Automatic Tissue Segmentation Research Articles

Research on terahertz image analysis of thin-shell seeds based on semantic segmentation

Assessing crop seed phenotypic traits is essential for breeding innovations and germplasm enhancement. However, the tough outer layers of thin-shelled seeds present significant challenges for traditional methods aimed at the rapid assessment of their internal structures and quality attributes. This study explores the potential of combining terahertz (THz) time-domain spectroscopy and imaging with semantic segmentation models for the rapid and non-destructive examination of these traits. A total of 120 watermelon seed samples from three distinct varieties, were curated in this study, facilitating a comprehensive analysis of both their outer layers and inner kernels. Utilizing a transmission imaging modality, THz spectral images were acquired and subsequently reconstructed employing a correlation coefficient method. Deep learning-based SegNet and DeepLab V3+ models were employed for automatic tissue segmentation. Our research revealed that DeepLab V3+ significantly surpassed SegNet in both speed and accuracy. Specifically, DeepLab V3+ achieved a pixel accuracy of 96.69 % and an intersection over the union of 91.3 % for the outer layer, with the inner kernel results closely following. These results underscore the proficiency of DeepLab V3+ in distinguishing between the seed coat and kernel, thereby furnishing precise phenotypic trait analyses for seeds with thin shells. Moreover, this study accentuates the instrumental role of deep learning technologies in advancing agricultural research and practices.

BounTI (boundary-preserving threshold iteration): A user-friendly tool for automatic hard tissue segmentation.

X-ray Computed Tomography (CT) images are widely used in various fields of natural, physical, and biological sciences. 3D reconstruction of the images involves segmentation of the structures of interest. Manual segmentation has been widely used in the field of biological sciences for complex structures composed of several sub-parts and can be a time-consuming process. Many tools have been developed to automate the segmentation process, all with various limitations and advantages, however, multipart segmentation remains a largely manual process. The aim of this study was to develop an open-access and user-friendly tool for the automatic segmentation of calcified tissues, specifically focusing on craniofacial bones. Here we describe BounTI, a novel segmentation algorithm which preserves boundaries between separate segments through iterative thresholding. This study outlines the working principles behind this algorithm, investigates the effect of several input parameters on its outcome, and then tests its versatility on CT images of the craniofacial system from different species (e.g. a snake, a lizard, an amphibian, a mouse and a human skull) with various scan qualities. The case studies demonstrate that this algorithm can be effectively used to segment the craniofacial system of a range of species automatically. High-resolution microCT images resulted in more accurate boundary-preserved segmentation, nonetheless significantly lower-quality clinical images could still be segmented using the proposed algorithm. Methods for manual intervention are included in this tool when the scan quality is insufficient to achieve the desired segmentation results. While the focus here was on the craniofacial system, BounTI can be used to automatically segment any hard tissue. The tool presented here is available as an Avizo/Amira add-on, a stand-alone Windows executable, and a Python library. We believe this accessible and user-friendly segmentation tool can benefit the wider anatomical community.

Fully automatic AI segmentation of oral surgery-related tissues based on cone beam computed tomography images

Accurate segmentation of oral surgery-related tissues from cone beam computed tomography (CBCT) images can significantly accelerate treatment planning and improve surgical accuracy. In this paper, we propose a fully automated tissue segmentation system for dental implant surgery. Specifically, we propose an image preprocessing method based on data distribution histograms, which can adaptively process CBCT images with different parameters. Based on this, we use the bone segmentation network to obtain the segmentation results of alveolar bone, teeth, and maxillary sinus. We use the tooth and mandibular regions as the ROI regions of tooth segmentation and mandibular nerve tube segmentation to achieve the corresponding tasks. The tooth segmentation results can obtain the order information of the dentition. The corresponding experimental results show that our method can achieve higher segmentation accuracy and efficiency compared to existing methods. Its average Dice scores on the tooth, alveolar bone, maxillary sinus, and mandibular canal segmentation tasks were 96.5%, 95.4%, 93.6%, and 94.8%, respectively. These results demonstrate that it can accelerate the development of digital dentistry.

Modern Magnetic Resonance Imaging Modalities to Advance Neuroimaging in Astronauts.

INTRODUCTION: The rapid development of the space industry requires a deeper understanding of spaceflight's impact on the brain. MRI research reports brain volume changes following spaceflight in astronauts, potentially affecting cognition. Recently, we have demonstrated that this evidence of volumetric changes, as measured by typical T1-weighted sequences (e.g., magnetization-prepared rapid gradient echo sequence; MPRAGE), is error-prone due to the microgravity-related redistribution of cerebrospinal fluid in the brain. More modern neuroimaging methods, particularly dual-echo MPRAGE (DEMPRAGE) and magnetization-prepared rapid gradient echo sequence utilizing two inversion pulses (MP2RAGE), have been suggested to be resilient to this error. Here, we tested if these imaging modalities offered consistent segmentation performance improvements in some commonly employed neuroimaging software packages.METHODS: We conducted manual gray matter tissue segmentation in traditional T1w MRI images to utilize for comparison. Automated tissue segmentation was performed for traditional T1w imaging, as well as on DEMPRAGE and MP2RAGE images from the same subjects. Statistical analysis involved a comparison of total gray matter volumes for each modality, and the extent of tissue segmentation agreement was assessed using a test of similarity (Dice coefficient).RESULTS: Neither DEMPRAGE nor MP2RAGE exhibited consistent segmentation performance across all toolboxes tested.DISCUSSION: This research indicates that customized data collection and processing methods are necessary for reliable and valid structural MRI segmentation in astronauts, as current methods provide erroneous classification and hence inaccurate claims of neuroplastic brain changes in the astronaut population.Berger L, Burles F, Jaswal T, Williams R, Iaria G. Modern magnetic resonance imaging modalities to advance neuroimaging in astronauts. Aerosp Med Hum Perform. 2024; 95(5):245-253.

MicroBundleCompute: Automated segmentation, tracking, and analysis of subdomain deformation in cardiac microbundles.

Advancing human induced pluripotent stem cell derived cardiomyocyte (hiPSC-CM) technology will lead to significant progress ranging from disease modeling, to drug discovery, to regenerative tissue engineering. Yet, alongside these potential opportunities comes a critical challenge: attaining mature hiPSC-CM tissues. At present, there are multiple techniques to promote maturity of hiPSC-CMs including physical platforms and cell culture protocols. However, when it comes to making quantitative comparisons of functional behavior, there are limited options for reliably and reproducibly computing functional metrics that are suitable for direct cross-system comparison. In addition, the current standard functional metrics obtained from time-lapse images of cardiac microbundle contraction reported in the field (i.e., post forces, average tissue stress) do not take full advantage of the available information present in these data (i.e., full-field tissue displacements and strains). Thus, we present "MicroBundleCompute," a computational framework for automatic quantification of morphology-based mechanical metrics from movies of cardiac microbundles. Briefly, this computational framework offers tools for automatic tissue segmentation, tracking, and analysis of brightfield and phase contrast movies of beating cardiac microbundles. It is straightforward to implement, runs without user intervention, requires minimal input parameter setting selection, and is computationally inexpensive. In this paper, we describe the methods underlying this computational framework, show the results of our extensive validation studies, and demonstrate the utility of exploring heterogeneous tissue deformations and strains as functional metrics. With this manuscript, we disseminate "MicroBundleCompute" as an open-source computational tool with the aim of making automated quantitative analysis of beating cardiac microbundles more accessible to the community.

Application of clustering strategy for automatic segmentation of tissue regions in mass spectrometry imaging.

Mass spectrometry imaging (MSI) has been widely used in biomedical research fields. Each pixel in MSI consists of a mass spectrum that reflects the molecule feature of the tissue spot. Because MSI contains high-dimensional datasets, it is highly desired to develop computational methods for data mining and constructing tissue segmentation maps. To visualize different tissue regions based on mass spectrum features and improve the efficiency in processing enormous data, we proposed a computational strategy that consists of four procedures including preprocessing, data reduction, clustering, and quantitative validation. In this study, we examined the combination of t-distributed stochastic neighbor embedding (t-SNE) and hierarchical clustering (HC) for MSI data analysis. Using publicly available MSI datasets, one dataset of mouse urinary bladder, and one dataset of human colorectal cancer, we demonstrated that the generated tissue segmentation maps from this combination were superior to other data reduction and clustering algorithms. Using the staining image as a reference, we assessed the performance of clustering algorithms with external and internal clustering validation measures, including purity, adjusted Rand index (ARI), Davies-Bouldin index (DBI), and spatial aggregation index (SAI). The result indicated that SAI delivered excellent performance for automatic segmentation of tissue regions in MSI. We used a clustering algorithm to construct tissue automatic segmentation in MSI datasets. The performance was evaluated by comparing it with the stained image and calculating clustering validation indexes. The results indicated that SAI is important for automatic tissue segmentation in MSI, different from traditional clustering validation measures. Compared to the reports that used internal clustering validation measures such as DBI, our method offers more effective evaluation of clustering results for MSI segmentation. We envision that the proposed automatic image segmentation strategy can facilitate deep learning in molecular feature extraction and biomarker discovery for the biomedical applications of MSI.

Dual-Attention Model Fusing CNN and Transformer for Pancreas Segmentation.

Pancreatic cancer is a highly malignant cancer of the digestive tract and is rapidly progressing and spreading clinically. Automatic and accurate pancreatic tissue segmentation in abdominal CT images is essential for the early diagnosis of pancreatic-related diseases. It is challenging that the pancreas is small in size and complex in morphology. To address this problem, we propose a dual-attention model fusing CNN and Transformer to effectively activate pancreas-related features expression. The CNN structure weights the importance of pancreas-related features at the channel level and weakens the background information. Transformer feature aggregation module constructs spatial correlations among long-distance pixels from a global perspective. This study is validated on the NIH-TCIA dataset and achieved a mean Dice Similarity Coefficient of 85.82%, which is outperforming than the state-of-the-art methods. The visualization of surface distance also demonstrates the effective segmentation of pancreas boundary details by the proposed model.

Development and Clinical Application of Artificial Intelligence Assistant System for Rotator Cuff Ultrasound Scanning

We developed an intelligent assistance system for shoulder ultrasound imaging, incorporating deep-learning algorithms to facilitate standard plane recognition and automatic tissue segmentation of the rotator cuff and its surrounding structures. We evaluated the system's performance using a dedicated data set of rotator cuff ultrasound images to assess its feasibility in clinical practice. To fulfill the system's primary functions, we designed a standard plane recognition module based on the ResNet50 network and an automatic tissue segmentation module using the Mask R-CNN model. The modules were trained on carefully curated data sets. The standard plane recognition module automatically identifies a specific standard plane based on the ultrasound image characteristics. The automatic tissue segmentation module effectively delineates and segments anatomical structures within the identified standard plane. With the use of 59,265 shoulder joint ultrasound images, the standard plane recognition model achieved an impressive recognition accuracy of 94.9% in the test set, with an average precision rate of 96.4%, recall rate of 95.4% and F1 score of 95.9%. The automatic tissue segmentation model, tested on 1886 images, exhibited a commendable average intersection over union value of 96.2%, indicating robustness and accuracy. The model achieved mean intersection over union values exceeding 90.0% for all standard planes, indicating its effectiveness in precisely delineating the anatomical structures. Our shoulder joint musculoskeletal intelligence system swiftly and accurately identifies standard planes and performs automatic tissue segmentation.

Automatic epicardial adipose tissue segmentation in pulmonary computed tomography venography using nnU-Net.

Epicardial adipose tissue (EAT) is a key aspect in the investigation of cardiac pathophysiology. We sought to develop a deep learning (DL) model for fully automatic extraction and quantification of EAT through pulmonary computed tomography venography (PCTV) images. In this retrospective study, we included 128 patients with atrial fibrillation and PCTV from 2 hospitals. A DL model for automated EAT segmentation was developed from a training set of 51 patients and a validation set of 13 patients from hospital A. The algorithm was further validated using an internal test set of 16 patients from hospital A and an external test set of 48 patients from hospital B. The consistency and measurement agreement of EAT quantification were compared between the DL model and the conventional manual protocol using the Dice score coefficient (DSC), Hausdorff distance (HD95), Pearson correlation coefficient, and Bland-Altman plot. In the internal and external test set, automated segmentation with DL was successful in all cases. The total analysis time was shorter for DL than for manual reconstruction (5.43±2.52 vs. 106.20±15.90 min; P<0.001). The EAT segmented with the DL model had good consistency with manual segmentation (the DSC of the internal and external test sets were 0.92±0.02 and 0.88±0.03, respectively). The quantification of EAT evaluated with the 2 methods showed excellent correlation (all correlation coefficients >0.9; all P values <0.001) and minimal measurement difference. The proposed DL model achieved fully automatic quantification of EAT from PCTV images. The yielded results were highly consistent with those of manual quantification.

Optimization of a deep convolution neural network (DCNN) for osteosarcoma necrosis quantification and outcome prediction.

e13541 Background: Deep convolutional neural network is a form of artificial intelligence that gained traction recently in diagnostic and prognostic applications such as automated tissue segmentation and outcome and response to immunotherapy prediction. Herein, we optimize DCNN to accurately segment and quantify necrosis in osteosarcoma (OSA) archival slides. We also correlate the AI driven necrosis ratios with those derived from experienced sarcoma pathologists (gold standard) and show their value in predicting patient’s outcomes. Methods: A total of 44 patients with bone resections for primary high-grade osteosarcomas post neoadjuvant treatment with conventional chemotherapy were included (Male n= 26, female n=19), median age = 16.4 years. 103 slides from 21 patients were originally manually segmented and used for initial training using 3-fold cross-validation split per patient and per slide. Here, an additional 119 slides from 23 patients (total of 222 slides) were added in the retraining for automated segmentation and to update the necrosis ratio calculation, and the correlation with pathologist calculated necrosis ratio (R spearman correlation). Based on this ratio, we performed a prediction of Overall survival using the Log rank test. Results: The segmentation classifier identified viable tumor, necrosis, normal tissue and artefacts within slides with an area under the curve ranging from 0.91 to 0.96. AI necrosis ratio showed a good correlation with pathologist driven ration (R=0.74 per slide, R=0.69 per patient). Some outliers with highest discrepancy for necrosis ratio included surface high grade periosteal osteosarcomas (known for cartilaginous differentiation that is not part of the original training of the algorithm). Patients with an AI necrosis ratio of ≥ 70% had a better Overall survival at 36 months (p=0.01). Conclusions: Our results show the potential of integration of DCNN for automated segmentation and necrosis quantification of central high OSA resection enhancing efficiency in interpreting these complex surgical specimens. The data also suggests a possible prognostic application of the AI derived necrosis ratio. Further validation of the DCNN performance on an independent cohort is underway.

Anthropometric estimators of abdominal fat volume in adults with overweight and obesity

Background/ObjectivesTo evaluate anthropometric measures for the prediction of whole-abdominal adipose tissue volumes VXAT (subcutaneous VSAT, visceral VVAT and total VTAT) in patients with obesity.Subjects/MethodsA total of 181 patients (108 women) with overweight or obesity were analyzed retrospectively. MRI data (1.5 T) were available from independent clinical trials at a single institution (Integrated Research and Treatment Center of Obesity, University of Leipzig). A custom-made software was used for automated tissue segmentation. Anthropometric parameters (AP) were circumferences of the waist (WC) and hip (HC), waist-to-hip ratio (WHR), waist-to-height ratio (WHtR) and the (hypothetical) hip-to-height ratio (HHtR). Agreement was evaluated by standard deviations sd% of percent differences between estimated volumes (using results of linear AP–VXAT regression) and measured ones as well as Pearson’s correlation coefficient r.ResultsFor SAT volume estimation, the smallest sd% for all patients was seen for HC (25.1%) closely followed by HHtR (25.2%). Sex-specific results for females (17.5% for BMI and 17.2% for HC) and males (20.7% for WC) agreed better. VAT volumes could not be estimated reliably by any of the anthropometric measures considered here. TAT volumes in a mixed population could be best estimated by BMI closely followed by WC (roughly 17.5%). A sex-specific consideration reduced the deviations to around 16% for females (BMI and WC) and below 14% for males (WC).ConclusionsWe suggest the use of sex-specific parameters–BMI or HC for females and WC for males–for the estimation of abdominal SAT and TAT volumes in patients with overweight or obesity.

Early-stage differentiation between Alzheimer's disease and frontotemporal lobe degeneration: Clinical, neuropsychology, and neuroimaging features.

Alzheimer's disease (AD) and frontotemporal lobar degeneration (FTLD) are the two most common forms of neurodegenerative dementia. Although both of them have well-established diagnostic criteria, achieving early diagnosis remains challenging. Here, we aimed to make the differential diagnosis of AD and FTLD from clinical, neuropsychological, and neuroimaging features. In this retrospective study, we selected 95 patients with PET-CT defined AD and 106 patients with PET-CT/biomarker-defined FTLD. We performed structured chart examination to collect clinical data and ascertain clinical features. A series of neuropsychological scales were used to assess the neuropsychological characteristics of patients. Automatic tissue segmentation of brain by Dr. Brain tool was used to collect multi-parameter volumetric measurements from different brain areas. All patients' structural neuroimage data were analyzed to obtain brain structure and white matter hyperintensities (WMH) quantitative data. The prevalence of vascular disease associated factors was higher in AD patients than that in FTLD group. 56.84% of patients with AD carried at least one APOE ε4 allele, which is much high than that in FTLD patients. The first symptoms of AD patients were mostly cognitive impairment rather than behavioral abnormalities. In contrast, behavioral abnormalities were the prominent early manifestations of FTLD, and few patients may be accompanied by memory impairment and motor symptoms. In direct comparison, patients with AD had slightly more posterior lesions and less frontal atrophy, whereas patients with FTLD had more frontotemporal atrophy and less posterior lesions. The WMH burden of AD was significantly higher, especially in cortical areas, while the WMH burden of FTLD was higher in periventricular areas. These results indicate that dynamic evaluation of cognitive function, behavioral and psychological symptoms, and multimodal neuroimaging are helpful for the early diagnosis and differentiation between AD and FTLD.

Reproducibility and Optimal Arterial Input Function Selection in Dynamic Contrast-Enhanced Perfusion MRI in the Healthy Brain.

Dynamic contrast-enhanced MRI (DCE-MRI) has seen increasing use for quantification of low level of blood-brain barrier (BBB) leakage in various pathological disease states and correlations with clinical outcomes. However, currently there exists limited studies on reproducibility in healthy controls, which is important for the establishment of a normality threshold for future research. To investigate the reproducibility of DCE-MRI and to evaluate the effect of arterial input function (AIF) selection and manual region of interests (ROI) delineation vs. automated global segmentation. Prospective. A total of 16 healthy controls; 11 females; mean age 28.7 years (SD 10.1). A 3T; GE DCE; 3D TFE T1WI. 2D TSE T2. The influx constant Ki , a measure of BBB permeability, and Vp , the blood plasma volume, was calculated using the Patlak model. Cerebral blood flow (CBF) was calculated using Tikhonov model free deconvolution. Manual tissue ROIs, drawn by H.J.S. (30+ years of experience), were compared to automatic tissue segmentation. Intraclass correlation coefficient (ICC) and repeatability coefficient (RC) was used to assess reproducibility. Bland-Altman plots were used to evaluate agreement between measurements day 1 vs. day 2, and manual vs. segmentation method. Ki showed excellent reproducibility in both white and gray matter with an ICC between 0.79 and 0.82 and excellent agreement between manual ROI and automatic segmentation, with an ICC of 0.89 for Ki in WM. Furthermore, Ki values in gray and white matter conforms with histological tissue characteristics, where gray matter generally has a 2-fold higher vessel density. The highest reproducibility measures of Ki (ICC=0.83), CBF (ICC=0.77) and Vd (ICC=0.83) was obtained with the AIF sampled in the internal carotid artery (ICA). DCE-MRI shows excellent reproducibility of pharmacokinetic variables derived from healthy controls. 2 TECHNICAL EFFICACY: Stage 2.

DBB - A Distorted Brain Benchmark for Automatic Tissue Segmentation in Paediatric Patients

T1-weighted magnetic resonance images provide a comprehensive view of the morphology of the human brain at the macro scale. These images are usually the input of a segmentation process that aims detecting the anatomical structures labeling them according to a predefined set of target tissues. Automated methods for brain tissue segmentation rely on anatomical priors of the human brain structures. This is the reason why their performance is quite accurate on healthy individuals. Nevertheless model-based tools become less accurate in clinical practice, specifically in the cases of severe lesions or highly distorted cerebral anatomy. More recently there are empirical evidences that a data-driven approach can be more robust in presence of alterations of brain structures, even though the learning model is trained on healthy brains. Our contribution is a benchmark to support an open investigation on how the tissue segmentation of distorted brains can be improved by adopting a supervised learning approach. We formulate a precise definition of the task and propose an evaluation metric for a fair and quantitative comparison. The training sample is composed of almost one thousand healthy individuals. Data include both T1-weighted MR images and their labeling of brain tissues. The test sample is a collection of several tens of individuals with severe brain distortions. Data and code are openly published on BrainLife, an open science platform for reproducible neuroscience data analysis.

Histopathological Tissue Segmentation of Lung Cancer with Bilinear CNN and Soft Attention.

Automatic tissue segmentation in whole-slide images (WSIs) is a critical task in hematoxylin and eosin- (H&E-) stained histopathological images for accurate diagnosis and risk stratification of lung cancer. Patch classification and stitching the classification results can fast conduct tissue segmentation of WSIs. However, due to the tumour heterogeneity, large intraclass variability and small interclass variability make the classification task challenging. In this paper, we propose a novel bilinear convolutional neural network- (Bilinear-CNN-) based model with a bilinear convolutional module and a soft attention module to tackle this problem. This method investigates the intraclass semantic correspondence and focuses on the more distinguishable features that make feature output variations relatively large between interclass. The performance of the Bilinear-CNN-based model is compared with other state-of-the-art methods on the histopathological classification dataset, which consists of 107.7 k patches of lung cancer. We further evaluate our proposed algorithm on an additional dataset from colorectal cancer. Extensive experiments show that the performance of our proposed method is superior to that of previous state-of-the-art ones and the interpretability of our proposed method is demonstrated by Grad-CAM.

Simultaneous and Spatially-Resolved Analysis of T-Lymphocytes, Macrophages and PD-L1 Immune Checkpoint in Rare Cancers.

Simple SummaryTo study various biomarkers, it is necessary to analyze multiple tissue sections through serial histological sections, which is challenging when only a small tissue sample is available. In this work we have developed a validated and objective method for combined biomarker immunostaining and its digital image analysis using open informatics tools, which is necessary for comprehensive understanding of the tumor microenvironment in rare cancers and in cases of limited samples with very significant clinical features.Penile, vulvar and anal neoplasms show an incidence lower than 0.5% of the population per year and therefore can be considered as rare cancers but with a dramatic impact on quality of life and survival. This work describes the experience of a Chilean cancer center using multiplexed immunofluorescence to study a case series of four penile cancers, two anal cancers and one vulvar cancer and simultaneous detection of CD8, CD68, PD-L1, Cytokeratin and Ki-67 in FFPE samples. Fluorescent image analyses were performed using open sources for automated tissue segmentation and cell phenotyping. Our results showed an objective and reliable counting of objects with a single or combined labeling or within a specific tissue compartment. The variability was below 10%, and the correlation between analytical events was 0.92–0.97. Critical cell phenotypes, such as TILs, PD-L1+ or proliferative tumor cells were detected in a supervised and unsupervised manner with a limit of detection of less than 1% of relative abundance. Finally, the observed diversity and abundance of the different cell phenotypes within the tumor microenvironment for the three studied tumor types confirmed that our methodology is useful and robust to be applicable for many other solid tumors.

Evaluation and Correlation Analysis of Mental and Psychological Factors and Premature Ejaculation in Patients with Benign Prostatic Hyperplasia in Mobile Medical System.

Chronic prostatitis is a very common and very difficult disease. Based on the mobile medical system, this paper carried out a correlation analysis on the psychological factors and the evaluation of premature ejaculation in patients with benign prostatic hyperplasia. The article first analyzes the application in the field of mobile medical and then introduces the prostate image segmentation method based on the geometric active contour model. The emergence of automatic organ tissue segmentation technology is timely; it can help clinicians save a lot of manual segmentation time and has better reversibility and objectivity, and the theory of curve evolution is analyzed. Finally, this paper introduces the experimental research object and purpose, makes a statistical analysis of the symptoms of benign prostatic hyperplasia, and compares the incidence of psychological symptoms in patients with prostatitis under different factors. The experimental results of this paper show that 90% of prostatitis patients have mild psychological problems, and 10% have moderate psychological problems. Among them, the main reason for the psychological disorder of patients is depression, for which we should strengthen the care for patients.

The effect of lesion filling on brain network analysis in multiple sclerosis using structural magnetic resonance imaging

BackgroundGraph theoretical network analysis with structural magnetic resonance imaging (MRI) of multiple sclerosis (MS) patients can be used to assess subtle changes in brain networks. However, the presence of multiple focal brain lesions might impair the accuracy of automatic tissue segmentation methods, and hamper the performance of graph theoretical network analysis. Applying “lesion filling” by substituting the voxel intensities of a lesion with the voxel intensities of nearby voxels, thus creating an image devoid of lesions, might improve segmentation and graph theoretical network analysis. This study aims to determine if brain networks are different between MS subtypes and healthy controls (HC) and if the assessment of these differences is affected by lesion filling.MethodsThe study included 49 MS patients and 19 HC that underwent a T1w, and T2w-FLAIR MRI scan. Graph theoretical network analysis was performed from grey matter fractions extracted from the original T1w-images and T1w-images after lesion filling.ResultsArtefacts in lesion-filled T1w images correlated positively with total lesion volume (r = 0.84, p < 0.001) and had a major impact on grey matter segmentation accuracy. Differences in sensitivity for network alterations were observed between original T1w data and after application of lesion filling: graph theoretical network analysis obtained from lesion-filled T1w images produced more differences in network organization in MS patients.ConclusionLesion filling might reduce variability across subjects resulting in an increased detection rate of network alterations in MS, but also induces significant artefacts, and therefore should be applied cautiously especially in individuals with higher lesions loads.

Semantic segmentation of multispectral photoacoustic images using deep learning

Photoacoustic (PA) imaging has the potential to revolutionize functional medical imaging in healthcare due to the valuable information on tissue physiology contained in multispectral photoacoustic measurements. Clinical translation of the technology requires conversion of the high-dimensional acquired data into clinically relevant and interpretable information. In this work, we present a deep learning-based approach to semantic segmentation of multispectral photoacoustic images to facilitate image interpretability. Manually annotated photoacoustic and ultrasound imaging data are used as reference and enable the training of a deep learning-based segmentation algorithm in a supervised manner. Based on a validation study with experimentally acquired data from 16 healthy human volunteers, we show that automatic tissue segmentation can be used to create powerful analyses and visualizations of multispectral photoacoustic images. Due to the intuitive representation of high-dimensional information, such a preprocessing algorithm could be a valuable means to facilitate the clinical translation of photoacoustic imaging.

3D U-Net Improves Automatic Brain Extraction for Isotropic Rat Brain Magnetic Resonance Imaging Data.

Brain extraction is a critical pre-processing step in brain magnetic resonance imaging (MRI) analytical pipelines. In rodents, this is often achieved by manually editing brain masks slice-by-slice, a time-consuming task where workloads increase with higher spatial resolution datasets. We recently demonstrated successful automatic brain extraction via a deep-learning-based framework, U-Net, using 2D convolutions. However, such an approach cannot make use of the rich 3D spatial-context information from volumetric MRI data. In this study, we advanced our previously proposed U-Net architecture by replacing all 2D operations with their 3D counterparts and created a 3D U-Net framework. We trained and validated our model using a recently released CAMRI rat brain database acquired at isotropic spatial resolution, including T2-weighted turbo-spin-echo structural MRI and T2*-weighted echo-planar-imaging functional MRI. The performance of our 3D U-Net model was compared with existing rodent brain extraction tools, including Rapid Automatic Tissue Segmentation, Pulse-Coupled Neural Network, SHape descriptor selected External Regions after Morphologically filtering, and our previously proposed 2D U-Net model. 3D U-Net demonstrated superior performance in Dice, Jaccard, center-of-mass distance, Hausdorff distance, and sensitivity. Additionally, we demonstrated the reliability of 3D U-Net under various noise levels, evaluated the optimal training sample sizes, and disseminated all source codes publicly, with a hope that this approach will benefit rodent MRI research community.Significant Methodological Contribution: We proposed a deep-learning-based framework to automatically identify the rodent brain boundaries in MRI. With a fully 3D convolutional network model, 3D U-Net, our proposed method demonstrated improved performance compared to current automatic brain extraction methods, as shown in several qualitative metrics (Dice, Jaccard, PPV, SEN, and Hausdorff). We trust that this tool will avoid human bias and streamline pre-processing steps during 3D high resolution rodent brain MRI data analysis. The software developed herein has been disseminated freely to the community.