3D CNN
Dotaz
Zobrazit nápovědu
Schizophrenia is a severe neuropsychiatric disease whose diagnosis, unfortunately, lacks an objective diagnostic tool supporting a thorough psychiatric examination of the patient. We took advantage of today's computational abilities, structural magnetic resonance imaging, and modern machine learning methods, such as stacked autoencoders (SAE) and 3D convolutional neural networks (3D CNN), to teach them to classify 52 patients with schizophrenia and 52 healthy controls. The main aim of this study was to explore whether complex feature extraction methods can help improve the accuracy of deep learning-based classifiers compared to minimally preprocessed data. Our experiments employed three commonly used preprocessing steps to extract three different feature types. They included voxel-based morphometry, deformation-based morphometry, and simple spatial normalization of brain tissue. In addition to classifier models, features and their combination, other model parameters such as network depth, number of neurons, number of convolutional filters, and input data size were also investigated. Autoencoders were trained on feature pools of 1000 and 5000 voxels selected by Mann-Whitney tests, and 3D CNNs were trained on whole images. The most successful model architecture (autoencoders) achieved the highest average accuracy of 69.62% (sensitivity 68.85%, specificity 70.38%). The results of all experiments were statistically compared (the Mann-Whitney test). In conclusion, SAE outperformed 3D CNN, while preprocessing using VBM helped SAE improve the results.
- Klíčová slova
- 3D CNN, autoencoders, classification, deep learning, deformation-based morphometry, schizophrenia, voxel-based morphometry,
- Publikační typ
- časopisecké články MeSH
BACKGROUND AND OBJECTIVE: We present a fully automatic system based on learning approaches, which aims to localization and identification (labeling) of vertebrae in 3D computed tomography (CT) scans of possibly incomplete spines in patients with bone metastases and vertebral compressions. METHODS: The framework combines a set of 3D algorithms for i) spine detection using a convolution neural network (CNN) ii) spinal cord tracking based on combination of a CNN and a novel growing sphere method with a population optimization, iii) intervertebral discs localization using a novel approach of spatially variant filtering of intensity profiles and iv) vertebra labeling using a CNN-based classification combined with global dynamic optimization. RESULTS: The proposed algorithm has been validated in testing databases, including also a publicly available dataset. The mean error of intervertebral discs localization is 4.4 mm, and for vertebra labeling, the average rate of correctly identified vertebrae is 87.1%, which can be considered a good result with respect to the large share of highly distorted spines and incomplete spine scans. CONCLUSIONS: The proposed framework, which combines several advanced methods including also three CNNs, works fully automatically even with incomplete spine scans and with distorted pathological cases. The achieved results allow including the presented algorithms as the first phase to the fully automated computer-aided diagnosis (CAD) system for automatic spine-bone lesion analysis in oncological patients.
- Klíčová slova
- Convolution neural network, Learning-based approach, Pathological vertebrae, Vertebra detection,
- MeSH
- algoritmy MeSH
- databáze faktografické MeSH
- diagnóza počítačová MeSH
- lidé MeSH
- metastázy nádorů MeSH
- meziobratlová ploténka diagnostické zobrazování patologie MeSH
- nádory kostí diagnostické zobrazování patologie MeSH
- nemoci páteře diagnostické zobrazování MeSH
- neuronové sítě MeSH
- páteř diagnostické zobrazování patologie MeSH
- počítačová rentgenová tomografie * MeSH
- počítačové zpracování obrazu MeSH
- reprodukovatelnost výsledků MeSH
- rozpoznávání automatizované MeSH
- software MeSH
- zobrazování trojrozměrné metody MeSH
- Check Tag
- lidé MeSH
- Publikační typ
- časopisecké články MeSH
This paper aims to address the segmentation and classification of lytic and sclerotic metastatic lesions that are difficult to define by using spinal 3D Computed Tomography (CT) images obtained from highly pathologically affected cases. As the lesions are ill-defined and consequently it is difficult to find relevant image features that would enable detection and classification of lesions by classical methods of texture and shape analysis, the problem is solved by automatic feature extraction provided by a deep Convolutional Neural Network (CNN). Our main contributions are: (i) individual CNN architecture, and pre-processing steps that are dependent on a patient data and a scan protocol - it enables work with different types of CT scans; (ii) medial axis transform (MAT) post-processing for shape simplification of segmented lesion candidates with Random Forest (RF) based meta-analysis; and (iii) usability of the proposed method on whole-spine CTs (cervical, thoracic, lumbar), which is not treated in other published methods (they work with thoracolumbar segments of spine only). Our proposed method has been tested on our own dataset annotated by two mutually independent radiologists and has been compared to other published methods. This work is part of the ongoing complex project dealing with spine analysis and spine lesion longitudinal studies.
- Klíčová slova
- CT analysis, Computer aided detection, Convolutional neural network, Spinal metastasis,
- MeSH
- lidé středního věku MeSH
- lidé MeSH
- nádory páteře diagnostické zobrazování sekundární MeSH
- neuronové sítě * MeSH
- počítačová rentgenová tomografie * MeSH
- rentgenový obraz - interpretace počítačová metody MeSH
- senioři nad 80 let MeSH
- senioři MeSH
- zobrazování trojrozměrné * MeSH
- Check Tag
- lidé středního věku MeSH
- lidé MeSH
- mužské pohlaví MeSH
- senioři nad 80 let MeSH
- senioři MeSH
- ženské pohlaví MeSH
- Publikační typ
- časopisecké články MeSH
- práce podpořená grantem MeSH
This study tested whether machine learning (ML) methods can effectively separate individual plants from complex 3D canopy laser scans as a prerequisite to analyzing particular plant features. For this, we scanned mung bean and chickpea crops with PlantEye (R) laser scanners. Firstly, we segmented the crop canopies from the background in 3D space using the Region Growing Segmentation algorithm. Then, Convolutional Neural Network (CNN) based ML algorithms were fine-tuned for plant counting. Application of the CNN-based (Convolutional Neural Network) processing architecture was possible only after we reduced the dimensionality of the data to 2D. This allowed for the identification of individual plants and their counting with an accuracy of 93.18% and 92.87% for mung bean and chickpea plants, respectively. These steps were connected to the phenotyping pipeline, which can now replace manual counting operations that are inefficient, costly, and error-prone. The use of CNN in this study was innovatively solved with dimensionality reduction, addition of height information as color, and consequent application of a 2D CNN-based approach. We found there to be a wide gap in the use of ML on 3D information. This gap will have to be addressed, especially for more complex plant feature extractions, which we intend to implement through further research.
- Klíčová slova
- 3D point clouds, computer vision, machine learning, phenotyping, plant detection,
- MeSH
- algoritmy * MeSH
- neuronové sítě MeSH
- strojové učení * MeSH
- Publikační typ
- časopisecké články MeSH
Segmentation of macromolecular structures is the primary bottleneck for studying biomolecules and their organization with electron microscopy in 2D/3D - requiring months of manual effort. Transformer-based Rapid Dimensionless Instance Segmentation (TARDIS) is a deep learning framework that automatically and accurately annotates membranes and filaments. Pre-trained TARDIS models can segment electron tomography (ET) reconstructions from both 3D and 2D electron micrographs of cryo and plastic-embedded samples. Furthermore, by implementing a novel geometric transformer architecture, TARDIS is the only method to provide accurate instance segmentations of these structures. Reducing the annotation time for ET data from months to minutes, we demonstrate segmentation of membranes and filaments in over 13,000 tomograms in the CZII Data Portal. TARDIS thus enables quantitative biophysical analysis at scale for the first time. We show this in application to kinetochore-microtubule attachment and viral-membrane interactions. TARDIS can be extended to new biomolecules and applications and open-source at https://github.com/SMLC-NYSBC/TARDIS.
- Klíčová slova
- Actin, CNN, Cryo-EM, Cryo-ET, DIST, Filaments, Instance Segmentation, Membranes, Microtubules, Point Cloud, Segmentation, Semantic Segmentation, TARDIS, TEM EM/ET,
- Publikační typ
- časopisecké články MeSH
- preprinty MeSH
DNA double-strand breaks (DSBs), marked by ionizing radiation-induced (repair) foci (IRIFs), are the most serious DNA lesions and are dangerous to human health. IRIF quantification based on confocal microscopy represents the most sensitive and gold-standard method in radiation biodosimetry and allows research on DSB induction and repair at the molecular and single-cell levels. In this study, we introduce DeepFoci - a deep learning-based fully automatic method for IRIF counting and morphometric analysis. DeepFoci is designed to work with 3D multichannel data (trained for 53BP1 and γH2AX) and uses U-Net for nucleus segmentation and IRIF detection, together with maximally stable extremal region-based IRIF segmentation. The proposed method was trained and tested on challenging datasets consisting of mixtures of nonirradiated and irradiated cells of different types and IRIF characteristics - permanent cell lines (NHDFs, U-87) and primary cell cultures prepared from tumors and adjacent normal tissues of head and neck cancer patients. The cells were dosed with 0.5-8 Gy γ-rays and fixed at multiple (0-24 h) postirradiation times. Under all circumstances, DeepFoci quantified the number of IRIFs with the highest accuracy among current advanced algorithms. Moreover, while the detection error of DeepFoci remained comparable to the variability between two experienced experts, the software maintained its sensitivity and fidelity across dramatically different IRIF counts per nucleus. In addition, information was extracted on IRIF 3D morphometric features and repair protein colocalization within IRIFs. This approach allowed multiparameter IRIF categorization of single- or multichannel data, thereby refining the analysis of DSB repair processes and classification of patient tumors, with the potential to identify specific cell subclones. The developed software improves IRIF quantification for various practical applications (radiotherapy monitoring, biodosimetry, etc.) and opens the door to advanced DSB focus analysis and, in turn, a better understanding of (radiation-induced) DNA damage and repair.
- Klíčová slova
- 53BP1, P53-binding protein 1, Biodosimetry, CNN, convolutional neural network, Confocal Microscopy, Convolutional Neural Network, DNA Damage and Repair, DSB, DNA double-strand break, Deep Learning, FOV, field of view, GUI, graphical user interface, IRIF, ionizing radiation-induced (repair) foci, Image Analysis, Ionizing Radiation-Induced Foci (IRIFs), MSER, maximally stable extremal region (algorithm), Morphometry, NHDFs, normal human dermal fibroblasts, RAD51, DNA repair protein RAD51 homolog 1, U-87, U-87 glioblastoma cell line, γH2AX, histone H2AX phosphorylated at serine 139,
- Publikační typ
- časopisecké články MeSH
Correct virtual reconstruction of a defective skull is a prerequisite for successful cranioplasty and its automatization has the potential for accelerating and standardizing the clinical workflow. This work provides a deep learning-based method for the reconstruction of a skull shape and cranial implant design on clinical data of patients indicated for cranioplasty. The method is based on a cascade of multi-branch volumetric CNNs that enables simultaneous training on two different types of cranioplasty ground-truth data: the skull patch, which represents the exact shape of the missing part of the original skull, and which can be easily created artificially from healthy skulls, and expert-designed cranial implant shapes that are much harder to acquire. The proposed method reaches an average surface distance of the reconstructed skull patches of 0.67 mm on a clinical test set of 75 defective skulls. It also achieves a 12% reduction of a newly proposed defect border Gaussian curvature error metric, compared to a baseline model trained on synthetic data only. Additionally, it produces directly 3D printable cranial implant shapes with a Dice coefficient 0.88 and a surface error of 0.65 mm. The outputs of the proposed skull reconstruction method reach good quality and can be considered for use in semi- or fully automatic clinical cranial implant design workflows.
- Klíčová slova
- 3D convolutional neural networks, Cranial implant design, Cranioplasty, Skull reconstruction,
- MeSH
- deep learning * MeSH
- lebka diagnostické zobrazování chirurgie MeSH
- lidé MeSH
- protézy a implantáty MeSH
- zákroky plastické chirurgie * MeSH
- Check Tag
- lidé MeSH
- Publikační typ
- časopisecké články MeSH
The complex shape of embryonic cartilage represents a true challenge for phenotyping and basic understanding of skeletal development. X-ray computed microtomography (μCT) enables inspecting relevant tissues in all three dimensions; however, most 3D models are still created by manual segmentation, which is a time-consuming and tedious task. In this work, we utilised a convolutional neural network (CNN) to automatically segment the most complex cartilaginous system represented by the developing nasal capsule. The main challenges of this task stem from the large size of the image data (over a thousand pixels in each dimension) and a relatively small training database, including genetically modified mouse embryos, where the phenotype of the analysed structures differs from the norm. We propose a CNN-based segmentation model optimised for the large image size that we trained using a unique manually annotated database. The segmentation model was able to segment the cartilaginous nasal capsule with a median accuracy of 84.44% (Dice coefficient). The time necessary for segmentation of new samples shortened from approximately 8 h needed for manual segmentation to mere 130 s per sample. This will greatly accelerate the throughput of μCT analysis of cartilaginous skeletal elements in animal models of developmental diseases.
Identifying the presence and extent of early ischemic changes (EIC) on Non-Contrast Computed Tomography (NCCT) is key to diagnosing and making time-sensitive treatment decisions in patients that present with Acute Ischemic Stroke (AIS). Segmenting EIC on NCCT is however a challenging task. In this study, we investigated a 3D CNN based on nnU-Net, a self-adapting CNN technique that has become the state-of-the-art in medical image segmentation, for segmenting EIC in NCCT of AIS patients. We trained and tested this model on a sizeable and heterogenous dataset of 534 patients, split into 438 for training and validation and 96 for testing. On this test set, we additionally assessed the inter-rater performance by comparing the proposed approach against two reference segmentation annotations by expert neuroradiologist readers, using this as the benchmark against which to compare our model. In terms of spatial agreement, we report median Dice Similarity Coefficients (DSCs) of 39.8% for the model vs. Reader-1, 39.4% for the model vs. Reader-2, and 55.6% for Reader-2 vs. Reader-1. In terms of lesion volume agreement, we report Intraclass Correlation Coefficients (ICCs) of 83.4% for model vs. Reader-1, 80.4% for model vs. Reader-2, and 94.8% for Reader-2 vs. Reader-1. Based on these results, we conclude that our model performs well relative to expert human performance and therefore may be useful as a decision-aid for clinicians.
- Klíčová slova
- Acute ischemic stroke, Brain lesion, Computed tomography, Computer vision, Deep learning, Machine learning, Neurovascular imaging, Segmentation,
- MeSH
- cévní mozková příhoda * diagnostické zobrazování MeSH
- ischemická cévní mozková příhoda * diagnostické zobrazování MeSH
- lidé MeSH
- počítačová rentgenová tomografie MeSH
- počítačové zpracování obrazu metody MeSH
- Check Tag
- lidé MeSH
- Publikační typ
- časopisecké články MeSH
- práce podpořená grantem MeSH
Achieving a reliable and accurate biomedical image segmentation is a long-standing problem. In order to train or adapt the segmentation methods or measure their performance, reference segmentation masks are required. Usually gold-standard annotations, i.e. human-origin reference annotations, are used as reference although they are very hard to obtain. The increasing size of the acquired image data, large dimensionality such as 3D or 3D + time, limited human expert time, and annotator variability, typically result in sparsely annotated gold-standard datasets. Reliable silver-standard annotations, i.e. computer-origin reference annotations, are needed to provide dense segmentation annotations by fusing multiple computer-origin segmentation results. The produced dense silver-standard annotations can then be either used as reference annotations directly, or converted into gold-standard ones with much lighter manual curation, which saves experts' time significantly. We propose a novel full-resolution multi-rater fusion convolutional neural network (CNN) architecture for biomedical image segmentation masks, called DeepFuse, which lacks any down-sampling layers. Staying everywhere at the full resolution enables DeepFuse to fully benefit from the enormous feature extraction capabilities of CNNs. DeepFuse outperforms the popular and commonly used fusion methods, STAPLE, SIMPLE and other majority-voting-based approaches with statistical significance on a wide range of benchmark datasets as demonstrated on examples of a challenging task of 2D and 3D cell and cell nuclei instance segmentation for a wide range of microscopy modalities, magnifications, cell shapes and densities. A remarkable feature of the proposed method is that it can apply specialized post-processing to the segmentation masks of each rater separately and recover under-segmented object parts during the refinement phase even if the majority of inputs vote otherwise. Thus, DeepFuse takes a big step towards obtaining fast and reliable computer-origin segmentation annotations for biomedical images.
- MeSH
- lidé MeSH
- neuronové sítě * MeSH
- počítačové zpracování obrazu * metody MeSH
- Check Tag
- lidé MeSH
- Publikační typ
- časopisecké články MeSH
