Showing papers in "BMC Medical Imaging in 2022"

Transfer learning for medical image classification: a literature review

Abstract: Transfer learning (TL) with convolutional neural networks aims to improve performances on a new task by leveraging the knowledge of similar tasks learned in advance. It has made a major contribution to medical image analysis as it overcomes the data scarcity problem as well as it saves time and hardware resources. However, transfer learning has been arbitrarily configured in the majority of studies. This review paper attempts to provide guidance for selecting a model and TL approaches for the medical image classification task.425 peer-reviewed articles were retrieved from two databases, PubMed and Web of Science, published in English, up until December 31, 2020. Articles were assessed by two independent reviewers, with the aid of a third reviewer in the case of discrepancies. We followed the PRISMA guidelines for the paper selection and 121 studies were regarded as eligible for the scope of this review. We investigated articles focused on selecting backbone models and TL approaches including feature extractor, feature extractor hybrid, fine-tuning and fine-tuning from scratch.The majority of studies (n = 57) empirically evaluated multiple models followed by deep models (n = 33) and shallow (n = 24) models. Inception, one of the deep models, was the most employed in literature (n = 26). With respect to the TL, the majority of studies (n = 46) empirically benchmarked multiple approaches to identify the optimal configuration. The rest of the studies applied only a single approach for which feature extractor (n = 38) and fine-tuning from scratch (n = 27) were the two most favored approaches. Only a few studies applied feature extractor hybrid (n = 7) and fine-tuning (n = 3) with pretrained models.The investigated studies demonstrated the efficacy of transfer learning despite the data scarcity. We encourage data scientists and practitioners to use deep models (e.g. ResNet or Inception) as feature extractors, which can save computational costs and time without degrading the predictive power.

3D AGSE-VNet: an automatic brain tumor MRI data segmentation framework

Abstract: Glioma is the most common brain malignant tumor, with a high morbidity rate and a mortality rate of more than three percent, which seriously endangers human health. The main method of acquiring brain tumors in the clinic is MRI. Segmentation of brain tumor regions from multi-modal MRI scan images is helpful for treatment inspection, post-diagnosis monitoring, and effect evaluation of patients. However, the common operation in clinical brain tumor segmentation is still manual segmentation, lead to its time-consuming and large performance difference between different operators, a consistent and accurate automatic segmentation method is urgently needed. With the continuous development of deep learning, researchers have designed many automatic segmentation algorithms; however, there are still some problems: (1) The research of segmentation algorithm mostly stays on the 2D plane, this will reduce the accuracy of 3D image feature extraction to a certain extent. (2) MRI images have gray-scale offset fields that make it difficult to divide the contours accurately.To meet the above challenges, we propose an automatic brain tumor MRI data segmentation framework which is called AGSE-VNet. In our study, the Squeeze and Excite (SE) module is added to each encoder, the Attention Guide Filter (AG) module is added to each decoder, using the channel relationship to automatically enhance the useful information in the channel to suppress the useless information, and use the attention mechanism to guide the edge information and remove the influence of irrelevant information such as noise.We used the BraTS2020 challenge online verification tool to evaluate our approach. The focus of verification is that the Dice scores of the whole tumor, tumor core and enhanced tumor are 0.68, 0.85 and 0.70, respectively.Although MRI images have different intensities, AGSE-VNet is not affected by the size of the tumor, and can more accurately extract the features of the three regions, it has achieved impressive results and made outstanding contributions to the clinical diagnosis and treatment of brain tumor patients.

Automated detection of pulmonary embolism from CT-angiograms using deep learning

Abstract: The aim of this study was to develop and evaluate a deep neural network model in the automated detection of pulmonary embolism (PE) from computed tomography pulmonary angiograms (CTPAs) using only weakly labelled training data.We developed a deep neural network model consisting of two parts: a convolutional neural network architecture called InceptionResNet V2 and a long-short term memory network to process whole CTPA stacks as sequences of slices. Two versions of the model were created using either chest X-rays (Model A) or natural images (Model B) as pre-training data. We retrospectively collected 600 CTPAs to use in training and validation and 200 CTPAs to use in testing. CTPAs were annotated only with binary labels on both stack- and slice-based levels. Performance of the models was evaluated with ROC and precision-recall curves, specificity, sensitivity, accuracy, as well as positive and negative predictive values.Both models performed well on both stack- and slice-based levels. On the stack-based level, Model A reached specificity and sensitivity of 93.5% and 86.6%, respectively, outperforming Model B slightly (specificity 90.7% and sensitivity 83.5%). However, the difference between their ROC AUC scores was not statistically significant (0.94 vs 0.91, p = 0.07).We show that a deep learning model trained with a relatively small, weakly annotated dataset can achieve excellent performance results in detecting PE from CTPAs.

Automatic lesion segmentation using atrous convolutional deep neural networks in dermoscopic skin cancer images

Abstract: Melanoma is the most dangerous and aggressive form among skin cancers, exhibiting a high mortality rate worldwide. Biopsy and histopathological analysis are standard procedures for skin cancer detection and prevention in clinical settings. A significant step in the diagnosis process is the deep understanding of the patterns, size, color, and structure of lesions based on images obtained through dermatoscopes for the infected area. However, the manual segmentation of the lesion region is time-consuming because the lesion evolves and changes its shape over time, making its prediction challenging. Moreover, it is challenging to predict melanoma at the initial stage as it closely resembles other skin cancer types that are not malignant as melanoma; thus, automatic segmentation techniques are required to design a computer-aided system for accurate and timely detection.As deep learning approaches have gained significant attention in recent years due to their remarkable performance, therefore, in this work, we proposed a novel design of a convolutional neural network (CNN) framework based on atrous convolutions for automatic lesion segmentation. This architecture is built based on the concept of atrous/dilated convolutions which are effective for semantic segmentation. A deep neural network is designed from scratch employing several building blocks consisting of convolutional, batch normalization, leakyReLU layer, and fine-tuned hyperparameters contributing altogether towards higher performance.The network was tested on three benchmark datasets provided by International Skin Imaging Collaboration (ISIC), i.e., ISIC 2016, ISIC 2017, and ISIC 2018. The experimental results showed that the proposed network achieved an average Jaccard index of 90.4% on ISIC 2016, 81.8% on ISIC 2017, and 89.1% on ISIC 2018 datasets, respectively which is recorded as higher than the top three winners of the ISIC challenge and other state-of-the-art methods. Also, the model successfully extracts lesions from the whole image in one pass in less time, requiring no pre-processing step. The conclusions yielded that network is accurate in performing lesion segmentation on adopted datasets.

Vessel segmentation for X-ray coronary angiography using ensemble methods with deep learning and filter-based features

Abstract: Automated segmentation of coronary arteries is a crucial step for computer-aided coronary artery disease (CAD) diagnosis and treatment planning. Correct delineation of the coronary artery is challenging in X-ray coronary angiography (XCA) due to the low signal-to-noise ratio and confounding background structures.A novel ensemble framework for coronary artery segmentation in XCA images is proposed, which utilizes deep learning and filter-based features to construct models using the gradient boosting decision tree (GBDT) and deep forest classifiers. The proposed method was trained and tested on 130 XCA images. For each pixel of interest in the XCA images, a 37-dimensional feature vector was constructed based on (1) the statistics of multi-scale filtering responses in the morphological, spatial, and frequency domains; and (2) the feature maps obtained from trained deep neural networks. The performance of these models was compared with those of common deep neural networks on metrics including precision, sensitivity, specificity, F1 score, AUROC (the area under the receiver operating characteristic curve), and IoU (intersection over union).With hybrid under-sampling methods, the best performing GBDT model achieved a mean F1 score of 0.874, AUROC of 0.947, sensitivity of 0.902, and specificity of 0.992; while the best performing deep forest model obtained a mean F1 score of 0.867, AUROC of 0.95, sensitivity of 0.867, and specificity of 0.993. Compared with the evaluated deep neural networks, both models had better or comparable performance for all evaluated metrics with lower standard deviations over the test images.The proposed feature-based ensemble method outperformed common deep convolutional neural networks in most performance metrics while yielding more consistent results. Such a method can be used to facilitate the assessment of stenosis and improve the quality of care in patients with CAD.

WBC image classification and generative models based on convolutional neural network

Abstract: Computer-aided methods for analyzing white blood cells (WBC) are popular due to the complexity of the manual alternatives. Recent works have shown highly accurate segmentation and detection of white blood cells from microscopic blood images. However, the classification of the observed cells is still a challenge, in part due to the distribution of the five types that affect the condition of the immune system.(i) This work proposes W-Net, a CNN-based method for WBC classification. We evaluate W-Net on a real-world large-scale dataset that includes 6562 real images of the five WBC types. (ii) For further benefits, we generate synthetic WBC images using Generative Adversarial Network to be used for education and research purposes through sharing.(i) W-Net achieves an average accuracy of 97%. In comparison to state-of-the-art methods in the field of WBC classification, we show that W-Net outperforms other CNN- and RNN-based model architectures. Moreover, we show the benefits of using pre-trained W-Net in a transfer learning context when fine-tuned to specific task or accommodating another dataset. (ii) The synthetic WBC images are confirmed by experiments and a domain expert to have a high degree of similarity to the original images. The pre-trained W-Net and the generated WBC dataset are available for the community to facilitate reproducibility and follow up research work.This work proposed W-Net, a CNN-based architecture with a small number of layers, to accurately classify the five WBC types. We evaluated W-Net on a real-world large-scale dataset and addressed several challenges such as the transfer learning property and the class imbalance. W-Net achieved an average classification accuracy of 97%. We synthesized a dataset of new WBC image samples using DCGAN, which we released to the public for education and research purposes.

Preoperative prediction of extramural venous invasion in rectal cancer by dynamic contrast-enhanced and diffusion weighted MRI: a preliminary study

Abstract: To explore the value of the quantitative dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and diffusion-weighted imaging (DWI) parameters in assessing preoperative extramural venous invasion (EMVI) in rectal cancer.Eighty-two rectal adenocarcinoma patients who had underwent MRI preoperatively were enrolled in this study. The differences in quantitative DCE-MRI and DWI parameters including Krans, Kep and ADC values were analyzed between MR-detected EMVI (mrEMVI)-positive and -negative groups. Multivariate logistic regression analysis was performed to build the combined prediction model for pathologic EMVI (pEMVI) with statistically significant quantitative parameters. The performance of the model for predicting pEMVI was evaluated using receiver operating characteristic (ROC) curve.Of the 82 patients, 24 were mrEMVI-positive and 58 were -negative. In the mrEMVI positive group, the Ktrans and Kep values were significantly higher than those in the mrEMVI negative group (P < 0.01), but the ADC values were significantly lower (P < 0.01). A negative correlation was observed between the Ktrans vs ADC values and Kep vs ADC values in patients with rectal cancer. Among the four quantitative parameters, Ktrans and ADC value were independently associated with mrEMVI by multivariate logistic regression analysis. ROC analysis showed that combined prediction model based on quantitative DCE parameters and ADC values had a good prediction efficiency for pEMVI in rectal cancer.The quantitative DCE-MRI parameters, Krans, Kep and ADC values play important role in predicting EMVI of rectal cancer, with Ktrans and ADC value being independent predictors of EMVI in rectal cancer.

A novel adaptive momentum method for medical image classification using convolutional neural network

Abstract: AI for medical diagnosis has made a tremendous impact by applying convolutional neural networks (CNNs) to medical image classification and momentum plays an essential role in stochastic gradient optimization algorithms for accelerating or improving training convolutional neural networks. In traditional optimizers in CNNs, the momentum is usually weighted by a constant. However, tuning hyperparameters for momentum can be computationally complex. In this paper, we propose a novel adaptive momentum for fast and stable convergence.Applying adaptive momentum rate proposes increasing or decreasing based on every epoch's error changes, and it eliminates the need for momentum hyperparameter optimization. We tested the proposed method with 3 different datasets: REMBRANDT Brain Cancer, NIH Chest X-ray, COVID-19 CT scan. We compared the performance of a novel adaptive momentum optimizer with Stochastic gradient descent (SGD) and other adaptive optimizers such as Adam and RMSprop.Proposed method improves SGD performance by reducing classification error from 6.12 to 5.44%, and it achieved the lowest error and highest accuracy compared with other optimizers. To strengthen the outcomes of this study, we investigated the performance comparison for the state-of-the-art CNN architectures with adaptive momentum. The results shows that the proposed method achieved the highest with 95% compared to state-of-the-art CNN architectures while using the same dataset. The proposed method improves convergence performance by reducing classification error and achieves high accuracy compared with other optimizers.

The application research of AI image recognition and processing technology in the early diagnosis of the COVID-19

Abstract: This study intends to establish a combined prediction model that integrates the clinical symptoms,the lung lesion volume, and the radiomics features of patients with COVID-19, resulting in a new model to predict the severity of COVID-19.The clinical data of 386 patients with COVID-19 at several hospitals, as well as images of certain patients during their hospitalization, were collected retrospectively to create a database of patients with COVID-19 pneumonia. The contour of lungs and lesion locations may be retrieved from CT scans using a CT-image-based quantitative discrimination and trend analysis method for COVID-19 and the Mask R-CNN deep neural network model to create 3D data of lung lesions. The quantitative COVID-19 factors were then determined, on which the diagnosis of the development of the patients' symptoms could be established. Then, using an artificial neural network, a prediction model of the severity of COVID-19 was constructed by combining characteristic imaging features on CT slices with clinical factors. ANN neural network was used for training, and tenfold cross-validation was used to verify the prediction model. The diagnostic performance of this model is verified by the receiver operating characteristic (ROC) curve.CT radiomics features extraction and analysis based on a deep neural network can detect COVID-19 patients with an 86% sensitivity and an 85% specificity. According to the ROC curve, the constructed severity prediction model indicates that the AUC of patients with severe COVID-19 is 0.761, with sensitivity and specificity of 79.1% and 73.1%, respectively.The combined prediction model for severe COVID-19 pneumonia, which is based on deep learning and integrates clinical aspects, pulmonary lesion volume, and radiomics features of patients, has a remarkable differential ability for predicting the course of disease in COVID-19 patients. This may assist in the early prevention of severe COVID-19 symptoms.

BPI-MVQA: a bi-branch model for medical visual question answering

Abstract: Visual question answering in medical domain (VQA-Med) exhibits great potential for enhancing confidence in diagnosing diseases and helping patients better understand their medical conditions. One of the challenges in VQA-Med is how to better understand and combine the semantic features of medical images (e.g., X-rays, Magnetic Resonance Imaging(MRI)) and answer the corresponding questions accurately in unlabeled medical datasets.We propose a novel Bi-branched model based on Parallel networks and Image retrieval for Medical Visual Question Answering (BPI-MVQA). The first branch of BPI-MVQA is a transformer structure based on a parallel network to achieve complementary advantages in image sequence feature and spatial feature extraction, and multi-modal features are implicitly fused by using the multi-head self-attention mechanism. The second branch is retrieving the similarity of image features generated by the VGG16 network to obtain similar text descriptions as labels.The BPI-MVQA model achieves state-of-the-art results on three VQA-Med datasets, and the main metric scores exceed the best results so far by 0.2[Formula: see text], 1.4[Formula: see text], and 1.1[Formula: see text].The evaluation results support the effectiveness of the BPI-MVQA model in VQA-Med. The design of the bi-branch structure helps the model answer different types of visual questions. The parallel network allows for multi-angle image feature extraction, a unique feature extraction method that helps the model better understand the semantic information of the image and achieve greater accuracy in the multi-classification of VQA-Med. In addition, image retrieval helps the model answer irregular, open-ended type questions from the perspective of understanding the information provided by images. The comparison of our method with state-of-the-art methods on three datasets also shows that our method can bring substantial improvement to the VQA-Med system.

A method for improving semantic segmentation using thermographic images in infants

Abstract: Regulation of temperature is clinically important in the care of neonates because it has a significant impact on prognosis. Although probes that make contact with the skin are widely used to monitor temperature and provide spot central and peripheral temperature information, they do not provide details of the temperature distribution around the body. Although it is possible to obtain detailed temperature distributions using multiple probes, this is not clinically practical. Thermographic techniques have been reported for measurement of temperature distribution in infants. However, as these methods require manual selection of the regions of interest (ROIs), they are not suitable for introduction into clinical settings in hospitals. Here, we describe a method for segmentation of thermal images that enables continuous quantitative contactless monitoring of the temperature distribution over the whole body of neonates.The semantic segmentation method, U-Net, was applied to thermal images of infants. The optimal combination of Weight Normalization, Group Normalization, and Flexible Rectified Linear Unit (FReLU) was evaluated. U-Net Generative Adversarial Network (U-Net GAN) was applied to thermal images, and a Self-Attention (SA) module was finally applied to U-Net GAN (U-Net GAN + SA) to improve precision. The semantic segmentation performance of these methods was evaluated.The optimal semantic segmentation performance was obtained with application of FReLU and Group Normalization to U-Net, showing accuracy of 92.9% and Mean Intersection over Union (mIoU) of 64.5%. U-Net GAN improved the performance, yielding accuracy of 93.3% and mIoU of 66.9%, and U-Net GAN + SA showed further improvement with accuracy of 93.5% and mIoU of 70.4%.FReLU and Group Normalization are appropriate semantic segmentation methods for application to neonatal thermal images. U-Net GAN and U-Net GAN + SA significantly improved the mIoU of segmentation.

Practical utility of liver segmentation methods in clinical surgeries and interventions

Abstract: Clinical imaging (e.g., magnetic resonance imaging and computed tomography) is a crucial adjunct for clinicians, aiding in the diagnosis of diseases and planning of appropriate interventions. This is especially true in malignant conditions such as hepatocellular carcinoma (HCC), where image segmentation (such as accurate delineation of liver and tumor) is the preliminary step taken by the clinicians to optimize diagnosis, staging, and treatment planning and intervention (e.g., transplantation, surgical resection, radiotherapy, PVE, embolization, etc). Thus, segmentation methods could potentially impact the diagnosis and treatment outcomes. This paper comprehensively reviews the literature (during the year 2012-2021) for relevant segmentation methods and proposes a broad categorization based on their clinical utility (i.e., surgical and radiological interventions) in HCC. The categorization is based on the parameters such as precision, accuracy, and automation.

Accuracy of advanced deep learning with tensorflow and keras for classifying teeth developmental stages in digital panoramic imaging

Abstract: This study aims to propose the combinations of image processing and machine learning model to segment the maturity development of the mandibular premolars using a Keras-based deep learning convolutional neural networks (DCNN) model.A dataset consisting of 240 images (20 images per stage per sex) of retrospect digital dental panoramic imaging of patients between 5 and 14 years of age was retrieved. In image preprocessing, abounding box with a dimension of 250 × 250 pixels was assigned to the left mandibular first (P1) and second (P2) permanent premolars. The implementation of dynamic programming of active contour (DP-AC) and convolutions neural network on images that require the procedure of image filtration using Python TensorFlow and Keras libraries were performed in image segmentation and classification, respectively.Image segmentation using the DP-AC algorithm enhanced the visibility of the image features in the region of interest while suppressing the image's background noise. The proposed model has an accuracy of 97.74%, 96.63% and 78.13% on the training, validation, and testing set, respectively. In addition, moderate agreement (Kappa value = 0.58) between human observer and computer were identified. Nonetheless, a robust DCNN model was achieved as there is no sign of the model's over-or under-fitting upon the learning process.The application of digital imaging and deep learning techniques used by the DP-AC and convolutions neural network algorithms to segment and identify premolars provides promising results for semi-automated forensic dental staging in the future.

Characterizing and quantifying low-value diagnostic imaging internationally: a scoping review

Abstract: Inappropriate and wasteful use of health care resources is a common problem, constituting 10-34% of health services spending in the western world. Even though diagnostic imaging is vital for identifying correct diagnoses and administrating the right treatment, low-value imaging-in which the diagnostic test confers little to no clinical benefit-is common and contributes to inappropriate and wasteful use of health care resources. There is a lack of knowledge on the types and extent of low-value imaging. Accordingly, the objective of this study was to identify, characterize, and quantify the extent of low-value diagnostic imaging examinations for adults and children.A scoping review of the published literature was performed. Medline-Ovid, Embase-Ovid, Scopus, and Cochrane Library were searched for studies published from 2010 to September 2020. The search strategy was built from medical subject headings (Mesh) for Diagnostic imaging/Radiology OR Health service misuse/Medical overuse OR Procedures and Techniques Utilization/Facilities and Services Utilization. Articles in English, German, Dutch, Swedish, Danish, or Norwegian were included.A total of 39,986 records were identified and, of these, 370 studies were included in the final synthesis. Eighty-four low-value imaging examinations were identified. Imaging of atraumatic pain, routine imaging in minor head injury, trauma, thrombosis, urolithiasis, after thoracic interventions, fracture follow-up and cancer staging/follow-up were the most frequently identified low-value imaging examinations. The proportion of low-value imaging varied between 2 and 100% inappropriate or unnecessary examinations.A comprehensive list of identified low-value radiological examinations for both adults and children are presented. Future research should focus on reasons for low-value imaging utilization and interventions to reduce the use of low-value imaging internationally.PROSPERO: CRD42020208072.

The value of radiomics based on dual-energy CT for differentiating benign from malignant solitary pulmonary nodules

Abstract: To investigate the value of monochromatic dual-energy CT (DECT) images based on radiomics in differentiating benign from malignant solitary pulmonary nodules.This retrospective study was approved by the institutional review board, and informed consent was waived. Pathologically confirmed lung nodules smaller than 3 cm with integrated arterial phase and venous phase (AP and VP) gemstone spectral imaging were retrospectively identified. After extracting the radiomic features of each case, principal component analysis (PCA) was used for feature selection, and after training with the logistic regression method, three classification models (ModelAP, ModelVP and ModelCombination) were constructed. The performance was assessed by the area under the receiver operating curve (AUC), and the efficacy of the models was validated using an independent cohort.A total of 153 patients were included and divided into a training cohort (n = 107) and a validation cohort (n = 46). A total of 1130 radiomic features were extracted from each case. The PCA method selected 22, 25 and 35 principal components to construct the three models. The diagnostic accuracy of ModelAP, ModelVP and ModelCombination was 0.8043, 0.6739, and 0.7826 in the validation set, with AUCs of 0.8148 (95% CI 0.682-0.948), 0.7485 (95% CI 0.602-0.895), and 0.8772 (95% CI 0.780-0.974), respectively. The DeLong test showed that there were significant differences in the AUCs between ModelAP and ModelCombination (P = 0.0396) and between ModelVP and ModelCombination (P = 0.0465). However, the difference in AUCs between ModelAP and ModelVP was not significant (P = 0.5061). These results demonstrate that ModelCombination shows a better performance than the other models. Decision curve analysis proved the clinical utility of this model.We developed a radiomics model based on monochromatic DECT images to identify solitary pulmonary nodules. This model could serve as an effective tool for discriminating benign from malignant pulmonary nodules in patients. The combination of arterial phase and venous phase imaging could significantly improve the model performance.

Volumetric accuracy of different imaging modalities in acute intracerebral hemorrhage

Abstract: Follow-up imaging in intracerebral hemorrhage is not standardized and radiologists rely on different imaging modalities to determine hematoma growth. This study assesses the volumetric accuracy of different imaging modalities (MRI, CT angiography, postcontrast CT) to measure hematoma size.28 patients with acute spontaneous intracerebral hemorrhage referred to a tertiary stroke center were retrospectively included between 2018 and 2019. Inclusion criteria were (1) spontaneous intracerebral hemorrhage (supra- or infratentorial), (2) noncontrast CT imaging performed on admission, (3) follow-up imaging (CT angiography, postcontrast CT, MRI), and (4) absence of hematoma expansion confirmed by a third cranial image within 6 days. Two independent raters manually measured hematoma volume by drawing a region of interest on axial slices of admission noncontrast CT scans as well as on follow-up imaging (CT angiography, postcontrast CT, MRI) using a semi-automated segmentation tool (Visage image viewer; version 7.1.10). Results were compared using Bland-Altman plots.Mean admission hematoma volume was 18.79 ± 19.86 cc. All interrater and intrarater intraclass correlation coefficients were excellent (1; IQR 0.98-1.00). In comparison to hematoma volume on admission noncontrast CT volumetric measurements were most accurate in patients who received postcontrast CT (bias of - 2.47%, SD 4.67: n = 10), while CT angiography often underestimated hemorrhage volumes (bias of 31.91%, SD 45.54; n = 20). In MRI sequences intracerebral hemorrhage volumes were overestimated in T2* (bias of - 64.37%, SD 21.65; n = 10). FLAIR (bias of 6.05%, SD 35.45; n = 13) and DWI (bias of-14.6%, SD 31.93; n = 12) over- and underestimated hemorrhagic volumes.Volumetric measurements were most accurate in postcontrast CT while CT angiography and MRI sequences often substantially over- or underestimated hemorrhage volumes.

Clinical parameters combined with radiomics features of PET/CT can predict recurrence in patients with high-risk pediatric neuroblastoma

Abstract: This retrospective study aimed to develop and validate a combined model based [18F]FDG PET/CT radiomics and clinical parameters for predicting recurrence in high-risk pediatric neuroblastoma patients.Eighty-four high-risk neuroblastoma patients were retrospectively enrolled and divided into training and test sets according to the ratio of 3:2. [18F]FDG PET/CT images of the tumor were segmented by 3D Slicer software and the radiomics features were extracted. The effective features were selected by the least absolute shrinkage and selection operator to construct the radiomics score (Rad_score). And the radiomics model (R_model) was constructed based on Rad_score for prediction of recurrence. Then, univariate and multivariate analyses were used to screen out the independent clinical risk parameters and construct the clinical model (C_model). A combined model (RC_model) was developed based on the Rad_score and independent clinical risk parameters and presented as radiomics nomogram. The performance of the above three models was assessed by the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA).Seven radiomics features were selected for building the R_model. The AUCs of the C_model in training and test sets were 0.744 (95% confidence interval [CI], 0.595-0.874) and 0.750 (95% CI, 0.577-0.904), respectively. The R_model yielded AUCs of 0.813 (95% CI, 0.685-0.916) and 0.869 (95% CI, 0.715-0.985) in the training and test sets, respectively. The RC_model demonstrated the largest AUCs of 0.889 (95% CI, 0.794-0.963) and 0.892 (95% CI, 0.758-0.992) in the training and test sets, respectively. DCA demonstrated that RC_model added more net benefits than either the C_model or the R_model for predicting recurrence in high-risk pediatric neuroblastoma.The combined model performed well for predicting recurrence in high-risk pediatric neuroblastoma, which can facilitate disease follow-up and management in clinical practice.

AbdomenNet: deep neural network for abdominal organ segmentation in epidemiologic imaging studies

Abstract: Whole-body imaging has recently been added to large-scale epidemiological studies providing novel opportunities for investigating abdominal organs. However, the segmentation of these organs is required beforehand, which is time consuming, particularly on such a large scale.We introduce AbdomentNet, a deep neural network for the automated segmentation of abdominal organs on two-point Dixon MRI scans. A pre-processing pipeline enables to process MRI scans from different imaging studies, namely the German National Cohort, UK Biobank, and Kohorte im Raum Augsburg. We chose a total of 61 MRI scans across the three studies for training an ensemble of segmentation networks, which segment eight abdominal organs. Our network presents a novel combination of octave convolutions and squeeze and excitation layers, as well as training with stochastic weight averaging.Our experiments demonstrate that it is beneficial to combine data from different imaging studies to train deep neural networks in contrast to training separate networks. Combining the water and opposed-phase contrasts of the Dixon sequence as input channels, yields the highest segmentation accuracy, compared to single contrast inputs. The mean Dice similarity coefficient is above 0.9 for larger organs liver, spleen, and kidneys, and 0.71 and 0.74 for gallbladder and pancreas, respectively.Our fully automated pipeline provides high-quality segmentations of abdominal organs across population studies. In contrast, a network that is only trained on a single dataset does not generalize well to other datasets.

Differentiation between combined hepatocellular cholangiocarcinoma and hepatocellular carcinoma: comparison of diagnostic performance between ultrasomics-based model and CEUS LI-RADS v2017

Abstract: The imaging findings of combined hepatocellular cholangiocarcinoma (CHC) may be similar to those of hepatocellular carcinoma (HCC). CEUS LI-RADS may not perform well in distinguishing CHC from HCC. Studies have shown that radiomics has an excellent imaging analysis ability. This study aimed to establish and confirm an ultrasomics model for differentiating CHC from HCC.Between 2004 and 2016, we retrospectively identified 53 eligible CHC patients and randomly included 106 eligible HCC patients with a ratio of HCC:CHC = 2:1, all of whom were categorized according to Contrast-Enhanced (CE) ultrasonography (US) Liver Imaging Reporting and Data System (LI-RADS) version 2017. The model based on ultrasomics features of CE US was developed in 74 HCC and 37 CHC and confirmed in 32 HCC and 16 CHC. The diagnostic performance of the LI-RADS or ultrasomics model was assessed by the area under the curve (AUC), accuracy, sensitivity and specificity.In the entire and validation cohorts, 67.0% and 81.3% of HCC cases were correctly assigned to LR-5 or LR-TIV contiguous with LR-5, and 73.6% and 87.5% of CHC cases were assigned to LR-M correctly. Up to 33.0% of HCC and 26.4% of CHC were misclassified by CE US LI-RADS. A total of 90.6% of HCC as well as 87.5% of CHC correctly diagnosed by the ultrasomics model in the validation cohort. The AUC, accuracy, sensitivity of the ultrasomics model were higher though without significant difference than those of CE US LI-RADS in the validation cohort.The proposed ultrasomics model showed higher ability though the difference was not significantly different for differentiating CHC from HCC, which may be helpful in clinical diagnosis.

Improved U-Net3+ with stage residual for brain tumor segmentation

Abstract: For the encoding part of U-Net3+,the ability of brain tumor feature extraction is insufficient, as a result, the features can not be fused well during up-sampling, and the accuracy of segmentation will reduce.In this study, we put forward an improved U-Net3+ segmentation network based on stage residual. In the encoder part, the encoder based on the stage residual structure is used to solve the vanishing gradient problem caused by the increasing in network depth, and enhances the feature extraction ability of the encoder which is instrumental in full feature fusion when up-sampling in the network. What's more, we replaced batch normalization (BN) layer with filter response normalization (FRN) layer to eliminate batch size impact on the network. Based on the improved U-Net3+ two-dimensional (2D) model with stage residual, IResUnet3+ three-dimensional (3D) model is constructed. We propose appropriate methods to deal with 3D data, which achieve accurate segmentation of the 3D network.The experimental results showed that: the sensitivity of WT, TC, and ET increased by 1.34%, 4.6%, and 8.44%, respectively. And the Dice coefficients of ET and WT were further increased by 3.43% and 1.03%, respectively. To facilitate further research, source code can be found at: https://github.com/YuOnlyLookOne/IResUnet3Plus .The improved network has a significant improvement in the segmentation task of the brain tumor BraTS2018 dataset, compared with the classical networks u-net, v-net, resunet and u-net3+, the proposed network has smaller parameters and significantly improved accuracy.

Ultrasound-based radiomics analysis for preoperative prediction of central and lateral cervical lymph node metastasis in papillary thyroid carcinoma: a multi-institutional study

Abstract: An accurate preoperative assessment of cervical lymph node metastasis (LNM) is important for choosing an optimal therapeutic strategy for papillary thyroid carcinoma (PTC) patients. This study aimed to develop and validate two ultrasound (US) nomograms for the individual prediction of central and lateral compartment LNM in patients with PTC.A total of 720 PTC patients from 3 institutions were enrolled in this study. They were categorized into a primary cohort, an internal validation, and two external validation cohorts. Radiomics features were extracted from conventional US images. LASSO regression was used to select optimized features to construct the radiomics signature. Two nomograms integrating independent clinical variables and radiomics signature were established with multivariate logistic regression. The performance of the nomograms was assessed with regard to discrimination, calibration, and clinical usefulness.The radiomics scores were significantly higher in patients with central/lateral LNM. A radiomics nomogram indicated good discrimination for central compartment LNM, with an area under the curve (AUC) of 0.875 in the training set, the corresponding value in the validation sets were 0.856, 0.870 and 0.870, respectively. Another nomogram for predicting lateral LNM also demonstrated good performance with an AUC of 0.938 and 0.905 in the training and internal validation cohorts, respectively. The AUC for the two external validation cohorts were 0.881 and 0.903, respectively. The clinical utility of the nomograms was confirmed by the decision curve analysis.The nomograms proposed here have favorable performance for preoperatively predicting cervical LNM, hold promise for optimizing the personalized treatment, and might greatly facilitate the decision-making in clinical practice.

Deep transfer learning to quantify pleural effusion severity in chest X-rays

Abstract: The detection of pleural effusion in chest radiography is crucial for doctors to make timely treatment decisions for patients with chronic obstructive pulmonary disease. We used the MIMIC-CXR database to develop a deep learning model to quantify pleural effusion severity in chest radiographs.The Medical Information Mart for Intensive Care Chest X-ray (MIMIC-CXR) dataset was divided into patients 'with' or 'without' chronic obstructive pulmonary disease (COPD). The label of pleural effusion severity was obtained from the extracted COPD radiology reports and classified into four categories: no effusion, small effusion, moderate effusion, and large effusion. A total of 200 datasets were randomly sampled to manually check each item and determine whether the tags are correct. A professional doctor re-tagged these items as a verification cohort without knowing their previous tags. The learning models include eight common network structures including Resnet, DenseNet, and GoogleNET. Three data processing methods (no sampling, downsampling, and upsampling) and two loss algorithms (focal loss and cross-entropy loss) were used for unbalanced data. The Neural Network Intelligence tool was applied to train the model. Receiver operating characteristic curves, Area under the curve, and confusion matrix were employed to evaluate the model results. Grad-CAM was used for model interpretation.Among the 8533 patients, 15,620 chest X-rays with clearly marked pleural effusion severity were obtained (no effusion, 5685; small effusion, 4877; moderate effusion, 3657; and large effusion, 1401). The error rate of the manual check label was 6.5%, and the error rate of the doctor's relabeling was 11.0%. The highest accuracy rate of the optimized model was 73.07. The micro-average AUCs of the testing and validation cohorts was 0.89 and 0.90, respectively, and their macro-average AUCs were 0.86 and 0.89, respectively. The AUC of the distinguishing results of each class and the other three classes were 0.95 and 0.94, 0.76 and 0.83, 0.85 and 0.83, and 0.87 and 0.93.The deep transfer learning model can grade the severity of pleural effusion.

Correlative analysis of lung CT findings in patients with Birt–Hogg–Dubé Syndrome and the occurrence of spontaneous pneumothorax: a preliminary study

Abstract: The diagnosis of patients with Birt-Hogg-Dubé (BHD) syndrome is always delayed (even for more than 10 years). Improving the understanding and diagnosis of this disease is vital for clinicians and radiologists. In this study we presented the chest computed tomography (CT) findings of BHD syndrome and offered suggestions for BHD cases with spontaneous pneumothorax.Twenty-six BHD patients from 11 families (10 men, 16 women; mean age: 46 ± 12 years, 20-68 years) were included. The clinical features of the patients included pneumothorax, renal lesions, and skin lesions. Twenty-three patients underwent chest CT imaging. The cyst condition of each patient derived from reconstructed chest CT imaging was recorded, including the cyst number, size, volume, pattern, and distribution.Pneumothorax occurred in 54% (14/26) of patients. Among them, 43% (6/14) had pneumothorax more than twice. However, typical skin and renal lesions were absent. Four patients had renal hamartoma. CT showed that 23 (100%) patients had lung cysts. Pulmonary cysts were bilateral and multiple, round, irregular, or willow-like. And 93.6% of the large cysts (long-axis diameter ≥ 20 mm) were under the pleura, and near the mediastinum and spine. The long-axis diameter, short-axis diameter and volume of the largest cysts were associated with the occurrence of pneumothorax (all P < 0.05).Chest CT imaging can reveal some characteristic features of BHD syndrome. The occurrence of pneumothorax in BHD patients is closely related to their pulmonary cystic lesions.

Frequency and anatomic variability of the mandibular lingual foramina: a cone-beam CT study

Abstract: To evaluate the distribution of lingual foramina (LF) and their correlation with demographic characteristics and mandible width, shape, and bone thickness in Caucasian Italian patients subjected to cone-beam CT (CBCT).CBCTs were reviewed to assess the number of all LF, midline and lateral LF. We also assessed the relationship of the number of lateral LF with gender and mandibular width, shape, and bone thickness using the Chi Square test. A p value < 0.05 was considered statistically significant.Three-hundred patients (180 males; age range: 21-87 years) were included. The highest frequency per patient was of 2 LF (97/300, 32.3%), followed by 3 (81/300, 27%) and 4 (53/300, 17.7%). No LF were observed in 2/300 patients (0.7%), while the highest number was of 8 LF in one patient. The highest frequency of midline LF per person was of 2 LF (57.3%, 172/300), while the highest number per person was 5 LF in one patient (0.3%). The highest frequency of midline LF located above and below the genial tubercle was of 1 in 197/300 patients (65.7%) and in 169/300 patients (56.3%), respectively. Concerning lateral LF, the highest frequencies were of 0 (113/300, 37.7%) and of 1 (112/300, 37.3%). We did not observe any significant difference of the number of midline and lateral LF based on gender (P = .438 and P = .195, respectively) or mandible width (P = .069 and P = .114, respectively). The mandible shape was normal in 188 cases, with facial constriction in 42, lingual constriction in 54, and hour glass constriction in 16. The mean bone thickness was 10.76 mm in the symphysis, 10.92 mm in the right hemiarches, and 10.68 in the left hemiarches. No significant differences in the distribution of LF were observed also based on mandibular shape and bone thickness (both with P > .05).We have shown the high variability of number and anatomic distribution of LF in an Italian group of patients subjected to CBCT without reporting any association with gender and mandible width, shape, and bone thickness.

Pulmonary vascular volume is associated with DLCO and fibrotic score in idiopathic pulmonary fibrosis: an observational study

Abstract: Abstract Background Idiopathic pulmonary fibrosis (IPF) is a disease that primarily occurs in elderly individuals. However, it is difficult to diagnose and has a complex disease course. High-resolution computed tomography (HRCT) and lung function testing are crucial for its diagnosis and follow-up. However, the correlation of HRCT findings with lung function test results has not been extensively investigated. Methods This study retrospectively analysed the medical records and images of patients with IPF. Patients with evident emphysema and lung cancer were excluded. The diagnosis of all the included cases was confirmed following a discussion among specialists from multiple disciplines. The correlation of HRCT findings, including fibrotic score, HRCT lung volume, pulmonary artery trunk (PA) diameter and pulmonary vascular volume (PVV), with lung function test parameters, such as forced vital capacity (FVC) and diffusing capacity for carbon monoxide (DLCO), was analysed. Results A total of 32 patients were included. Higher fibrotic and PVV scores were significantly correlated with lower DLCO (r = − 0.59, p = 0.01; r = − 0.43, p = 0.03, respectively) but not with FVC. Higher PVV score significantly correlated with higher fibrotic score (r = 0.59, p < 0.01) and PA diameter (r = 0.47, p = 0.006). Conclusion Our study demonstrated the structural and functional correlation of IPF. The extent of lung fibrosis (fibrotic score) and PVV score were associated with DLCO but not with FVC. The PA diameter, which reflects the pulmonary artery pressure, was found to be associated with the PVV score.

The efficacy of deep learning models in the diagnosis of endometrial cancer using MRI: a comparison with radiologists

Abstract: To compare the diagnostic performance of deep learning models using convolutional neural networks (CNN) with that of radiologists in diagnosing endometrial cancer and to verify suitable imaging conditions.This retrospective study included patients with endometrial cancer or non-cancerous lesions who underwent MRI between 2015 and 2020. In Experiment 1, single and combined image sets of several sequences from 204 patients with cancer and 184 patients with non-cancerous lesions were used to train CNNs. Subsequently, testing was performed using 97 images from 51 patients with cancer and 46 patients with non-cancerous lesions. The test image sets were independently interpreted by three blinded radiologists. Experiment 2 investigated whether the addition of different types of images for training using the single image sets improved the diagnostic performance of CNNs.The AUC of the CNNs pertaining to the single and combined image sets were 0.88-0.95 and 0.87-0.93, respectively, indicating non-inferior diagnostic performance than the radiologists. The AUC of the CNNs trained with the addition of other types of single images to the single image sets was 0.88-0.95.CNNs demonstrated high diagnostic performance for the diagnosis of endometrial cancer using MRI. Although there were no significant differences, adding other types of images improved the diagnostic performance for some single image sets.

DNL-Net: deformed non-local neural network for blood vessel segmentation

Abstract: The non-local module has been primarily used in literature to capturing long-range dependencies. However, it suffers from prohibitive computational complexity and lacks the interactions among positions across the channels.We present a deformed non-local neural network (DNL-Net) for medical image segmentation, which has two prominent components; deformed non-local module (DNL) and multi-scale feature fusion. The former optimizes the structure of the non-local block (NL), hence, reduces the problem of excessive computation and memory usage, significantly. The latter is derived from the attention mechanisms to fuse the features of different levels and improve the ability to exchange information across channels. In addition, we introduce a residual squeeze and excitation pyramid pooling (RSEP) module that is like spatial pyramid pooling to effectively resample the features at different scales and improve the network receptive field.The proposed method achieved 96.63% and 92.93% for Dice coefficient and mean intersection over union, respectively, on the intracranial blood vessel dataset. Also, DNL-Net attained 86.64%, 96.10%, and 98.37% for sensitivity, accuracy and area under receiver operation characteristic curve, respectively, on the DRIVE dataset.The overall performance of DNL-Net outperforms other current state-of-the-art vessel segmentation methods, which indicates that the proposed network is more suitable for blood vessel segmentation, and is of great clinical significance.

Safely reducing unnecessary benign breast biopsies by applying non-mass and DWI directional variance filters to ADC thresholding

Abstract: Thresholding apparent diffusion coefficient (ADC) maps obtained from Diffusion-Weighted-Imaging (DWI) has been proposed for identifying benign lesions that can safely avoid biopsy. The presence of malignancies with high ADC values leads to high thresholds, limiting numbers of avoidable biopsies.We evaluate two previously reported methods for identifying avoidable biopsies: using case-set dependent ADC thresholds that assure 100% sensitivity and using negative likelihood ratio (LR-) with a fixed ADC threshold of 1.50 × 10-3 mm2/s. We evaluated improvements in efficacy obtained by excluding non-mass lesions and lesions with anisotropic intra-lesion morphologic characteristics.Prospective.55 adult females with dense breasts with 69 BI-RADS 4 or 5 lesions (38 malignant, 31 benign) identified on ultrasound and mammography and imaged with MRI prior to biopsy.1.5 T and 3.0 T. DWI.Analysis of DWI, including directional images was done on an ROI basis. ROIs were drawn on DWI images acquired prior to biopsy, referencing all available images including DCE, and mean ADC was measured. Anisotropy was quantified via variation in ADC values in the lesion core across directional DWI images.Improvement in specificity at 100% sensitivity was evaluated with exact McNemar test with 1-sided p-value < 0.05 indicating statistical significance.Using ADC thresholding that assures 100% sensitivity, non-mass and directional variance filtering improved the percent of avoidable biopsies to 42% from baseline of 10% achieved with ADC thresholding alone. Using LR-, filtering improved outcome to 0.06 from baseline 0.25 with ADC thresholding alone. ADC thresholding showed a lower percentage of avoidable biopsies in our cohort than reported in prior studies. When ADC thresholding was supplemented with filtering, the percentage of avoidable biopsies exceeded those of prior studies.Supplementing ADC thresholding with filters excluding non-mass lesions and lesions with anisotropic characteristics on DWI can result in an increased number of avoidable biopsies.

Analysis of the value of enhanced CT combined with texture analysis in the differential diagnosis of pulmonary sclerosing pneumocytoma and atypical peripheral lung cancer: a feasibility study

Abstract: As a rare benign lung tumour, pulmonary sclerosing pneumocytoma (PSP) is often misdiagnosed as atypical peripheral lung cancer (APLC) on routine imaging examinations. This study explored the value of enhanced CT combined with texture analysis to differentiate between PSP and APLC.Forty-eight patients with PSP and fifty patients with APLC were retrospectively enrolled. The CT image features of the two groups of lesions were analysed, and MaZda software was used to evaluate the texture of CT venous phase thin-layer images. Independent sample t-test, Mann-Whitney U tests or χ2 tests were used to compare between groups. The intra-class correlation coefficient (ICC) was used to analyse the consistency of the selected texture parameters. Spearman correlation analysis was used to evaluate the differences in texture parameters between the two groups. Based on the statistically significant CT image features and CT texture parameters, the independent influencing factors between PSP and APLC were analysed by multivariate logistic regression. Extremely randomized trees (ERT) was used as the classifier to build models, and the models were evaluated by the five-fold cross-validation method.Logistic regression analysis based on CT image features showed that calcification and arterial phase CT values were independent factors for distinguishing PSP from APLC. The results of logistic regression analysis based on CT texture parameters showed that WavEnHL_s-1 and Perc.01% were independent influencing factors to distinguish the two. Compared with the single-factor model (models A and B), the classification accuracy of the model based on image features combined with texture parameters was 0.84 ± 0.04, the AUC was 0.84 ± 0.03, and the sensitivity and specificity were 0.82 ± 0.13 and 0.87 ± 0.12, respectively.Enhanced CT combined with texture analysis showed good diagnostic value for distinguishing PSP and APLC, which may contribute to clinical decision-making and prognosis evaluation.

High-resolution compressed sensing time-of-flight MR angiography outperforms CT angiography for evaluating patients with Moyamoya disease after surgical revascularization

Abstract: To evaluate the utility of high-resolution compressed sensing time-of-fight MR angiography (CS TOF-MRA) for assessing patients with moyamoya disease (MMD) after surgical revascularization, by comparison with computer tomography angiography (CTA).Twenty patients with MMD after surgical revascularizations who underwent CS TOF-MRA and CTA were collected. The scan time of CS TOF-MRA was 5 min and 4 s, with a reconstructed resolution of 0.4 × 0.4 × 0.4 mm3. Visualization of superficial temporal artery and middle cerebral artery (STA-MCA) bypass, neovascularization into the brain pial surface and Moyamoya vessels (MMVs) were independently ranked by two neuroradiologists on CS TOF-MRA and CTA, respectively. The patency of anastomosis was assessed as patent or occluded, using digital subtraction angiography and expert's consensus as ground truth. Interobserver agreement was calculated using the weighted kappa statistic. Wilcoxon signed-rank or Chi-square test was performed to investigate diagnostic difference between CS TOF-MRA and CTA.Twenty-two hemispheres from 20 patients were analyzed. The inter-reader agreement for evaluating STA-MCA bypass, neovascularization and anastomosis patency was good to excellent (κCS TOF-MRA, 0.738-1.000; κCTA, 0.743-0.909). The STA-MCA bypass and MMVs were better visualized on CS TOF-MRA than CTA (both P < 0.05). CS TOF-MRA had a higher sensitivity than CTA (94.7% vs. 73.7%) for visualizing anastomoses. Neovascularization was better observed in 13 (59.1%) sides on CS TOF-MRA, in comparison to 7 (31.8%) sides on CTA images (P = 0.005).High-resolution CS TOF-MRA outperforms CTA for visualization of STA-MCA bypass, neovascularization and MMVs within a clinically reasonable time in MMD patients after revascularization.