Fully automated measurement on coronal alignment of lower limbs using deep convolutional neural networks on radiographic images.

BACKGROUND: A deep convolutional neural network (DCNN) system is proposed to measure the lower limb parameters of the mechanical lateral distal femur angle (mLDFA), medial proximal tibial angle (MPTA), lateral distal tibial angle (LDTA), joint line convergence angle (JLCA), and mechanical axis of the lower limbs. METHODS: Standing X-rays of 1000 patients' lower limbs were examined for the DCNN and assigned to training, validation, and test sets. A coarse-to-fine network was employed to locate 20 key landmarks on both limbs that first recognised the regions of hip, knee, and ankle, and subsequently outputted the key points in each sub-region from a full-length X-ray. Finally, information from these key landmark locations was used to calculate the above five parameters. RESULTS: The DCNN system showed high consistency (intraclass correlation coefficient > 0.91) for all five lower limb parameters. Additionally, the mean absolute error (MAE) and root mean squared error (RMSE) of all angle predictions were lower than 3° for both the left and right limbs. The MAE of the mechanical axis of the lower limbs was 1.124 mm and 1.416 mm and the RMSE was 1.032 mm and 1.321 mm, for the right and left limbs, respectively. The measurement time of the DCNN system was 1.8 ± 1.3 s, which was significantly shorter than that of experienced radiologists (616.8 ± 48.2 s, t = -180.4, P < 0.001). CONCLUSIONS: The proposed DCNN system can automatically measure mLDFA, MPTA, LDTA, JLCA, and the mechanical axis of the lower limbs, thus helping physicians manage lower limb alignment accurately and efficiently.

Radiomics-based machine learning analysis and characterization of breast lesions with multiparametric diffusion-weighted MR.

BACKGROUND: This study aimed to evaluate the utility of radiomics-based machine learning analysis with multiparametric DWI and to compare the diagnostic performance of radiomics features and mean diffusion metrics in the characterization of breast lesions. METHODS: This retrospective study included 542 lesions from February 2018 to November 2018. One hundred radiomics features were computed from mono-exponential (ME), biexponential (BE), stretched exponential (SE), and diffusion-kurtosis imaging (DKI). Radiomics-based analysis was performed by comparing four classifiers, including random forest (RF), principal component analysis (PCA), L1 regularization (L1R), and support vector machine (SVM). These four classifiers were trained on a training set with 271 patients via ten-fold cross-validation and tested on an independent testing set with 271 patients. The diagnostic performance of the mean diffusion metrics of ME (mADCall b, mADC0-1000), BE (mD, mD*, mf), SE (mDDC, mα), and DKI (mK, mD) were also calculated for comparison. The area under the receiver operating characteristic curve (AUC) was used to compare the diagnostic performance. RESULTS: RF attained higher AUCs than L1R, PCA and SVM. The AUCs of radiomics features for the differential diagnosis of breast lesions ranged from 0.80 (BE_D*) to 0.85 (BE_D). The AUCs of the mean diffusion metrics ranged from 0.54 (BE_mf) to 0.79 (ME_mADC0-1000). There were significant differences in the AUCs between the mean values of all diffusion metrics and radiomics features of AUCs (all P < 0.001) for the differentiation of benign and malignant breast lesions. Of the radiomics features computed, the most important sequence was BE_D (AUC: 0.85), and the most important feature was FO-10 percentile (Feature Importance: 0.04). CONCLUSIONS: The radiomics-based analysis of multiparametric DWI by RF enables better differentiation of benign and malignant breast lesions than the mean diffusion metrics.

Automated segmentation and diagnosis of pneumothorax on chest X-rays with fully convolutional multi-scale ScSE-DenseNet: a retrospective study.

BACKGROUND: Pneumothorax (PTX) may cause a life-threatening medical emergency with cardio-respiratory collapse that requires immediate intervention and rapid treatment. The screening and diagnosis of pneumothorax usually rely on chest radiographs. However, the pneumothoraces in chest X-rays may be very subtle with highly variable in shape and overlapped with the ribs or clavicles, which are often difficult to identify. Our objective was to create a large chest X-ray dataset for pneumothorax with pixel-level annotation and to train an automatic segmentation and diagnosis framework to assist radiologists to identify pneumothorax accurately and timely. METHODS: In this study, an end-to-end deep learning framework is proposed for the segmentation and diagnosis of pneumothorax on chest X-rays, which incorporates a fully convolutional DenseNet (FC-DenseNet) with multi-scale module and spatial and channel squeezes and excitation (scSE) modules. To further improve the precision of boundary segmentation, we propose a spatial weighted cross-entropy loss function to penalize the target, background and contour pixels with different weights. RESULTS: This retrospective study are conducted on a total of eligible 11,051 front-view chest X-ray images (5566 cases of PTX and 5485 cases of Non-PTX). The experimental results show that the proposed algorithm outperforms the five state-of-the-art segmentation algorithms in terms of mean pixel-wise accuracy (MPA) with [Formula: see text] and dice similarity coefficient (DSC) with [Formula: see text], and achieves competitive performance on diagnostic accuracy with 93.45% and [Formula: see text]-score with 92.97%. CONCLUSION: This framework provides substantial improvements for the automatic segmentation and diagnosis of pneumothorax and is expected to become a clinical application tool to help radiologists to identify pneumothorax on chest X-rays.

Automatic Segmentation of the Corpus Callosum Using a Cell-Competition Algorithm: Diffusion Tensor Imaging-Based Evaluation of Callosal Atrophy and Tissue Alterations in Patients With Systemic Lupus Erythematosus.

OBJECTIVE: Patients with neuropsychiatric systemic lupus erythematosus (NPSLE) may exhibit corpus callosal atrophy and tissue alterations. Measuring the callosal volume and tissue integrity using diffusion tensor imaging (DTI) could help to differentiate patients with NPSLE from patients without NPSLE. Hence, this study aimed to use an automatic cell-competition algorithm to segment the corpus callosum and to investigate the effects of central nervous system (CNS) involvement on the callosal volume and tissue integrity in patients with SLE. METHODS: Twenty-two SLE patients with (N = 10, NPSLE) and without (N = 12, non-NPSLE) CNS involvement and 22 control subjects were enrolled in this study. For volumetric measurement, a cell-competition algorithm was used to automatically delineate corpus callosal boundaries based on a midsagittal fractional anisotropy (FA) map. After obtaining corpus callosal boundaries for all subjects, the volume, FA, and mean diffusivity (MD) of the corpus callosum were calculated. A post hoc Tamhane's T2 analysis was performed to statistically compare differences among NPSLE, non-NPSLE, and control subjects. A receiver operating characteristic curve analysis was also performed to compare the performance of the volume, FA, and MD of the corpus callosum in differentiating NPSLE from other subjects. RESULTS: Patients with NPSLE had significant decreases in volume and FA but an increase in MD in the corpus callosum compared with control subjects, whereas no significant difference was noted between patients without NPSLE and control subjects. The FA was found to have better performance in differentiating NPSLE from other subjects. CONCLUSIONS: A cell-competition algorithm could be used to automatically evaluate callosal atrophy and tissue alterations. Assessments of the corpus callosal volume and tissue integrity helped to demonstrate the effects of CNS involvement in patients with SLE.

Image synthesis with disentangled attributes for chest X-ray nodule augmentation and detection.

Lung nodule detection in chest X-ray (CXR) images is common to early screening of lung cancers. Deep-learning-based Computer-Assisted Diagnosis (CAD) systems can support radiologists for nodule screening in CXR images. However, it requires large-scale and diverse medical data with high-quality annotations to train such robust and accurate CADs. To alleviate the limited availability of such datasets, lung nodule synthesis methods are proposed for the sake of data augmentation. Nevertheless, previous methods lack the ability to generate nodules that are realistic with the shape/size attributes desired by the detector. To address this issue, we introduce a novel lung nodule synthesis framework in this paper, which decomposes nodule attributes into three main aspects including the shape, the size, and the texture, respectively. A GAN-based Shape Generator firstly models nodule shapes by generating diverse shape masks. The following Size Modulation then enables quantitative control on the diameters of the generated nodule shapes in pixel-level granularity. A coarse-to-fine gated convolutional Texture Generator finally synthesizes visually plausible nodule textures conditioned on the modulated shape masks. Moreover, we propose to synthesize nodule CXR images by controlling the disentangled nodule attributes for data augmentation, in order to better compensate for the nodules that are easily missed in the detection task. Our experiments demonstrate the enhanced image quality, diversity, and controllability of the proposed lung nodule synthesis framework. We also validate the effectiveness of our data augmentation strategy on greatly improving nodule detection performance.

Segment aorta and localize landmarks simultaneously on noncontrast CT using a multitask learning framework for patients without severe vascular disease.

BACKGROUND: Non-contrast chest CT is widely used for lung cancer screening, and its images carry potential information of the thoracic aorta. The morphological assessment of the thoracic aorta may have potential value in the presymptomatic detection of thoracic aortic-related diseases and the risk prediction of future adverse events. However, due to low vasculature contrast in such images, visual assessment of aortic morphology is challenging and highly depends on physicians' experience. PURPOSE: The main objective of this study is to propose a novel multi-task framework based on deep learning for simultaneous aortic segmentation and localization of key landmarks on unenhanced chest CT. The secondary objective is to use the algorithm to measure quantitative features of thoracic aorta morphology. METHODS: The proposed network is composed of two subnets to carry out segmentation and landmark detection, respectively. The segmentation subnet aims to demarcate the aortic sinuses of the Valsalva, aortic trunk and aortic branches, whereas the detection subnet is devised to locate five landmarks on the aorta to facilitate morphology measures. The networks share a common encoder and run decoders in parallel, taking full advantage of the synergy of the segmentation and landmark detection tasks. Furthermore, the volume of interest (VOI) module and the squeeze-and-excitation (SE) block with attention mechanisms are incorporated to further boost the capability of feature learning. RESULTS: Benefiting from the multitask framework, we achieved a mean Dice score of 0.95, average symmetric surface distance of 0.53 mm, Hausdorff distance of 2.13 mm for aortic segmentation, and mean square error (MSE) of 3.23 mm for landmark localization in 40 testing cases. CONCLUSION: We proposed a multitask learning framework which can perform segmentation of the thoracic aorta and localization of landmarks simultaneously and achieved good results. It can support quantitative measurement of aortic morphology for further analysis of aortic diseases, such as hypertension.

Stochastic region competition algorithm for Doppler sonography segmentation.

PURPOSE: A probabilistic image segmentation algorithm called stochastic region competition is proposed for performing Doppler sonography segmentation. METHODS: The image segmentation is conducted by maximizing a posteriori that models histogram likelihood, gradient likelihood, and a spatial prior. The optimization is done by a modified expectation and maximization (EM) method that aims to improve computation efficiency and avoid local optima. RESULTS: The algorithm was tested on 155 color Doppler sonograms and compared with manual delineations. The qualitative assessment shows that our algorithm is able to segment mass lesions under the condition of low image quality and the interference of the color-encoded Doppler information. The quantitative assessment analysis shows that the average distance between the algorithm-generated boundaries and manual delineations is statistically comparable to the variability of manual delineations. The ratio of the overlapping area between the algorithm-generated boundaries and manual delineations is also comparable to that between different sets of manual delineations. A reproductivity test was conducted to confirm that the result is statistically reproducible. CONCLUSIONS: The algorithm can be used to perform Doppler sonography segmentation and to replace the tedious manual delineation task in clinical application.

Cascaded-Recalibrated Multiple Instance Deep Model for Pathologic-Level Lung Cancer Prediction in CT Images.

Lung cancer accounts for the greatest number of cancer-related mortality, while the accurate evaluation of pulmonary nodules in computed tomography (CT) images can significantly increase the 5-year relative survival rate. Despite deep learning methods that have recently been introduced to the identification of malignant nodules, a substantial challenge remains due to the limited datasets. In this study, we propose a cascaded-recalibrated multiple instance learning (MIL) model based on multiattribute features transfer for pathologic-level lung cancer prediction in CT images. This cascaded-recalibrated MIL deep model incorporates a cascaded recalibration mechanism at the nodule level and attribute level, which fuses the informative attribute features into nodule embeddings and then the key nodule features can be converged into the patient-level embedding to improve the performance of lung cancer prediction. We evaluated the proposed cascaded-recalibrated MIL model on the public Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) benchmark dataset and compared it to the latest approaches. The experimental results showed a significant performance boost by the cascaded-recalibrated MIL model over the higher-order transfer learning, instance-space MIL, and embedding-space MIL models and the radiologists. In addition, the recalibration coefficients of the nodule and attribute feature for the final decision were also analyzed to reveal the underlying relationship between the confirmed diagnosis and its highly-correlated attributes. The cascaded recalibration mechanism enables the MIL model to pay more attention to those important nodules and attributes while suppressing less-useful feature embeddings, and the cascaded-recalibrated MIL model provides substantial improvements for the pathologic-level lung cancer prediction by using the CT images. The identification of the important nodules and attributes also provides better interpretability for model decision-making, which is very important for medical applications.

CENTS: cortical enhanced neonatal tissue segmentation.

The acquisition of high-quality magnetic resonance (MR) images of neonatal brains is largely hampered by their characteristically small head size and insufficient tissue contrast. As a result, subsequent image processing and analysis, especially brain tissue segmentation, are often affected. To overcome this problem, a dedicated phased array neonatal head coil is utilized to improve MR image quality by augmenting signal-to-noise ratio and spatial resolution without lengthening data acquisition time. In addition, a specialized hybrid atlas-based tissue segmentation algorithm is developed for the delineation of fine structures in the acquired neonatal brain MR images. The proposed tissue segmentation method first enhances the sheet-like cortical gray matter (GM) structures in the to-be-segmented neonatal image with a Hessian filter for generation of a cortical GM confidence map. A neonatal population atlas is then generated by averaging the presegmented images of a population, weighted by their cortical GM similarity with respect to the to-be-segmented image. Finally, the neonatal population atlas is combined with the GM confidence map, and the resulting enhanced tissue probability maps for each tissue form a hybrid atlas is used for atlas-based segmentation. Various experiments are conducted to compare the segmentations of the proposed method with manual segmentation (on both images acquired with a dedicated phased array coil and a conventional volume coil), as well as with the segmentations of two population-atlas-based methods. Results show the proposed method is capable of segmenting the neonatal brain with the best accuracy, and also preserving the most structural details in the cortical regions.

Learning Hierarchical Attention for Weakly-Supervised Chest X-Ray Abnormality Localization and Diagnosis.

We consider the problem of abnormality localization for clinical applications. While deep learning has driven much recent progress in medical imaging, many clinical challenges are not fully addressed, limiting its broader usage. While recent methods report high diagnostic accuracies, physicians have concerns trusting these algorithm results for diagnostic decision-making purposes because of a general lack of algorithm decision reasoning and interpretability. One potential way to address this problem is to further train these models to localize abnormalities in addition to just classifying them. However, doing this accurately will require a large amount of disease localization annotations by clinical experts, a task that is prohibitively expensive to accomplish for most applications. In this work, we take a step towards addressing these issues by means of a new attention-driven weakly supervised algorithm comprising a hierarchical attention mining framework that unifies activation- and gradient-based visual attention in a holistic manner. Our key algorithmic innovations include the design of explicit ordinal attention constraints, enabling principled model training in a weakly-supervised fashion, while also facilitating the generation of visual-attention-driven model explanations by means of localization cues. On two large-scale chest X-ray datasets (NIH ChestX-ray14 and CheXpert), we demonstrate significant localization performance improvements over the current state of the art while also achieving competitive classification performance.

Computer-aided US diagnosis of breast lesions by using cell-based contour grouping.

PURPOSE: To develop a computer-aided diagnostic algorithm with automatic boundary delineation for differential diagnosis of benign and malignant breast lesions at ultrasonography (US) and investigate the effect of boundary quality on the performance of a computer-aided diagnostic algorithm. MATERIALS AND METHODS: This was an institutional review board-approved retrospective study with waiver of informed consent. A cell-based contour grouping (CBCG) segmentation algorithm was used to delineate the lesion boundaries automatically. Seven morphologic features were extracted. The classifier was a logistic regression function. Five hundred twenty breast US scans were obtained from 520 subjects (age range, 15-89 years), including 275 benign (mean size, 15 mm; range, 5-35 mm) and 245 malignant (mean size, 18 mm; range, 8-29 mm) lesions. The newly developed computer-aided diagnostic algorithm was evaluated on the basis of boundary quality and differentiation performance. The segmentation algorithms and features in two conventional computer-aided diagnostic algorithms were used for comparative study. RESULTS: The CBCG-generated boundaries were shown to be comparable with the manually delineated boundaries. The area under the receiver operating characteristic curve (AUC) and differentiation accuracy were 0.968 +/- 0.010 and 93.1% +/- 0.7, respectively, for all 520 breast lesions. At the 5% significance level, the newly developed algorithm was shown to be superior to the use of the boundaries and features of the two conventional computer-aided diagnostic algorithms in terms of AUC (0.974 +/- 0.007 versus 0.890 +/- 0.008 and 0.788 +/- 0.024, respectively). CONCLUSION: The newly developed computer-aided diagnostic algorithm that used a CBCG segmentation method to measure boundaries achieved a high differentiation performance.

ACCOMP: Augmented cell competition algorithm for breast lesion demarcation in sonography.

PURPOSE: Fully automatic and high-quality demarcation of sonographical breast lesions remains a far-reaching goal. This article aims to develop an image segmentation algorithm that provides quality delineation of breast lesions in sonography with a simple and friendly semiautomatic scheme. METHODS: A data-driven image segmentation algorithm, named as augmented cell competition (ACCOMP) algorithm, is developed to delineate breast lesion boundaries in ultrasound images. Inspired by visual perceptual experience and Gestalt principles, the ACCOMP algorithm is constituted of two major processes, i.e., cell competition and cell-based contour grouping. The cell competition process drives cells, i.e., the catchment basins generated by a two-pass watershed transformation, to merge and split into prominent components. A prominent component is defined as a relatively large and homogeneous region circumscribed by a perceivable boundary. Based on the prominent component tessellation, cell-based contour grouping process seeks the best closed subsets of edges in the prominent component structure as the desirable boundary candidates. Finally, five boundary candidates with respect to five devised boundary cost functions are suggested by the ACCOMP algorithm for user selection. To evaluate the efficacy of the ACCOMP algorithm on breast lesions with complicated echogenicity and shapes, 324 breast sonograms, including 199 benign and 125 malignant lesions, are adopted as testing data. The boundaries generated by the ACCOMP algorithm are compared to manual delineations, which were confirmed by four experienced medical doctors. Four assessment metrics, including the modified Williams index, percentage statistic, overlapping ratio, and difference ratio, are employed to see if the ACCOMP-generated boundaries are comparable to manual delineations. A comparative study is also conducted by implementing two pixel-based segmentation algorithms. The same four assessment metrics are employed to evaluate the boundaries generated by two conventional pixel-based algorithms based on the same set of manual delineations. RESULTS: The ACCOMP-generated boundaries are shown to be comparable to the manual delineations. Particularly, the modified Williams indices of the boundaries generated by the ACCOMP algorithm and the first and second pixel-based algorithms are 1.069 +/- 0.024, 0.935 +/- 0.024, and 0.579 +/- 0.013, respectively. If the modified Williams index is greater than or equal to 1, the average distance between the computer-generated boundaries and manual delineations is deemed to be comparable to that between the manual delineations. CONCLUSIONS: The boundaries derived by the ACCOMP algorithm are shown to reasonably demarcate sonographic breast lesions, especially for the cases with complicated echogenicity and shapes. It suggests that the ACCOMP-generated boundaries can potentially serve as the basis for further morphological or quantitative analysis.

Self-co-attention neural network for anatomy segmentation in whole breast ultrasound.

The automated whole breast ultrasound (AWBUS) is a new breast imaging technique that can depict the whole breast anatomy. To facilitate the reading of AWBUS images and support the breast density estimation, an automatic breast anatomy segmentation method for AWBUS images is proposed in this study. The problem at hand is quite challenging as it needs to address issues of low image quality, ill-defined boundary, large anatomical variation, etc. To address these issues, a new deep learning encoder-decoder segmentation method based on a self-co-attention mechanism is developed. The self-attention mechanism is comprised of spatial and channel attention module (SC) and embedded in the ResNeXt (i.e., Res-SC) block in the encoder path. A non-local context block (NCB) is further incorporated to augment the learning of high-level contextual cues. The decoder path of the proposed method is equipped with the weighted up-sampling block (WUB) to attain class-specific better up-sampling effect. Meanwhile, the co-attention mechanism is also developed to improve the segmentation coherence among two consecutive slices. Extensive experiments are conducted with comparison to several the state-of-the-art deep learning segmentation methods. The experimental results corroborate the effectiveness of the proposed method on the difficult breast anatomy segmentation problem on AWBUS images.

Dual-domain convolutional neural networks for improving structural information in 3 T MRI.

We propose a novel dual-domain convolutional neural network framework to improve structural information of routine 3 T images. We introduce a parameter-efficient butterfly network that involves two complementary domains: a spatial domain and a frequency domain. The butterfly network allows the interaction of these two domains in learning the complex mapping from 3 T to 7 T images. We verified the efficacy of the dual-domain strategy and butterfly network using 3 T and 7 T image pairs. Experimental results demonstrate that the proposed framework generates synthetic 7 T-like images and achieves performance superior to state-of-the-art methods.

Corrections to "Accurate Cervical Cell Segmentation From Overlapping Clumps in Pap Smear Images" [Jan 17 288-300].

In [1], Baiying Lei was indicated as the corresponding author. Tianfu Wang and Baiying Lei should have been indicated as the corresponding authors.

Dual-Domain Cascaded Regression for Synthesizing 7T from 3T MRI.

Due to the high cost and low accessibility of 7T magnetic resonance imaging (MRI) scanners, we propose a novel dual-domain cascaded regression framework to synthesize 7T images from the routine 3T images. Our framework is composed of two parallel and interactive multi-stage regression streams, where one stream regresses on spatial domain and the other regresses on frequency domain. These two streams complement each other and enable the learning of complex mappings between 3T and 7T images. We evaluated the proposed framework on a set of 3T and 7T images by leave-one-out cross-validation. Experimental results demonstrate that the proposed framework generates realistic 7T images and achieves better results than state-of-the-art methods.

Automatic Fetal Head Circumference Measurement in Ultrasound Using Random Forest and Fast Ellipse Fitting.

Head circumference (HC) is one of the most important biometrics in assessing fetal growth during prenatal ultrasound examinations. However, the manual measurement of this biometric by doctors often requires substantial experience. We developed a learning-based framework that used prior knowledge and employed a fast ellipse fitting method (ElliFit) to measure HC automatically. We first integrated the prior knowledge about the gestational age and ultrasound scanning depth into a random forest classifier to localize the fetal head. We further used phase symmetry to detect the center line of the fetal skull and employed ElliFit to fit the HC ellipse for measurement. The experimental results from 145 HC images showed that our method had an average measurement error of 1.7 mm and outperformed traditional methods. The experimental results demonstrated that our method shows great promise for applications in clinical practice.

Cell-based two-region competition algorithm with a map framework for boundary delineation of a series of 2D ultrasound images.

To ensure the delineated boundaries of a series of 2-D images closely following the visually perceivable edges with high boundary coherence between consecutive slices, a cell-based two-region competition algorithm based on a maximum a posteriori (MAP) framework is proposed. It deforms the region boundary in a cell-by-cell fashion through a cell-based two-region competition process. The cell-based deformation is guided by a cell-based MAP framework with a posterior function characterizing the distribution of the cell means in each region, the salience and shape complexity of the region boundary and the boundary coherence of the consecutive slices. The proposed algorithm has been validated using 10 series of breast sonograms, including seven compression series and three freehand series. The compression series contains two carcinoma and five fibroadenoma cases and the freehand series contains two carcinoma and one fibroadenoma cases. The results show that >70% of the derived boundaries fall within the span of the manually delineated boundaries. The robustness of the proposed algorithm to the variation of regions-of-interest is confirmed by the Friedman tests and the p-values of which are 0.517 and 0.352 for the compression and freehand series groups, respectively. The Pearson's correlations between the lesion sizes derived by the proposed algorithm and those defined by the average manually delineated boundaries are all higher than 0.990. The overlapping and difference ratios between the derived boundaries and the average manually delineated boundaries are mostly higher than 0.90 and lower than 0.13, respectively. For both series groups, all assessments conclude that the boundaries derived by the proposed algorithm be comparable to those delineated manually. Moreover, it is shown that the proposed algorithm is superior to the Chan and Vese level set method based on the paired-sample t-tests on the performance indices at a 5% significance level.

Accurate Cervical Cell Segmentation from Overlapping Clumps in Pap Smear Images.

Accurate segmentation of cervical cells in Pap smear images is an important step in automatic pre-cancer identification in the uterine cervix. One of the major segmentation challenges is overlapping of cytoplasm, which has not been well-addressed in previous studies. To tackle the overlapping issue, this paper proposes a learning-based method with robust shape priors to segment individual cell in Pap smear images to support automatic monitoring of changes in cells, which is a vital prerequisite of early detection of cervical cancer. We define this splitting problem as a discrete labeling task for multiple cells with a suitable cost function. The labeling results are then fed into our dynamic multi-template deformation model for further boundary refinement. Multi-scale deep convolutional networks are adopted to learn the diverse cell appearance features. We also incorporated high-level shape information to guide segmentation where cell boundary might be weak or lost due to cell overlapping. An evaluation carried out using two different datasets demonstrates the superiority of our proposed method over the state-of-the-art methods in terms of segmentation accuracy.

Ultrasound Standard Plane Detection Using a Composite Neural Network Framework.

Ultrasound (US) imaging is a widely used screening tool for obstetric examination and diagnosis. Accurate acquisition of fetal standard planes with key anatomical structures is very crucial for substantial biometric measurement and diagnosis. However, the standard plane acquisition is a labor-intensive task and requires operator equipped with a thorough knowledge of fetal anatomy. Therefore, automatic approaches are highly demanded in clinical practice to alleviate the workload and boost the examination efficiency. The automatic detection of standard planes from US videos remains a challenging problem due to the high intraclass and low interclass variations of standard planes, and the relatively low image quality. Unlike previous studies which were specifically designed for individual anatomical standard planes, respectively, we present a general framework for the automatic identification of different standard planes from US videos. Distinct from conventional way that devises hand-crafted visual features for detection, our framework explores in- and between-plane feature learning with a novel composite framework of the convolutional and recurrent neural networks. To further address the issue of limited training data, a multitask learning framework is implemented to exploit common knowledge across detection tasks of distinctive standard planes for the augmentation of feature learning. Extensive experiments have been conducted on hundreds of US fetus videos to corroborate the better efficacy of the proposed framework on the difficult standard plane detection problem.