is achieved by analysing multiple segmentations of the same images. Results demonstrate the importance of accurately defining vessel radii and lengths when generating high-fidelity patient-specific haemodynamics models. KEY POINTS: This study introduces novel algorithms for generating labelled directed trees from medical images, focusing on accurate junction node placement and radius extraction using change points to provide haemodynamic predictions with uncertainty within expected measurement error. Geometric features, such as vessel dimension (length and radius) and network size, significantly impact pressure and flow predictions in both pulmonary and aortic arterial networks. Standardizing networks to a consistent number of vessels is crucial for meaningful comparisons and decreases haemodynamic uncertainty. Change points are valuable to understanding structural transitions in vascular data, providing an automated and efficient way to detect shifts in vessel characteristics and ensure reliable extraction of representative vessel radii.

CalciSeg: A versatile approach for unsupervised segmentation of calcium imaging data.

Recent advances in calcium imaging, including the development of fast and sensitive genetically encoded indicators, high-resolution camera chips for wide-field imaging, and resonant scanning mirrors in laser scanning microscopy, have notably improved the temporal and spatial resolution of functional imaging analysis. Nonetheless, the variability of imaging approaches and brain structures challenges the development of versatile and reliable segmentation methods. Standard techniques, such as manual selection of regions of interest or machine learning solutions, often fall short due to either user bias, non-transferability among systems, or computational demand. To overcome these issues, we developed CalciSeg, a data-driven and reproducible approach for unsupervised functional calcium imaging data segmentation. CalciSeg addresses the challenges associated with brain structure variability and user bias by offering a computationally efficient solution for automatic image segmentation based on two parameters: regions' size limits and number of refinement iterations. We evaluated CalciSeg efficacy on datasets of varied complexity, different insect species (locusts, bees, and cockroaches), and imaging systems (wide-field, confocal, and multiphoton), showing the robustness and generality of our approach. Finally, the user-friendly nature and open-source availability of CalciSeg facilitate the integration of this algorithm into existing analysis pipelines.

Light3DHS: A lightweight 3D hippocampus segmentation method using multiscale convolution attention and vision transformer.

The morphological analysis and volume measurement of the hippocampus are crucial to the study of many brain diseases. Therefore, an accurate hippocampal segmentation method is beneficial for the development of clinical research in brain diseases. U-Net and its variants have become prevalent in hippocampus segmentation of Magnetic Resonance Imaging (MRI) due to their effectiveness, and the architecture based on Transformer has also received some attention. However, some existing methods focus too much on the shape and volume of the hippocampus rather than its spatial information, and the extracted information is independent of each other, ignoring the correlation between local and global features. In addition, many methods cannot be effectively applied to practical medical image segmentation due to many parameters and high computational complexity. To this end, we combined the advantages of CNNs and ViTs (Vision Transformer) and proposed a simple and lightweight model: Light3DHS for the segmentation of the 3D hippocampus. In order to obtain richer local contextual features, the encoder first utilizes a multi-scale convolutional attention module (MCA) to learn the spatial information of the hippocampus. Considering the importance of local features and global semantics for 3D segmentation, we used a lightweight ViT to learn high-level features of scale invariance and further fuse local-to-global representation. To evaluate the effectiveness of encoder feature representation, we designed three decoders of different complexity to generate segmentation maps. Experiments on three common hippocampal datasets demonstrate that the network achieves more accurate hippocampus segmentation with fewer parameters. Light3DHS performs better than other state-of-the-art algorithms.

Systematic review and meta-analysis of automated methods for quantifying enlarged perivascular spaces in the brain.

Research into magnetic resonance imaging (MRI)-visible perivascular spaces (PVS) has recently increased, as results from studies in different diseases and populations are cementing their association with sleep, disease phenotypes, and overall health indicators. With the establishment of worldwide consortia and the availability of large databases, computational methods that allow to automatically process all this wealth of information are becoming increasingly relevant. Several computational approaches have been proposed to assess PVS from MRI, and efforts have been made to summarise and appraise the most widely applied ones. We systematically reviewed and meta-analysed all publications available up to September 2023 describing the development, improvement, or application of computational PVS quantification methods from MRI. We analysed 67 approaches and 60 applications of their implementation, from 112 publications. The two most widely applied were the use of a morphological filter to enhance PVS-like structures, with Frangi being the choice preferred by most, and the use of a U-Net configuration with or without residual connections. Older adults or population studies comprising adults from 18 years old onwards were, overall, more frequent than studies using clinical samples. PVS were mainly assessed from T2-weighted MRI acquired in 1.5T and/or 3T scanners, although combinations using it with T1-weighted and FLAIR images were also abundant. Common associations researched included age, sex, hypertension, diabetes, white matter hyperintensities, sleep and cognition, with occupation-related, ethnicity, and genetic/hereditable traits being also explored. Despite promising improvements to overcome barriers such as noise and differentiation from other confounds, a need for joined efforts for a wider testing and increasing availability of the most promising methods is now paramount.

recolorize: An R package for flexible colour segmentation of biological images.

Colour pattern variation provides biological information in fields ranging from disease ecology to speciation dynamics. Comparing colour pattern geometries across images requires colour segmentation, where pixels in an image are assigned to one of a set of colour classes shared by all images. Manual methods for colour segmentation are slow and subjective, while automated methods can struggle with high technical variation in aggregate image sets. We present recolorize, an R package toolbox for human-subjective colour segmentation with functions for batch-processing low-variation image sets and additional tools for handling images from diverse (high-variation) sources. The package also includes export options for a variety of formats and colour analysis packages. This paper illustrates recolorize for three example datasets, including high variation, batch processing and combining with reflectance spectra, and demonstrates the downstream use of methods that rely on this output.

IGZO/WO3-x-Heterostructured Artificial Optoelectronic Synaptic Devices Mimicking Image Segmentation and Motion Capture.

Currently, artificial neural networks (ANNs) based on memristors are limited to recognizing static images of objects when simulating human visual system, preventing them from performing high-dimensional information perception, and achieving more complex biomimetic functions is subject to certain limitations. In this work, indium gallium zinc oxide (IGZO)/tungsten oxide (WO3-x)-heterostructured artificial optoelectronic synaptic devices mimicking image segmentation and motion capture exhibiting high-performance optoelectronic synaptic responses are proposed and demonstrated. Upon electrical and optical stimulations, the device shows a variety of fundamental and advanced electrical and optical synaptic plasticity. Most importantly, outstanding and repeatable linear synaptic weight changes are attained by the developed memristor. By taking advantage of the notable linear synaptic weight changes, ANNs have been constructed and successfully utilized to demonstrate two applications in the field of computer vision, including image segmentation and object tracking. The accuracy attained by the memristor-based ANNs is similar to that of the computer algorithms, while its power has been significantly reduced by 105 orders of magnitude. With successful emulations of the human brain reactions when observing objects, the demonstrated memristor and related ANNs can be effectively utilized in constructing artificial optoelectronic synaptic devices and show promising potential in emulating human visual perception.

Deep learning to overcome Zernike phase-contrast nanoCT artifacts for automated micro-nano porosity segmentation in bone.

Bone material contains a hierarchical network of micro- and nano-cavities and channels, known as the lacuna-canalicular network (LCN), that is thought to play an important role in mechanobiology and turnover. The LCN comprises micrometer-sized lacunae, voids that house osteocytes, and submicrometer-sized canaliculi that connect bone cells. Characterization of this network in three dimensions is crucial for many bone studies. To quantify X-ray Zernike phase-contrast nanotomography data, deep learning is used to isolate and assess porosity in artifact-laden tomographies of zebrafish bones. A technical solution is proposed to overcome the halo and shade-off domains in order to reliably obtain the distribution and morphology of the LCN in the tomographic data. Convolutional neural network (CNN) models are utilized with increasing numbers of images, repeatedly validated by `error loss' and `accuracy' metrics. U-Net and Sensor3D CNN models were trained on data obtained from two different synchrotron Zernike phase-contrast transmission X-ray microscopes, the ANATOMIX beamline at SOLEIL (Paris, France) and the P05 beamline at PETRA III (Hamburg, Germany). The Sensor3D CNN model with a smaller batch size of 32 and a training data size of 70 images showed the best performance (accuracy 0.983 and error loss 0.032). The analysis procedures, validated by comparison with human-identified ground-truth images, correctly identified the voids within the bone matrix. This proposed approach may have further application to classify structures in volumetric images that contain non-linear artifacts that degrade image quality and hinder feature identification.

Generative Adversarial Networks in Digital Histopathology: Current Applications, Limitations, Ethical Considerations, and Future Directions.

Generative adversarial networks (GANs) have gained significant attention in the field of image synthesis, particularly in computer vision. GANs consist of a generative model and a discriminative model trained in an adversarial setting to generate realistic and novel data. In the context of image synthesis, the generator produces synthetic images, whereas the discriminator determines their authenticity by comparing them with real examples. Through iterative training, the generator allows the creation of images that are indistinguishable from real ones, leading to high-quality image generation. Considering their success in computer vision, GANs hold great potential for medical diagnostic applications. In the medical field, GANs can generate images of rare diseases, aid in learning, and be used as visualization tools. GANs can leverage unlabeled medical images, which are large in size, numerous in quantity, and challenging to annotate manually. GANs have demonstrated remarkable capabilities in image synthesis and have the potential to significantly impact digital histopathology. This review article focuses on the emerging use of GANs in digital histopathology, examining their applications and potential challenges. Histopathology plays a crucial role in disease diagnosis, and GANs can contribute by generating realistic microscopic images. However, ethical considerations arise because of the reliance on synthetic or pseudogenerated images. Therefore, the manuscript also explores the current limitations and highlights the ethical considerations associated with the use of this technology. In conclusion, digital histopathology has seen an emerging use of GANs for image enhancement, such as color (stain) normalization, virtual staining, and ink/marker removal. GANs offer significant potential in transforming digital pathology when applied to specific and narrow tasks (preprocessing enhancements). Evaluating data quality, addressing biases, protecting privacy, ensuring accountability and transparency, and developing regulation are imperative to ensure the ethical application of GANs.

An immunofluorescence-guided segmentation model in H&E images is enabled by tissue artifact correction by CycleGAN.

Despite recent advances, the adoption of computer vision methods into clinical and commercial applications has been hampered by the limited availability of accurate ground truth tissue annotations required to train robust supervised models. Generating such ground truth can be accelerated by annotating tissue molecularly using immunofluorescence staining (IF) and mapping these annotations to a post-IF H&E (terminal H&E). Mapping the annotations between the IF and the terminal H&E increases both the scale and accuracy by which ground truth could be generated. However, discrepancies between terminal H&E and conventional H&E caused by IF tissue processing have limited this implementation. We sought to overcome this challenge and achieve compatibility between these parallel modalities using synthetic image generation, in which a cycle-consistent generative adversarial network (CycleGAN) was applied to transfer the appearance of conventional H&E such that it emulates the terminal H&E. These synthetic emulations allowed us to train a deep learning (DL) model for the segmentation of epithelium in the terminal H&E that could be validated against the IF staining of epithelial-based cytokeratins. The combination of this segmentation model with the CycleGAN stain transfer model enabled performative epithelium segmentation in conventional H&E images. The approach demonstrates that the training of accurate segmentation models for the breadth of conventional H&E data can be executed free of human-expert annotations by leveraging molecular annotation strategies such as IF, so long as the tissue impacts of the molecular annotation protocol are captured by generative models that can be deployed prior to the segmentation process.

Anterior cruciate ligament injury and age affect knee cartilage T2 but not thickness.

OBJECTIVE: To investigate the effect of unilateral anterior cruciate ligament (ACL) injury on cartilage thickness and composition, specifically laminar transverse relaxation time (T2) by magnetic resonance imaging (MRI), in younger and older participants and to compare within-person side differences in these parameters between ACL-injured and healthy controls. DESIGN: Quantitative double-echo steady-state 3 Tesla MRI-sequences were acquired in both knees of 85 participants in four groups: 20-30 years: healthy, HEA20-30, n = 24; ACL-injured, ACL20-30, n = 23; 40-60 years: healthy, HEA40-60, n = 24; ACL-injured, ACL40-60, n = 14 (ACL injury 2-10 years prior to study inclusion). Weight-bearing femorotibial cartilages were manually segmented; cartilage T2 and thickness were computed using custom software. Mean and side differences in subregional cartilage thickness, superficial and deep cartilage T2 were compared within and between groups using non-parametric statistics. RESULTS: Cartilage thickness did not differ within or between groups. Only the side difference in medial femorotibial cartilage thickness was greater in ACL20-30 than in HEA20-30. Deep zone T2 was longer in the ACL-injured than in the contralateral uninjured knees and than in healthy controls, especially in the lateral compartment. Most ACL-injured participants had side differences in femorotibial deep zone T2 above the threshold derived from controls. CONCLUSION: In the ACL-injured knee, early compositional differences in femorotibial cartilage (T2) appear to occur in the deep zone and precede cartilage thickness loss. These results suggest that monitoring laminar T2 after ACL injury may be useful in diagnosing and monitoring early articular cartilage changes.

A Lightweight Convolutional Neural Network Based on Dynamic Level-Set Loss Function for Spine MR Image Segmentation.

BACKGROUND: Spine MR image segmentation is important foundation for computer-aided diagnostic (CAD) algorithms of spine disorders. Convolutional neural networks segment effectively, but require high computational costs. PURPOSE: To design a lightweight model based on dynamic level-set loss function for high segmentation performance. STUDY TYPE: Retrospective. POPULATION: Four hundred forty-eight subjects (3163 images) from two separate datasets. Dataset-1: 276 subjects/994 images (53.26% female, mean age 49.02 ± 14.09), all for disc degeneration screening, 188 had disc degeneration, 67 had herniated disc. Dataset-2: public dataset with 172 subjects/2169 images, 142 patients with vertebral degeneration, 163 patients with disc degeneration. FIELD STRENGTH/SEQUENCE: T2 weighted turbo spin echo sequences at 3T. ASSESSMENT: Dynamic Level-set Net (DLS-Net) was compared with four mainstream (including U-net++) and four lightweight models, and manual label made by five radiologists (vertebrae, discs, spinal fluid) used as segmentation evaluation standard. Five-fold cross-validation are used for all experiments. Based on segmentation, a CAD algorithm of lumbar disc was designed for assessing DLS-Net's practicality, and the text annotation (normal, bulging, or herniated) from medical history data were used as evaluation standard. STATISTICAL TESTS: All segmentation models were evaluated with DSC, accuracy, precision, and AUC. The pixel numbers of segmented results were compared with manual label using paired t-tests, with P < 0.05 indicating significance. The CAD algorithm was evaluated with accuracy of lumbar disc diagnosis. RESULTS: With only 1.48% parameters of U-net++, DLS-Net achieved similar accuracy in both datasets (Dataset-1: DSC 0.88 vs. 0.89, AUC 0.94 vs. 0.94; Dataset-2: DSC 0.86 vs. 0.86, AUC 0.93 vs. 0.93). The segmentation results of DLS-Net showed no significant differences with manual labels in pixel numbers for discs (Dataset-1: 1603.30 vs. 1588.77, P = 0.22; Dataset-2: 863.61 vs. 886.4, P = 0.14) and vertebrae (Dataset-1: 3984.28 vs. 3961.94, P = 0.38; Dataset-2: 4806.91 vs. 4732.85, P = 0.21). Based on DLS-Net's segmentation results, the CAD algorithm achieved higher accuracy than using non-cropped MR images (87.47% vs. 61.82%). DATA CONCLUSION: The proposed DLS-Net has fewer parameters but achieves similar accuracy to U-net++, helps CAD algorithm achieve higher accuracy, which facilitates wider application. EVIDENCE LEVEL: 2 TECHNICAL EFFICACY: Stage 1.

A multiangle polarised imaging-based method for thin section segmentation.

The most crucial task of petroleum geology is to explore oil and gas reservoirs in the deep underground. As one of the analysis techniques in petroleum geological research, rock thin section identification method includes particle segmentation, which is one of the key steps. A conventional sandstone thin section image typically contains hundreds of mineral particles with blurred boundaries and complex microstructures inside the particles. Moreover, the complex lithology and low porosity of tight sandstone make traditional image segmentation methods unsuitable for solving the complex thin section segmentation problems. This paper combines petrology and image processing technologies. First, polarised sequence images are aligned, and then the images are transformed to the HSV colour space to extract pores. Second, particles are extracted according to their extinction characteristics. Last, a concavity and corner detection matching method is used to process the extracted particles, thereby completing the segmentation of sandstone thin section images. The experimental results show that our proposed method can more accurately fit the boundaries of mineral particles in sandstone images than existing image segmentation methods. Additionally, when applied in actual production scenarios, our method exhibits excellent performance, greatly improving thin section identification efficiency and significantly assisting experts in identification.

Highly accurate and effective deep neural networks in pathological diagnosis of prostate cancer.

PURPOSE: To established an AI system to make the pathological diagnosis of prostate cancer. METHODS: Prostate histopathological whole mount (WM) sections from patients underwent robot-assisted laparoscopic prostatectomy were prepared. All the prostate WM pathological sections were converted to digital image data and marked with different colors on the basis of the ISUP Gleason grade group. The image was then fed into a segmentation algorithm. We chose modified U-Net as our fundamental network architecture. RESULTS: 172 patients were involved in this study. 896 pieces of prostate WM pathological sections from 160 patients, in which 826 pieces of WM sections from 148 patients were assigned to the training set randomly. After image segmentation there were totally 2,138,895 patches, of which 1,646,535 patches were valid for training. The other WM section was arranged for testing. Based on the whole image testing, AI and pathologists presented the same answers among 21 of 22 pieces of sections. To evaluate the diagnostic results at the pixel level, we anticipated correct cancer or non-cancer diagnose from this AI system. The area under the ROC curve as 96.8%. The value of pixel accuracy of three methods (binary analysis, clinically oriented analysis and analysis for different ISUP Gleason grade) were 96.93%, 95.43% and 93.88%, respectively. The value of frequency weighted IoU were 94.32%, 92.13% and 90.21%, respectively. CONCLUSIONS: This AI system is able to assist pathologists to make a final diagnosis, indicating the great potential and a wide-range of applications of AI in the medical field.

Segmentation of hyphae and yeast in fungi-infected tissue slice images and its application in analyzing antifungal blue light therapy.

Candida albicans is a pathogenic fungus that undergoes morphological transitions between hyphal and yeast forms, adapting to diverse environmental stimuli and exhibiting distinct virulence. Existing research works on antifungal blue light (ABL) therapy have either focused solely on hyphae or neglected to differentiate between morphologies, obscuring potential differential effects. To address this gap, we established a novel dataset of 150 C. albicans-infected mouse skin tissue slice images with meticulously annotated hyphae and yeast. Eleven representative convolutional neural networks were trained and evaluated on this dataset using seven metrics to identify the optimal model for segmenting hyphae and yeast in original high pixel size images. Leveraging the segmentation results, we analyzed the differential impact of blue light on the invasion depth and density of both morphologies within the skin tissue. U-Net-BN outperformed other models in segmentation accuracy, achieving the best overall performance. While both hyphae and yeast exhibited significant reductions in invasion depth and density at the highest ABL dose (180 J/cm2), only yeast was significantly inhibited at the lower dose (135 J/cm2). This novel finding emphasizes the importance of developing more effective treatment strategies for both morphologies.

We studied the effects of blue light therapy on hyphal and yeast forms of Candida albicans. Through image segmentation techniques, we discovered that the changes in invasion depth and density differed between these two forms after exposure to blue light.

Evaluating artificial intelligence for comparative radiography.

INTRODUCTION: Comparative radiography is a forensic identification and shortlisting technique based on the comparison of skeletal structures in ante-mortem and post-mortem images. The images (e.g., 2D radiographs or 3D computed tomographies) are manually superimposed and visually compared by a forensic practitioner. It requires a significant amount of time per comparison, limiting its utility in large comparison scenarios. METHODS: We propose and validate a novel framework for automating the shortlisting of candidates using artificial intelligence. It is composed of (1) a segmentation method to delimit skeletal structures' silhouettes in radiographs, (2) a superposition method to generate the best simulated "radiographs" from 3D images according to the segmented radiographs, and (3) a decision-making method for shortlisting all candidates ranked according to a similarity metric. MATERIAL: The dataset is composed of 180 computed tomographies and 180 radiographs where the frontal sinuses are visible. Frontal sinuses are the skeletal structure analyzed due to their high individualization capability. RESULTS: Firstly, we validate two deep learning-based techniques for segmenting the frontal sinuses in radiographs, obtaining high-quality results. Secondly, we study the framework's shortlisting capability using both automatic segmentations and superimpositions. The obtained superimpositions, based only on the superimposition metric, allowed us to filter out 40% of the possible candidates in a completely automatic manner. Thirdly, we perform a reliability study by comparing 180 radiographs against 180 computed tomographies using manual segmentations. The results allowed us to filter out 73% of the possible candidates. Furthermore, the results are robust to inter- and intra-expert-related errors.

Improved Robustness for Deep Learning-based Segmentation of Multi-Center Myocardial Perfusion MRI Datasets Using Data Adaptive Uncertainty-guided Space-time Analysis.

BACKGROUND: Fully automatic analysis of myocardial perfusion MRI datasets enables rapid and objective reporting of stress/rest studies in patients with suspected ischemic heart disease. Developing deep learning techniques that can analyze multi-center datasets despite limited training data and variations in software (pulse sequence) and hardware (scanner vendor) is an ongoing challenge. METHODS: Datasets from 3 medical centers acquired at 3T (n = 150 subjects; 21,150 first-pass images) were included: an internal dataset (inD; n = 95) and two external datasets (exDs; n = 55) used for evaluating the robustness of the trained deep neural network (DNN) models against differences in pulse sequence (exD-1) and scanner vendor (exD-2). A subset of inD (n = 85) was used for training/validation of a pool of DNNs for segmentation, all using the same spatiotemporal U-Net architecture and hyperparameters but with different parameter initializations. We employed a space-time sliding-patch analysis approach that automatically yields a pixel-wise "uncertainty map" as a byproduct of the segmentation process. In our approach, dubbed Data Adaptive Uncertainty-Guided Space-time (DAUGS) analysis, a given test case is segmented by all members of the DNN pool and the resulting uncertainty maps are leveraged to automatically select the "best" one among the pool of solutions. For comparison, we also trained a DNN using the established approach with the same settings (hyperparameters, data augmentation, etc.). RESULTS: The proposed DAUGS analysis approach performed similarly to the established approach on the internal dataset (Dice score for the testing subset of inD: 0.896 ± 0.050 vs. 0.890 ± 0.049; p = n.s.) whereas it significantly outperformed on the external datasets (Dice for exD-1: 0.885 ± 0.040 vs. 0.849 ± 0.065, p < 0.005; Dice for exD-2: 0.811 ± 0.070 vs. 0.728 ± 0.149, p < 0.005). Moreover, the number of image series with "failed" segmentation (defined as having myocardial contours that include bloodpool or are noncontiguous in ≥1 segment) was significantly lower for the proposed vs. the established approach (4.3% vs. 17.1%, p < 0.0005). CONCLUSIONS: The proposed DAUGS analysis approach has the potential to improve the robustness of deep learning methods for segmentation of multi-center stress perfusion datasets with variations in the choice of pulse sequence, site location or scanner vendor.

Land Resources for Wind Energy Development Requires Regionalized Characterizations.

Estimates of the land area occupied by wind energy differ by orders of magnitude due to data scarcity and inconsistent methodology. We developed a method that combines machine learning-based imagery analysis and geographic information systems and examined the land area of 318 wind farms (15,871 turbines) in the U.S. portion of the Western Interconnection. We found that prior land use and human modification in the project area are critical for land-use efficiency and land transformation of wind projects. Projects developed in areas with little human modification have a land-use efficiency of 63.8 ± 8.9 W/m2 (mean ±95% confidence interval) and a land transformation of 0.24 ± 0.07 m2/MWh, while values for projects in areas with high human modification are 447 ± 49.4 W/m2 and 0.05 ± 0.01 m2/MWh, respectively. We show that land resources for wind can be quantified consistently with our replicable method, a method that obviates >99% of the workload using machine learning. To quantify the peripheral impact of a turbine, buffered geometry can be used as a proxy for measuring land resources and metrics when a large enough impact radius is assumed (e.g., >4 times the rotor diameter). Our analysis provides a necessary first step toward regionalized impact assessment and improved comparisons of energy alternatives.