Liver CT Images Research Articles

Deep learning-based reconstruction of virtual monoenergetic images of kVp-switching dual energy CT for evaluation of hypervascular liver lesions: Comparison with standard reconstruction technique

ObjectiveTo investigate clinical applicability of deep learning(DL)-based reconstruction of virtual monoenergetic images(VMIs) of arterial phase liver CT obtained by rapid kVp-switching dual-energy CT for evaluation of hypervascular liver lesions. Materials and MethodsWe retrospectively included 109 patients who had available late arterial phase liver CT images of the liver obtained with a rapid switching kVp DECT scanner for suspicious intra-abdominal malignancies. Two VMIs of 70 keV and 40 keV were reconstructed using adaptive statistical iterative reconstruction (ASiR-V) for arterial phase scans. VMIs at 40 keV were additionally reconstructed with a vendor-agnostic DL-based reconstruction technique (ClariCT.AI, ClariPi, DL 40 keV). Qualitative, quantitative image quality and subjective diagnostic acceptability were compared according to reconstruction techniques. ResultsIn qualitative analysis, DL 40 keV images showed less image noise (4.55 vs 3.11 vs 3.95, p < 0.001), better image sharpness (4.75 vs 4.16 vs 4.3, p < 0.001), better image contrast (4.98 vs 4.72 vs 4.19, p < 0.017), better lesion conspicuity (4.61 vs 4.23 vs 3.4, p < 0.001) and diagnostic acceptability (4.59 vs 3.88 vs 4.09, p < 0.001) compared with ASiR-V 40 keV or 70 keV image sets. In quantitative analysis, DL 40 keV significantly reduced image noise relative to ASiR-V 40 keV images (49.9%, p < 0.001) and ASiR-V 70 keV images (85.2%, p = 0.012). DL 40 keV images showed significantly higher CNRlesion to the liver and SNRliver than ASiR-V 40 keV image and 70 keV images (p < 0.001). ConclusionDL-based reconstruction of 40 keV images using vendor-agnostic software showed greater noise reduction, better lesion conspicuity, image contrast, image sharpness, and higher overall image diagnostic acceptability than ASiR for 40 keV or 70 keV images in patients with hypervascular liver lesions.

MCI-Net: Multi-scale context integrated network for liver CT image segmentation

Owing to the various object scales and high similarity with the surrounding organs (e.g., kidney, stomach, and spleen), it is difficult to accurately segment the liver region from the abdominal computed tomography images. In this study, we propose a multi-scale context integration network called MCI-Net for liver image segmentation. Specifically, we first design a simplified residual module to prevent network degradation. Given the scale variability of objects, we propose a multi-scale context extraction module by combining four cascaded branches of hybrid dilated convolutions to capture broader and deeper features. In addition, we introduce an external attention mechanism based on two external, learnable and shared memory units, which helps to perceive the most discriminative information and suppress redundant features. Finally, we provide a boundary correction block to further improve the localization ability of boundary information. Extensive experiments on two liver CT image benchmark datasets qualitatively and quantitatively illustrate that our method is effective in improving liver segmentation accuracy and outperforms several state-of-the-art methods.

Automatic scan range for dose-reduced multiphase CT imaging of the liver utilizing CNNs and Gaussian models.

Multiphase CT scanning of the liver is performed for several clinical applications; however, radiation exposure from CT scanning poses a nontrivial cancer risk to the patients. The radiation dose may be reduced by determining the scan range of the subsequent scans by the location of the target of interest in the first scan phase. The purpose of this study is to present and assess an automatic method for determining the scan range for multiphase CT scans. Our strategy is to first apply a CNN-based method for detecting the liver in 2D slices, and to use a liver range search algorithm for detecting the liver range in the scout volume. The target liver scan range for subsequent scans can be obtained by adding safety margins achieved from Gaussian liver motion models to the scan range determined from the scout. Experiments were performed on 657 multiphase CT volumes obtained from multiple hospitals. The experiment shows that the proposed liver detection method can detect the liver in 223 out of a total of 224 3D volumes on average within one second, with mean intersection of union, wall distance and centroid distance of 85.5%, 5.7mm and 9.7mm, respectively. In addition, the performance of the proposed liver detection method is comparable to the best of the state-of-the-art 3D liver detectors in the liver detection accuracy while it requires less processing time. Furthermore, we apply the liver scan range generation method on the liver CT images acquired from radiofrequency ablation and Y-90 transarterial radioembolization (selective internal radiation therapy) interventions of 46 patients from two hospitals. The result shows that the automatic scan range generation can significantly reduce the effective radiation dose by an average of 14.5% (2.56 mSv) compared to manual performance by the radiographer from Y-90 transarterial radioembolization, while no statistically significant difference in performance was found with the CT images from intra RFA intervention (p = 0.81). Finally, three radiologists assess both the original and the range-reduced images for evaluating the effect of the range reduction method on their clinical decisions. We conclude that the automatic liver scan range generation method is able to reduce excess radiation compared to the manual performance with a high accuracy and without penalizing the clinical decision.

Optimized cross-modality guided contrast enhancement for liver tumor segmentation

Radiologists currently use CT images with intravenous contrast infusion, in order to detect lesions and vessels in the liver. This step is quite time consuming when it is done manually. There are many algorithms developed for segmentation which are based on edge or region characteristics. These are dependent on the quality of the image. The contrast of a CT or MRI image plays an important role in identifying region of interest i.e. lesion(s). Traditional enhancement methods suffer from limitations such as saturation, over-enhancement, and uneven contrast spatial distribution that may result from the uncontrolled CE process. One way to overcome such limitations is to combine the contrast enhancement approach with a quality control scheme. One way to overcome such limitations is to combine the contrast enhancement approach with a quality control scheme. Inspired by the guided filtering approach and the simplicity of context-aware histogram-based image quality enhancement, propose in this paper a cross-modality guided histogram specification technique to improve the contrast of liver CT images using MRI images as guiding input data. The proposed method is based on two concepts, namely guided image enhancement and image quality control through an optimization scheme. The proposed OPTimized Guided Contrast Enhancement (OPTGCE) scheme exploits both contextual information from the guidance image and structural information from the input image. Tumor segmentation algorithm is applied on the enhanced images to analyze the performance of the proposed method in facilitating tumor segmentation. The qualitative and quantitative analysis using metrics including entropy, MCCEE, and MIGLCM shows the superiority of the proposed method in comparison with the existing methods that do not include guidance mechanism Keywords—: Cross-modality, contrast enhancement, 2D histogram specification (HS), SSIM gradient, tumor segmentation.

Efficient Johnson-SB Mixture Model for Segmentation of CT Liver Image.

To overcome the problem that the traditional Gaussian mixture model (GMM) cannot well describe the skewness distribution of the gray-level histogram of a liver CT slice, we propose a novel segmentation method for liver CT images by introducing the Johnson-SB mixture model (JSBMM). The Johnson-SB model not only has a flexible asymmetrical distribution but also covers a variety of other distributions as well. In this article, the parameter optimization formulas for JSBMM were derived by employing the expectation-maximization (EM) algorithm and maximum likelihood. The implementation process of the JSBMM-based segmentation algorithm is provided in detail. To make better use of the skewness of Johnson-SB and improve the segmentation accuracy, we devise an idea to divide the histogram into two parts and calculate the segmentation threshold for each part, respectively, which is called JSBMM-TDH. By analyzing and comparing the segmentation thresholds with different cluster numbers, it is illustrated that the segmentation threshold of JSBMM-TDH will tend to be stable with the increasing of cluster number, while that of GMM is sensitive to different cluster numbers. The proposed JSBMM-TDH is applied to segment four randomly obtained abdominal CT image sequences, and the segmentation results and robustness have been compared between JSBMM-TDH and GMM. It is verified that JSBMM-TDH has preferable segmentation results and better robustness than GMM for the segmentation of liver CT images.

A Novel Approach for Small Object Detection in Medical Images through Deep Ensemble Convolution Neural Network

Small objects detection in medical image becomes an interesting field of research that helps the medical practitioners to focus on in-depth evaluation of diseases. The accurate localization and classification of objects face tremendous difficulty due to lower intensity of the images and distraction of pixel points that vary the decision on identifying the shape, structure etc. In many real-time cases, detection and classification of tiny objects in the medically treated images becomes mandatory. The proposed system is designed in the same criteria in which the semantic segmentation of tiny objects in the medical images is considered. The system design focused on implementing the model for different kinds of human organs such as lung and liver. The axial CT or PET images of Lung and Liver are considered as the prime input for the given system. Detection of tiny objects in the CT-PET images, segmenting it from the background and classification of segmented part as Tumor or Nodule is discussed. The preprocessed images are feature extracted after the morphology segmentation that determines the structural features of the tiny object being segmented. The feature vectors are nothing but the feature points from Kaze feature extraction and Morphology segmented image. These two inputs are fetched to the Deep ensemble Convolution neural network (DECNN) to obtain the dual classification results. Performing the quantitative measurements to evaluate the decision making system for nodule or tumor class is determined. The performance measure is done using accuracy, precision, recall and F1Score.

Developing an explainable deep learning boundary correction method by incorporating cascaded x-Dim models to improve segmentation defects in liver CT images

Deep learning methods achieved remarkable results in medical image analysis tasks but it has not yet been widely used by medical professionals. One of the main reasons for this restricted usage is the uncertainty of the reasons that influence the decision of the model. Explainable AI methods have been developed to improve the transparency, interpretability, and explainability of the black-box AI methods. The result of an explainable segmentation method will be more trusted by experts. In this study, we designed an explainable deep correction method by incorporating cascaded 1D and 2D models to refine the output of other models and provide reliable yet accurate results. We implemented a 2-step loop with a 1D local boundary validation model in the first step, and a 2D image patch segmentation model in the second step, to refine incorrect segmented regions slice-by-slice. The proposed method improved the result of the CNN segmentation models and achieved state-of-the-art results on 3D liver segmentation with the average dice coefficient of 98.27 on the Sliver07 dataset.

CT-Based Radiomics Score Can Accurately Predict Esophageal Variceal Rebleeding in Cirrhotic Patients.

Purpose: This study aimed to develop a radiomics score (Rad-score) extracted from liver and spleen CT images in cirrhotic patients to predict the probability of esophageal variceal rebleeding.Methods: In total, 173 cirrhotic patients were enrolled in this retrospective study. A total of 2,264 radiomics features of the liver and spleen were extracted from CT images. Least absolute shrinkage and selection operator (LASSO) Cox regression was used to select features and generate the Rad-score. Then, the Rad-score was evaluated by the concordance index (C-index), calibration curves, and decision curve analysis (DCA). Kaplan–Meier analysis was used to assess the risk stratification ability of the Rad-score.Results: Rad-scoreLiver, Rad-scoreSpleen, and Rad-scoreLiver−Spleen were independent risk factors for EV rebleeding. The Rad-scoreLiver−Spleen, which consisted of ten features, showed good discriminative performance, with C-indexes of 0.853 [95% confidence interval (CI), 0.776–0.904] and 0.822 (95% CI, 0.749–0.875) in the training and validation cohorts, respectively. The calibration curve showed that the predicted probability of rebleeding was very close to the actual probability. DCA verified the usefulness of the Rad-scoreLiver−Spleen in clinical practice. The Rad-scoreLiver−Spleen showed good performance in stratifying patients into high-, intermediate- and low-risk groups in both the training and validation cohorts. The C-index of the Rad-scoreLiver−Spleen in the hepatitis B virus (HBV) cohort was higher than that in the non-HBV cohort.Conclusion: The radiomics score extracted from liver and spleen CT images can predict the risk of esophageal variceal rebleeding and stratify cirrhotic patients accordingly.

Development and validation of an automated methodology to assess perceptual in vivo noise texture in liver CT.

Purpose: Developing, validating, and evaluating a method for measuring noise texture directly from patient liver CT images (i.e., in vivo). Approach: The method identifies target regions within patient scans that are least likely to have major contribution of patient anatomy, detrends them locally, and measures noise power spectrum (NPS) there using a previously phantom-validated technique targeting perceptual noise-non-anatomical fluctuations in the image that may interfere with the detection of focal lesions. Method development and validation used scanner-specific CT simulations of computational, anthropomorphic phantom (XCAT phantom, three phases of contrast-enhancement) with known ground truth of the NPS. Simulations were based on a clinical scanner (Definition Flash, Siemens) and clinically relevant settings (tube voltage of 120kV at three dose levels). Images were reconstructed with filtered backprojection (kernel: B31, B41, and B50) and Sinogram Affirmed Iterative Reconstruction (kernel: I31, I41, and I50) using a manufacturer-specific reconstruction software (ReconCT, Siemens). All NPS measurements were made in the liver. Ground-truth NPS were taken as the sum of (1)a measurement in parenchymal regions of anatomy-subtracted (i.e., noise only) scans, and (2)a measurement in the same region of noise-free (pre-noise-insertion) images. To assess in vivo NPS performance, correlation of NPS average frequency ( ), was reported. Sensitivity of accuracy [root-mean-square-error (RMSE)] to number of pixels included in measurement was conducted via bootstrapped pixel-dropout. Sensitivity of NPS to dose and reconstruction kernel was assessed to confirm that ground truth NPS similarities were maintained in patient-specific measurements. Results: Pearson and Spearman correlation coefficients 0.97 and 0.96 for indicated good correlation. Results suggested accurate NPS measurements (within 5% total RMSE) could be acquired with . Conclusions: Relationships of similar NPS due to reconstruction kernel and dose were preserved between gold standard and observed in vivo estimations. The NPS estimation method was further deployed on clinical cases to demonstrate the feasibility of clinical analysis.

A novel multilevel thresholding algorithm based on quantum computing for abdominal CT liver images

A novel multilevel thresholding algorithm based on quantum computing for abdominal CT liver images

Anatomic Variations of the Hepatic Artery in 5625 Patients.

To analyze the origin and anatomic course of the hepatic arteries by using digital subtraction angiography (DSA) and multidetector CT in a large study sample. This retrospective study included 5625 patients who underwent liver CT and chemoembolization between January 2005 and December 2018 (mean age, 60 years ± 11 [range, 11-99 years]; 4464 males). The CT and DSA images were reviewed to evaluate the visceral arterial anatomy for variations in the celiac axis and hepatic arteries. Aberrant right hepatic arteries (aRHAs) and aberrant left hepatic arteries (aLHAs) were defined on the basis of their origin and anatomic course. Statistical analyses were performed to evaluate the association between aRHAs and aLHAs and the association between variations in the hepatic artery and celiac axis. Right hepatic arteries were categorized as being aRHAs (if originating from the proximal to middle common hepatic artery, gastroduodenal artery, superior mesenteric artery [SMA], celiac axis, aorta, splenic artery, or left gastric artery [LGA]) or as being aLHAs (if arising from the LGA, celiac axis, aorta, or SMA). The prevalence of aRHAs (15.63%; 879 of 5625) and the prevalence aLHAs (16.32%; 918 of 5625) were similar. Patients with an aRHA were more likely to have an aLHA than those without an aRHA (29.01% vs 13.97%; P < .001), and patients with an aLHA were more likely to have an aRHA than those without an aLHA (27.78% vs 13.26%; P < .001). There was no association between the hepatic arterial variations and celiac axis variations. A hypothetical anatomic model summarizing the observed variations was created. A comprehensive list of hepatic arterial variations and a three-dimensional hypothetical model for the observed variations were described.Keywords: CT, Angiography, Liver, Anatomy, Arteries© RSNA, 2021 Supplemental material is available for this article. See also commentary by Sutphin and Kalva in this issue.

A Densely Connected UNet3D Network Combined Attention Mechanism for Liver and Tumor Segmentation

In recent years, liver cancer has become one of the five dangerous cancers due to the highest mortality ratios worldwide. Automatic tumor segmentation is a most important task to help radiologists and oncologists to analyze liver CT images. With the rapid development of Convolutional Neural Network (CNN), UNet2D have been widely applied in medical image segmentation. But 2D convolutions cannot extract more important spatial information, making it difficult for the network to learn powerful features between slices. In order to address the problems, we proposed a new densely connected UNet3D network combined attention mechanism (Att-DialResUNet3D) for liver and tumor segmentation. During coding and decoding stages, UNet3D used residual convolution with jagged structure blocks to decrease spatial hierarchical information loss. UNet3D applied an attention mechanism by using the long-range connections between the encoder and decoder to increase the ability of learning important information network. Dense connection decreases the gradient dissipation, and deep supervision can train the shallow layer more fully. We evaluated the proposed approach on the MICCAI 2017 liver tumor segmentation challenge (LiTS) dataset. Our approach preceded other research methods and has gained superior performance for liver tumor segmentation.

RECONSTRUCTION OF HIGH-RESOLUTION HEPATIC TUMOR CT IMAGES USING AN AUGMENTATION-BASED SUPER-RESOLUTION TECHNIQUE

Improving the resolution of medical images is crucial in diagnosis, feature extraction, and data retrieval. A significant group of super-resolution algorithms is multi-frame techniques. However, they are not appropriate to medical data since they need several frames of the same scene, which bring a high risk of radiation or require a considerable acquisition time. We propose a new data augmentation technique and employ it in a multi-frame image reconstruction algorithm to improve the resolution of pathologic liver CT images. The input to our algorithm is a 3D CT-scan of the abdominal region. Neighboring slices are considered to increase the resolution of a single slice. Augmented slices are prepared using the nearby slices and the interpolation approach. The new data is aligned to the original slice, and it is used as an augmented version of the data. Then, a multi-frame scheme is utilized to reconstruct the high-resolution image. Our method’s novelty is to remove the need for multiple scans of a patent to employ multi-resolution techniques in medical applications. The results reveal that the proposed method is superior to conventional interpolation methods and available augmentation techniques. Compared to the tricubic interpolation, the proposed method improved the PSNR by 3.1. Concerning conventional augmentation techniques, it enhanced the SSIM measure by 0.06. The proposed algorithm improved the SSIM by 0.11 compared to traditional interpolation techniques and 0.1 for recent researches. Therefore, a multi-frame super-resolution technique has the potential to reconstruct medical data better.

Inter-reader reliability of CT Liver Imaging Reporting and Data System according to imaging analysis methodology: a systematic review and meta-analysis.

To establish inter-reader reliability of CT Liver Imaging Reporting and Data System (LI-RADS) and explore factors that affect it. MEDLINE and EMBASE databases were searched from January 2014 to March 2020 to identify original articles reporting the inter-reader reliability of CT LI-RADS. The imaging analysis methodology of each study was identified, and pooled intraclass correlation coefficient (ICC) or kappa values (κ) were calculated for lesion size, major features (arterial-phase hyperenhancement [APHE], nonperipheral washout [WO], andenhancing capsule [EC]), and LI-RADS categorization (LR) using random-effects models. Subgroup analyses of pooled κ were performed for the number of readers, average reader experience, differences in reader experience, and LI-RADS version. In the 12 included studies, the pooled ICC or κ of lesion size, APHE, WO, EC, and LR were 0.99 (0.96-1.00), 0.69 (0.58-0.81), 0.67 (0.53-0.82), 0.65 (0.54-0.76), and 0.70 (0.59-0.82), respectively. The experience and number of readers varied: studies using readers with ≥ 10 years of experience showed significantly higher κ for LR (0.82 vs. 0.45, p = 0.01) than those with < 10 years of reader experience. Studies with multiple readers including inexperienced readers showed significantly lower κ for APHE (0.55 vs. 0.76, p = 0.04) and LR (0.45 vs. 0.79, p = 0.02) than those with all experienced readers. CT LI-RADS showed substantial inter-reader reliability for major features and LR. Inter-reader reliability differed significantly according to average reader experience and differences in reader experience. Reported results for inter-reader reliability of CT LI-RADS should be understood with consideration of the imaging analysis methodology. • The CT Liver Imaging Reporting and Data System (LI-RADS) provides substantial inter-reader reliability for three major features and category assignment. • The imaging analysis methodology varied across studies. • The inter-reader reliability of CT LI-RADS differed significantly according to the average reader experience and the difference in reader experience.

Segmentation and Registration of the Liver in Dynamic Contrast-Enhanced Computed Tomography Images

Dynamic contrast-enhanced computed tomography (DCE-CT) is the main auxiliary diagnostic tool for liver diseases. Liver segmentation and registration in all stages of DCE-CT images are the key technology for big data analysis of liver disease diagnosis. The change of imaging conditions in different stages of DCE-CT brings enormous challenges to the segmentation of liver CT images. This study proposes an automatic model for liver segmentation from abdominal CT images in different stages of DCE on the basis of U-Net. The skip connection in U-Net can improve the ability of complex feature recognition. A total of 4863 CT slices from 16 patients with hepatocellular carcinoma (HCC) were selected as the training set, and 1754 CT slices from 6 patients with HCC were selected as the test set. The training and test sets included plain scan, hepatic arterial-dominant phase, and portal venous-dominant phase CT scans. Results showed that the Dice value of the proposed method was significantly higher than those of the full convolutional network and region-growing method. Then, 3D reconstruction and registration were performed on the segmentation results of the liver region of DCE-CT images. The proposed method obtained the best performance, which can provide technical support for the big data analysis of liver diseases.

MAGAN: Mask Attention Generative Adversarial Network for Liver Tumor CT Image Synthesis.

For deep learning, the size of the dataset greatly affects the final training effect. However, in the field of computer-aided diagnosis, medical image datasets are often limited and even scarce. We aim to synthesize medical images and enlarge the size of the medical image dataset. In the present study, we synthesized the liver CT images with a tumor based on the mask attention generative adversarial network (MAGAN). We masked the pixels of the liver tumor in the image as the attention map. And both the original image and attention map were loaded into the generator network to obtain the synthesized images. Then, the original images, the attention map, and the synthesized images were all loaded into the discriminator network to determine if the synthesized images were real or fake. Finally, we can use the generator network to synthesize liver CT images with a tumor. The experiments showed that our method outperformed the other state-of-the-art methods and can achieve a mean peak signal-to-noise ratio (PSNR) of 64.72 dB. All these results indicated that our method can synthesize liver CT images with a tumor and build a large medical image dataset, which may facilitate the progress of medical image analysis and computer-aided diagnosis. An earlier version of our study has been presented as a preprint in the following link: https://www.researchsquare.com/article/rs-41685/v1.

'Curved tunnel' sign on MRI: a typical radiological feature in hepatic trichinellosis.

To summarize the clinical and radiological features of hepatic trichinellosis. Patients diagnosed with trichinellosis and liver involvement in our institution since 2011 were included retrospectively. Then, qualified patients were divided into a seropositive group and a clinically diagnosed group. Preoperative liver CT and/or MRI images were evaluated independently by two radiologists. The correlations between the radiological and pathological features were evaluated by a radiologist and a pathologist. Independent t-tests were carried out to assess clinical data between groups and those reported in the previous literature. Values of P less than 0.05 were considered statistically significant. Eventually, ten patients were enrolled, five in each group. All patients recovered successfully after surgical (n = 8) or conservative treatment (n = 2). The clinical, radiological, and pathological features were highly consistent between the two groups. All patients were under 60years of age (48.0 ± 7.9years in the seropositive group, 48.6 ± 10.1years in the clinically diagnosed group), and they were older than those previously reported in the literature (30.5 ± 17years, n = 31; P = 0.032 and 0.028). Eosinophilia was found in four patients, and all were mildly increased (˂ 1.5 × 109/L). Lesions in all patients were in the right lobe of the liver, and nine of them showed a characteristic 'curved tunnel' sign on DCE-MRI images. None of the patients had signs or evidence of extrahepatic infection. The 'curved tunnel' sign on MRI is a typical radiological feature in hepatic trichinellosis.

Region-growing fully convolutional neural network interactive segmentation of liver CT images

Region-growing fully convolutional neural network interactive segmentation of liver CT images

Annotation-Efficient Learning for Medical Image Segmentation Based on Noisy Pseudo Labels and Adversarial Learning

Despite that deep learning has achieved state-of-the-art performance for medical image segmentation, its success relies on a large set of manually annotated images for training that are expensive to acquire. In this paper, we propose an annotation-efficient learning framework for segmentation tasks that avoids annotations of training images, where we use an improved Cycle-Consistent Generative Adversarial Network (GAN) to learn from a set of unpaired medical images and auxiliary masks obtained either from a shape model or public datasets. We first use the GAN to generate pseudo labels for our training images under the implicit high-level shape constraint represented by a Variational Auto-encoder (VAE)-based discriminator with the help of the auxiliary masks, and build a Discriminator-guided Generator Channel Calibration (DGCC) module which employs our discriminator's feedback to calibrate the generator for better pseudo labels. To learn from the pseudo labels that are noisy, we further introduce a noise-robust iterative learning method using noise-weighted Dice loss. We validated our framework with two situations: objects with a simple shape model like optic disc in fundus images and fetal head in ultrasound images, and complex structures like lung in X-Ray images and liver in CT images. Experimental results demonstrated that 1) Our VAE-based discriminator and DGCC module help to obtain high-quality pseudo labels. 2) Our proposed noise-robust learning method can effectively overcome the effect of noisy pseudo labels. 3) The segmentation performance of our method without using annotations of training images is close or even comparable to that of learning from human annotations.

Data Balancing Method for Training Segmentation Neural Networks

Data imbalance is a common problem in machine learning and image processing. The lack of training data for the rarest classes can lead to worse learning ability and negatively affect the quality of segmentation. In this paper, we focus on the problem of data balancing for the task of image segmentation. We review major trends in handling unbalanced data and propose a new method for data balancing, based on Distance Transform. This method is designed for using in segmentation convolutional neural networks (CNNs), but it is universal and can be used with any patch-based segmentation machine learning model. The evaluation of the proposed data balancing method is performed on two datasets. The first is medical dataset LiTS, containing CT images of liver with tumor abnormalities. The second one is a geological dataset, containing of photographs of polished sections of different ores. The proposed algorithm enhances the data balance between classes and improves the overall performance of CNN model.