Automatic Image Segmentation Research Articles

Automatic segmentation and quantification of OCT images before and after cataract surgery using deep learning

Obtaining quantitative geometry of the anterior segment of the eye, generally from optical coherence tomography (OCT) images, is important to construct 3D computer eye models, used to understand the optical quality of the normal and pathological eye and to improve treatment (for example, selecting the intraocular lens to be implanted in cataract surgery or guiding refractive surgery). An important step to quantify OCT images is segmentation (i.e., finding and labeling the surfaces of interest in the images), which, for the purpose of feeding optical models, needs to be automatic, accurate, robust, and fast. In this work, we designed a segmentation algorithm based on deep learning, which we applied to OCT images from pre- and post-cataract surgery eyes obtained using anterior segment OCT commercial systems. We proposed a feature pyramid network architecture with a pre-trained encoder and trained, validated, and tested the algorithm using 1640 OCT images. We showed that the proposed method outperformed a classical image-processing-based approach in terms of accuracy (from 91.4% to 93.2% accuracy), robustness (decreasing the standard deviation of accuracy across images by a factor of 1.7), and processing time (from 0.48 to 0.34 s/image). We also described a method for the 3D models’ construction and their quantification from the segmented images and applied the proposed segmentation/quantification algorithms to quantify 136 new eye measurements (780 images) obtained from OCT commercial systems.

Automatic segmentation of knee CT images of tibial plateau fractures based on three-dimensional U-Net: Assisting junior physicians with Schatzker classification

PurposeThis study aimed to automatically segment knee computed tomography (CT) images of tibial plateau fractures using a three-dimensional (3D) U-net-based method, accurately construct 3D maps of tibial plateau fractures, and examine their usefulness for Schatzker classification in clinical practice. MethodsWe retrospectively enrolled 234 cases with tibial plateau fractures from our hospital in this study. The four constituent bones of the knee were manually annotated using ITK-SNAP software. Finally, image features were extracted using deep learning. The usefulness of the results for Schatzker classification was examined by an orthopaedic and a radiology resident. ResultsOn average, our model required < 40 s to process a 3D CT scan of the knee. The average Dice coefficient for all four knee bones was higher than 0.950, and highly accurate 3D maps of the tibia were produced. With the aid of the results of our model, the accuracy, sensitivity, and specificity of the Schatzker classification of both residents improved. ConclusionsThe proposed method can rapidly and accurately segment knee CT images of tibial plateau fractures and assist residents with Schatzker classification, which can help improve diagnostic efficiency and reduce the workload of junior doctors in clinical practice.

Characterization of gray matter volume changes from one week to 6 months after termination of electroconvulsive therapy in depressed patients

BackgroundIncreased gray matter volume (GMV) following electroconvulsive therapy (ECT) has been well-documented, with limited studies reporting a subsequent decrease in GMV afterwards. ObjectiveThis study characterized the reversion pattern of GMV after ECT and its association with clinical depression outcome, using multi-site triple time-point data from the Global ECT-MRI Research Collaboration (GEMRIC). Methods86 subjects from the GEMRIC database were included, and GMV in 84 regions-of-interest (ROI) was obtained from automatic segmentation of T1 MRI images at three timepoints: pre-ECT (T0), within one-week post-ECT (T1), and one to six months post-ECT (T2). RM-ANOVAs were used to assess longitudinal changes and LMM analyses explored associations between GMV changes and demographical and clinical characteristics. Results63 of the 84 ROIs showed a significant increase-and-decrease pattern (RM-ANOVA, Bonferroni corrected p < 0.00059). Post hoc tests indicated a consistent pattern in each of these 63 ROIs: significant increase from T0 to T1inGMV, followed by significant decrease from T1 to T2 and no difference between T0 and T2, except for both amygdalae, right hippocampus and pars triangularis, which showed the same increase and decrease but GMV at T2 remained higher compared to T0. No consistent relationship was found between GMV change pattern and clinical status. ConclusionThe GEMRIC cohort confirmed a rapid increase of GMV after ECT followed by reversion of GMV one to six months thereafter. The lack of association between the GMV change pattern and depression outcome scores implies a transient neurobiological effect of ECT unrelated to clinical improvement.

Automatic segmentation of spine x-ray images based on multiscale feature enhancement network.

Automatic segmentation of vertebrae in spinal x-ray images is crucial for clinical diagnosis, case analysis, and surgical planning of spinal lesions. However, due to the inherent characteristics of x-ray images, including low contrast, high noise, and uneven grey scale, it remains a critical and challenging problem in computer-aided spine image analysis and disease diagnosis applications. In this paper, a Multiscale Feature Enhancement Network (MFENet), is proposed for segmenting whole spinal x-ray images, to aid doctors in diagnosing spinal-related diseases. To enhance feature extraction, the network incorporates a Dual-branch Feature Extraction Module (DFEM) and a Semantic Aggregation Module (SAM). The DFEM has a parallel dual-branch structure. The upper branch utilizes multiscale convolutional kernels to extract features from images. Employing convolutional kernels of different sizes helps capture details and structural information at different scales. The lower branch incorporates attention mechanisms to further optimize feature representation. By modeling the feature maps spatially and across channels, the network becomes more focused on key feature regions and suppresses task-irrelevant information. The SAM leverages contextual semantic information to compensate for details lost during pooling and convolution operations. It integrates high-level feature information from different scales to reduce segmentation result discontinuity. In addition, a hybrid loss function is employed to enhance the network's feature extraction capability. In this study, we conducted a multitude of experiments utilizing dataset provided by the Spine Surgery Department of Henan Provincial People's Hospital. The experimental results indicate that our proposed MFENet demonstrates superior segmentation performance in spinal segmentation on x-ray images compared to other advanced methods, achieving 92.61±0.431 for MIoU, 92.42±0.329 for DSC, and 99.51±0.037 for Global_accuracy. Our model is able to more effectively learn and extract global contextual semantic information, significantly improving spinal segmentation performance, further aiding doctors in analyzing patient conditions.

Enhanced Medical Image Segmentation using Transfer Learning with Res101_UNet: Experimental Insights and Comparative Performance Analysis

Throughout the past few decades, artificial intelligence and machine learning have seen a lot of active research in areas such as computer vision, natural language processing, and speech processing. As a result, deep learning models became state-of-the-art for computer vision tasks such as object detection, classification, segmentation, and other allied tasks. Of course, the fruits of this research are extended to the design of robust and reliable digital health systems as well as other applications in the healthcare sector. Many clinical applications require the automatic segmentation of medical images. Recent deep learning-based approaches have demonstrated state-of-the-art performance in medical image segmentation tasks. In addition to their ability to automatically extract features and generalize over large amounts of data, transfer learning based deep learning models have proven to be handy for data scared areas like medical domains. In this research, we investigate and demonstrate the efficacy of a DCNN-based transfer learning model -Res101_Unet, which has been trained and/or fine-tuned to execute tumor tissue segmentation tasks in MRI, CT, PET, and X-RAY pictures of medical organ scans with little data. For our experimental study, we employed two image datasets: 'Liver Tumor' and 'Gland Colon Cancer', both obtained from the Kaggle portal. This experimental setup includes an Open-Source segmentation model API. Our findings indicate that domain similarity-based transfer learning can be used to data-scarce sectors. We achieved 98.47% accuracy and a IoU score of 0.9891 on Liver Tumor data and 0.6956 accuracy and a IoU score of 0.7043 on gland colon dataset.

Skin Lesion Segmentation for Melanoma Using Dilated DenseUNet

Melanoma, a highly malignant form of skin cancer, affects individuals of all genders and is associated with high mortality rates, especially in advanced stages. The use of tele-dermatology has emerged as a proficient diagnostic approach for skin lesions and is particularly beneficial in rural areas with limited access to dermatologists. However, accurately, and efficiently segmenting melanoma remains a challenging task due to the significant diversity observed in the morphology, pigmentation, and dimensions of cutaneous nevi. To address this challenge, we propose a novel approach called DenseUNet-169 with a dilated convolution encoder-decoder for automatic segmentation of RGB dermascopic images. By incorporating dilated convolution, our model improves the receptive field of the kernels without increasing the number of parameters. Additionally, we used a method called Copy and Concatenation Attention Block (CCAB) for robust feature computation. To evaluate the performance of our proposed framework, we utilized the International Skin Imaging Collaboration (ISIC) 2017 dataset. The experimental results demonstrate the reliability and effectiveness of our suggested approach compared to existing methodologies. Our framework achieved a high level of accuracy (98.38%), precision (96.07%), recall (94.32%), dice score (95.07%), and Jaccard score (90.45%), outperforming current techniques.

Improvement of gram staining effect by ethanol pretreatment and quantization of staining image by unsupervised machine learning.

In this study, we propose an Ethanol Pretreatment Gram staining method that significantly enhances the color contrast of the stain, thereby improving the accuracy of judgement, and demonstrated the effectiveness of the modification by eliminating unaided-eye observational errors with unsupervised machine learning image analysis. By comparing the traditional Gram staining method with the improved method on various bacterial samples, results showed that the improved method offers distinct color contrast. Using multimodal assessment strategies, including unaided-eye observation, manual image segmentation, and advanced unsupervised machine learning automatic image segmentation, the practicality of ethanol pretreatment on Gram staining was comprehensively validated. In our quantitative analysis, the application of the CIEDE2000, and CMC color difference standards confirmed the significant effect of the method in enhancing the discrimination of Gram staining.This study not only improved the efficacy of Gram staining, but also provided a more accurate and standardized strategy for analyzing Gram staining results, which might provide an useful analytical tool in microbiological diagnostics.

A semi-supervised segmentation method for microscopic hyperspectral pathological images based on multi-consistency learning.

Pathological images are considered the gold standard for clinical diagnosis and cancer grading. Automatic segmentation of pathological images is a fundamental and crucial step in constructing powerful computer-aided diagnostic systems. Medical microscopic hyperspectral pathological images can provide additional spectral information, further distinguishing different chemical components of biological tissues, offering new insights for accurate segmentation of pathological images. However, hyperspectral pathological images have higher resolution and larger area, and their annotation requires more time and clinical experience. The lack of precise annotations limits the progress of research in pathological image segmentation. In this paper, we propose a novel semi-supervised segmentation method for microscopic hyperspectral pathological images based on multi-consistency learning (MCL-Net), which combines consistency regularization methods with pseudo-labeling techniques. The MCL-Net architecture employs a shared encoder and multiple independent decoders. We introduce a Soft-Hard pseudo-label generation strategy in MCL-Net to generate pseudo-labels that are closer to real labels for pathological images. Furthermore, we propose a multi-consistency learning strategy, treating pseudo-labels generated by the Soft-Hard process as real labels, by promoting consistency between predictions of different decoders, enabling the model to learn more sample features. Extensive experiments in this paper demonstrate the effectiveness of the proposed method, providing new insights for the segmentation of microscopic hyperspectral tissue pathology images.

Automatic segmentation of MRI in prospective breast volume evaluation: Comparison of different assessments for immediate breast reconstruction

Assessment of breast volume is essential in preoperative planning of immediate breast reconstruction (IBR) surgery to achieve satisfactory cosmetic outcome. This study introduced a breast volume measurement tool that can be used to perform automatic segmentation of magnetic resonance images (MRI) and calculation of breast volume. We compared the accuracy and reliability of this measurement method with four other conventional modalities. Patients who were scheduled to undergo mastectomy with IBR between 2016 and 2021 were enrolled in the study. Five different breast volume assessments, including automatic segmentation of MRI, manual segmentation of MRI, 3D surface imaging, mammography, and the BREAST-V formula, were used to evaluate different breast volumes. The results were validated using water displacement volumes of the mastectomy specimens. In this pilot study, a total of 50 female patients met the inclusion criteria and contributed 54 breast specimens to the volumetric analysis. There was a strong linear association between the MRI and water displacement methods (automatic segmentation: r=0.911, p<0.001; manual segmentation: r=0.924, p<0.001), followed by 3D surface imaging (r=0.858, p<0.001), mammography (r=0.841, p<0.001), and Breast-V formula (r=0.838, p<0.001). Breast volumes measured using automatic and manual segmentation of MRI had lower mean relative errors (30.3%±22.0% and 28.9%±19.8, respectively) than 3D surface imaging (38.9%±31.2), Breast-V formula (44.8%±25.8), and mammography (60.3%±37.6). Breast volume assessment using the MRI methods had better accuracy and reliability than the other methods used in our study. Breast volume measurement using automatic segmentation of MRI could be more efficient compared to the conventional methods.

A novel CNN-based image segmentation pipeline for individualized feline spinal cord stimulation modeling

Objective. Spinal cord stimulation (SCS) is a well-established treatment for managing certain chronic pain conditions. More recently, it has also garnered attention as a means of modulating neural activity to restore lost autonomic or sensory-motor function. Personalized modeling and treatment planning are critical aspects of safe and effective SCS (Rowald and Amft 2022 Front. Neurorobotics 16 983072, Wagner et al 2018 Nature 563 65–71). However, the generation of spine models at the required level of detail and accuracy requires time and labor intensive manual image segmentation by human experts. This study aims to develop a maximally automated segmentation routine capable of producing high-quality anatomical models, even with limited data, to facilitate safe and effective personalized SCS treatment planning. Approach. We developed an automated image segmentation and model generation pipeline based on a novel convolutional neural network (CNN) architecture trained on feline spinal cord magnetic resonance imaging data. The pipeline includes steps for image preprocessing, data augmentation, transfer learning, and cleanup. To assess the relative importance of each step in the pipeline and our choice of CNN architecture, we systematically dropped steps or substituted architectures, quantifying the downstream effects in terms of tissue segmentation quality (Jaccard index and Hausdorff distance) and predicted nerve recruitment (estimated axonal depolarization). Main results. The leave-one-out analysis demonstrated that each pipeline step contributed a small but measurable increment to mean segmentation quality. Surprisingly, minor differences in segmentation accuracy translated to significant deviations (ranging between 4% and 13% for each pipeline step) in predicted nerve recruitment, highlighting the importance of careful workflow design. Additionally, transfer learning techniques enhanced segmentation metric consistency and allowed generalization to a completely different spine region with minimal additional training data. Significance. To our knowledge, this work is the first to assess the downstream impacts of segmentation quality differences on neurostimulation predictions. It highlights the role of each step in the pipeline and paves the way towards fully automated, personalized SCS treatment planning in clinical settings.

Optimizing cardiovascular image segmentation through integrated hierarchical features and attention mechanisms.

Cardiovascular diseases are the top cause of death in China. Manual segmentation of cardiovascular images, prone to errors, demands an automated, rapid, and precise solution for clinical diagnosis. The paper highlights deep learning in automatic cardiovascular image segmentation, efficiently identifying pixel regions of interest for auxiliary diagnosis and research in cardiovascular diseases. In our study, we introduce innovative Region Weighted Fusion (RWF) and Shape Feature Refinement (SFR) modules, utilizing polarized self-attention for significant performance improvement in multiscale feature integration and shape fine-tuning. The RWF module includes reshaping, weight computation, and feature fusion, enhancing high-resolution attention computation and reducing information loss. Model optimization through loss functions offers a more reliable solution for cardiovascular medical image processing. Our method excels in segmentation accuracy, emphasizing the vital role of the RWF module. It demonstrates outstanding performance in cardiovascular image segmentation, potentially raising clinical practice standards. Our method ensures reliable medical image processing, guiding cardiovascular segmentation for future advancements in practical healthcare and contributing scientifically to enhanced disease diagnosis and treatment.

Automated quantification of photoreceptor outer segments in developing and degenerating retinas on microscopy images across scales

The functionality of photoreceptors, rods, and cones is highly dependent on their outer segments (POS), a cellular compartment containing highly organized membranous structures that generate biochemical signals from incident light. While POS formation and degeneration are qualitatively assessed on microscopy images, reliable methodology for quantitative analyses is still limited. Here, we developed methods to quantify POS (QuaPOS) maturation and quality on retinal sections using automated image analyses. POS formation was examined during the development and in adulthood of wild-type mice via light microscopy (LM) and transmission electron microscopy (TEM). To quantify the number, size, shape, and fluorescence intensity of POS, retinal cryosections were immunostained for the cone POS marker S-opsin. Fluorescence images were used to train the robust classifier QuaPOS-LM based on supervised machine learning for automated image segmentation. Characteristic features of segmentation results were extracted to quantify the maturation of cone POS. Subsequently, this quantification method was applied to characterize POS degeneration in “cone photoreceptor function loss 1” mice. TEM images were used to establish the ultrastructural quantification method QuaPOS-TEM for the alignment of POS membranes. Images were analyzed using a custom-written MATLAB code to extract the orientation of membranes from the image gradient and their alignment (coherency). This analysis was used to quantify the POS morphology of wild-type and two inherited retinal degeneration (“retinal degeneration 19” and “rhodopsin knock-out”) mouse lines. Both automated analysis technologies provided robust characterization and quantification of POS based on LM or TEM images. Automated image segmentation by the classifier QuaPOS-LM and analysis of the orientation of membrane stacks by QuaPOS-TEM using fluorescent or TEM images allowed quantitative evaluation of POS formation and quality. The assessments showed an increase in POS number, volume, and membrane coherency during wild-type postnatal development, while a decrease in all three observables was detected in different retinal degeneration mouse models. All the code used for the presented analysis is open source, including example datasets to reproduce the findings. Hence, the QuaPOS quantification methods are useful for in-depth characterization of POS on retinal sections in developmental studies, for disease modeling, or after therapeutic interventions affecting photoreceptors.

Detection of Road Crack Images Based on Multistage Feature Fusion and a Texture Awareness Method.

Structural health monitoring for roads is an important task that supports inspection of transportation infrastructure. This paper explores deep learning techniques for crack detection in road images and proposes an automatic pixel-level semantic road crack image segmentation method based on a Swin transformer. This method employs Swin-T as the backbone network to extract feature information from crack images at various levels and utilizes the texture unit to extract the texture and edge characteristic information of cracks. The refinement attention module (RAM) and panoramic feature module (PFM) then merge these diverse features, ultimately refining the segmentation results. This method is called FetNet. We collect four public real-world datasets and conduct extensive experiments, comparing FetNet with various deep-learning methods. FetNet achieves the highest precision of 90.4%, a recall of 85.3%, an F1 score of 87.9%, and a mean intersection over union of 78.6% on the Crack500 dataset. The experimental results show that the FetNet approach surpasses other advanced models in terms of crack segmentation accuracy and exhibits excellent generalizability for use in complex scenes.

Hi-gMISnet: generalized medical image segmentation using DWT based multilayer fusion and dual mode attention into high resolution pGAN

Objective. Automatic medical image segmentation is crucial for accurately isolating target tissue areas in the image from background tissues, facilitating precise diagnoses and procedures. While the proliferation of publicly available clinical datasets led to the development of deep learning-based medical image segmentation methods, a generalized, accurate, robust, and reliable approach across diverse imaging modalities remains elusive. Approach. This paper proposes a novel high-resolution parallel generative adversarial network (pGAN)-based generalized deep learning method for automatic segmentation of medical images from diverse imaging modalities. The proposed method showcases better performance and generalizability by incorporating novel components such as partial hybrid transfer learning, discrete wavelet transform (DWT)-based multilayer and multiresolution feature fusion in the encoder, and a dual mode attention gate in the decoder of the multi-resolution U-Net-based GAN. With multi-objective adversarial training loss functions including a unique reciprocal loss for enforcing cooperative learning in pGANs, it further enhances the robustness and accuracy of the segmentation map. Main results. Experimental evaluations conducted on nine diverse publicly available medical image segmentation datasets, including PhysioNet ICH, BUSI, CVC-ClinicDB, MoNuSeg, GLAS, ISIC-2018, DRIVE, Montgomery, and PROMISE12, demonstrate the proposed method’s superior performance. The proposed method achieves mean F1 scores of 79.53%, 88.68%, 82.50%, 93.25%, 90.40%, 94.19%, 81.65%, 98.48%, and 90.79%, respectively, on the above datasets, surpass state-of-the-art segmentation methods. Furthermore, our proposed method demonstrates robust multi-domain segmentation capabilities, exhibiting consistent and reliable performance. The assessment of the model’s proficiency in accurately identifying small details indicates that the high-resolution generalized medical image segmentation network (Hi-gMISnet) is more precise in segmenting even when the target area is very small. Significance. The proposed method provides robust and reliable segmentation performance on medical images, and thus it has the potential to be used in a clinical setting for the diagnosis of patients.

Shifting to machine supervision: annotation-efficient semi and self-supervised learning for automatic medical image segmentation and classification

Advancements in clinical treatment are increasingly constrained by the limitations of supervised learning techniques, which depend heavily on large volumes of annotated data. The annotation process is not only costly but also demands substantial time from clinical specialists. Addressing this issue, we introduce the S4MI (Self-Supervision and Semi-Supervision for Medical Imaging) pipeline, a novel approach that leverages advancements in self-supervised and semi-supervised learning. These techniques engage in auxiliary tasks that do not require labeling, thus simplifying the scaling of machine supervision compared to fully-supervised methods. Our study benchmarks these techniques on three distinct medical imaging datasets to evaluate their effectiveness in classification and segmentation tasks. Notably, we observed that self-supervised learning significantly surpassed the performance of supervised methods in the classification of all evaluated datasets. Remarkably, the semi-supervised approach demonstrated superior outcomes in segmentation, outperforming fully-supervised methods while using 50% fewer labels across all datasets. In line with our commitment to contributing to the scientific community, we have made the S4MI code openly accessible, allowing for broader application and further development of these methods. The code can be accessed at https://github.com/pranavsinghps1/S4MI.

An improvement method for pancreas CT segmentation using superpixel-based active contour

Objective. Pancreas is one of the most challenging organs for Computed Tomograph (CT) image automatic segmentation due to its complex shapes and fuzzy edges. It is simple and universal to use the traditional segmentation method as a post-processor of deep learning method for segmentation accuracy improvement. As the most suitable traditional segmentation method for pancreatic segmentation, the active contour model (ACM), still suffers from the problems of weak boundary leakage and slow contour evolution speed. Therefore, a convenient post-processor for any deep learning methods using superpixel-based active contour model (SbACM) is proposed to improve the segmentation accuracy. Approach. Firstly, the superpixels with strong adhesion to edges are used to guide the design of narrowband and energy function. A multi-scale evolution strategy is also proposed to reduce the weak boundary leakage and comprehensively improve the evolution speed. Secondly, using the original image and the coarse segmentation results obtained from deep learning methods as inputs, the proposed SbACM method is used as a post-processor for fine segmentation. Finally, the pancreatic segmentation public dataset TCIA from the National Institutes of Health(NIH, USA) is used for evaluation, and the Wilcoxon Test confirmed that the improvement of proposed method is statistically significant. Main results. (1) the superpixel-based narrowband shape and dynamic edge energy of the proposed SbACM work for boundary leakage reduction, as well as the multi-scale evolution strategy and dynamic narrowband width for the evolution speed improvement; (2) as a post-processor, SbACM can increase the Dice similarity coefficients (DSC) of five typical UNet-based models, including UNet, SS-UNet, PBR UNet, ResDSN, and nnUNet, 2.35% in average and 9.04% in maximum. (3) Based on the best backbone nnUNet, the proposed post-processor performs better than either adding edge awareness or adding edge loss in segmentation enhancement without increasing the complexity and training time of deep learning models. Significance. The proposed SbACM can improve segmentation accuracy with the lowest cost, especially in cases of squeezed fuzzy edges with similar neighborhood , and complex edges.

Segmentation of mature human oocytes provides interpretable and improved blastocyst outcome predictions by a machine learning model

Within the medical field of human assisted reproductive technology, a method for interpretable, non-invasive, and objective oocyte evaluation is lacking. To address this clinical gap, a workflow utilizing machine learning techniques has been developed involving automatic multi-class segmentation of two-dimensional images, morphometric analysis, and prediction of developmental outcomes of mature denuded oocytes based on feature extraction and clinical variables. Two separate models have been developed for this purpose—a model to perform multiclass segmentation, and a classifier model to classify oocytes as likely or unlikely to develop into a blastocyst (Day 5–7 embryo). The segmentation model is highly accurate at segmenting the oocyte, ensuring high-quality segmented images (masks) are utilized as inputs for the classifier model (mask model). The mask model displayed an area under the curve (AUC) of 0.63, a sensitivity of 0.51, and a specificity of 0.66 on the test set. The AUC underwent a reduction to 0.57 when features extracted from the ooplasm were removed, suggesting the ooplasm holds the information most pertinent to oocyte developmental competence. The mask model was further compared to a deep learning model, which also utilized the segmented images as inputs. The performance of both models combined in an ensemble model was evaluated, showing an improvement (AUC 0.67) compared to either model alone. The results of this study indicate that direct assessments of the oocyte are warranted, providing the first objective insights into key features for developmental competence, a step above the current standard of care—solely utilizing oocyte age as a proxy for quality.

PFMNet: Prototype-based feature mapping network for few-shot domain adaptation in medical image segmentation

Lack of data is one of the biggest hurdles for rare disease research using deep learning. Due to the lack of rare-disease images and annotations, training a robust network for automatic rare-disease image segmentation is very challenging. To address this challenge, few-shot domain adaptation (FSDA) has emerged as a practical research direction, aiming to leverage a limited number of annotated images from a target domain to facilitate adaptation of models trained on other large datasets in a source domain. In this paper, we present a novel prototype-based feature mapping network (PFMNet) designed for FSDA in medical image segmentation. PFMNet adopts an encoder–decoder structure for segmentation, with the prototype-based feature mapping (PFM) module positioned at the bottom of the encoder–decoder structure. The PFM module transforms high-level features from the target domain into the source domain-like features that are more easily comprehensible by the decoder. By leveraging these source domain-like features, the decoder can effectively perform few-shot segmentation in the target domain and generate accurate segmentation masks. We evaluate the performance of PFMNet through experiments on three typical yet challenging few-shot medical image segmentation tasks: cross-center optic disc/cup segmentation, cross-center polyp segmentation, and cross-modality cardiac structure segmentation. We consider four different settings: 5-shot, 10-shot, 15-shot, and 20-shot. The experimental results substantiate the efficacy of our proposed approach for few-shot domain adaptation in medical image segmentation.

Deep Learning for Automated Image Segmentation of the Middle Ear: A Scoping Review.

Convolutional neural networks (CNNs) have revolutionized medical image segmentation in recent years. This scoping review aimed to carry out a comprehensive review of the literature describing automated image segmentation of the middle ear using CNNs from computed tomography (CT) scans. A comprehensive literature search, generated jointly with a medical librarian, was performed on Medline, Embase, Scopus, Web of Science, and Cochrane, using Medical Subject Headingterms and keywords. Databases were searched from inception to July 2023. Reference lists of included papers were also screened. Ten studies were included for analysis, which contained a total of 866 scans which were used in model training/testing. Thirteen different architectures were described to perform automated segmentation. The best Dice similarity coefficient (DSC) for the entire ossicular chain was 0.87 using ResNet. The highest DSC for any structure was the incus using 3D-V-Net at 0.93. The most difficult structure to segment was the stapes, with the highest DSC of 0.84 using 3D-V-Net. Numerous architectures have demonstrated good performance in segmenting the middle ear using CNNs. To overcome some of the difficulties in segmenting the stapes, we recommend the development of an architecture trained on cone beam CTs to provide improved spatial resolution to assist with delineating the smallest ossicle. This has clinical applications for preoperative planning, diagnosis, and simulation.

Three-Dimensional Reconstruction Method of Rail Corrugation Based on the Vision Detection

In order to realize the visual identification of the three-dimensional geometric characteristics of rail corrugation, the three-dimensional reconstruction method of rail corrugation based on vision detection is developed in the present paper. The acquisition equipment is designed to acquire the images of rail corrugation and corresponding depth information. Then, the image preprocessing method is proposed to achieve the automatic segmentation of rail corrugation images. The fitting equation between the image and depth of rail corrugation is obtained, which allows the three-dimensional reconstruction of rail corrugation. Furthermore, based on the finite element model of the wheel-rail system and the wear calculation model, the practicality and effectiveness of the three-dimensional reconstruction method are verified. Results show that the depth-grayscale correlation equation is well fitted with the coefficient of determination of 0.94 and the root mean square error of 0.02. The three-dimensional reconstruction method has an error of 2.63% at the maximum wear depth. When the three-dimensional reconstruction method is applied to the wear prediction of rail corrugation, the wear prediction of rail corrugation with the three-dimensional reconstruction is approximately the same as the actual wear distribution, which has an error of 5.02% at the maximum wear depth. The aforementioned results demonstrate the effectiveness of the three-dimensional reconstruction method of rail corrugation based on the vision detection.