Unified medical image segmentation by learning from uncertainty in an end-to-end manner

A review of uncertainty quantification in medical image analysis: Probabilistic and non-probabilistic methods

The comprehensive integration of machine learning healthcare models within clinical practice remains suboptimal, notwithstanding the proliferation of high-performing solutions reported in the literature. A predominant factor hindering widespread adoption pertains to an insufficiency of evidence affirming the reliability of the aforementioned models. Recently, uncertainty quantification methods have been proposed as a potential solution to quantify the reliability of machine learning models and thus increase the interpretability and acceptability of the results. In this review, we offer a comprehensive overview of the prevailing methods proposed to quantify the uncertainty inherent in machine learning models developed for various medical image tasks. Contrary to earlier reviews that exclusively focused on probabilistic methods, this review also explores non-probabilistic approaches, thereby furnishing a more holistic survey of research pertaining to uncertainty quantification for machine learning models. Analysis of medical images with the summary and discussion on medical applications and the corresponding uncertainty evaluation protocols are presented, which focus on the specific challenges of uncertainty in medical image analysis. We also highlight some potential future research work at the end. Generally, this review aims to allow researchers from both clinical and technical backgrounds to gain a quick and yet in-depth understanding of the research in uncertainty quantification for medical image analysis machine learning models.

Towards consistency in pediatric brain tumor measurements: Challenges, solutions, and the role of artificial intelligence-based segmentation

MR imaging is central to the assessment of tumor burden and changes over time in neuro-oncology. Several response assessment guidelines have been set forth by the Response Assessment in Pediatric Neuro-Oncology (RAPNO) working groups in different tumor histologies; however, the visual delineation of tumor components using MRIs is not always straightforward, and complexities not currently addressed by these criteria can introduce inter- and intra-observer variability in manual assessments. Differentiation of non-enhancing tumors from peritumoral edema, mild enhancement from absence of enhancement, and various cystic components can be challenging; particularly given a lack of sufficient and uniform imaging protocols in clinical practice. Automated tumor segmentation with artificial intelligence (AI) may be able to provide more objective delineations, but rely on accurate and consistent training data created manually (ground truth). Herein, this paper reviews existing challenges and potential solutions to identifying and defining subregions of pediatric brain tumors (PBTs) that are not explicitly addressed by current guidelines. The goal is to assert the importance of defining and adopting criteria for addressing these challenges, as it will be critical to achieving standardized tumor measurements and reproducible response assessment in PBTs, ultimately leading to more precise outcome metrics and accurate comparisons among clinical studies.

Application of simultaneous uncertainty quantification and segmentation for oropharyngeal cancer use-case with Bayesian deep learning

BACKGROUND

Radiotherapy is a core treatment modality for oropharyngeal cancer (OPC), where the primary gross tumor volume (GTVp) is manually segmented with high interobserver variability. This calls for reliable and trustworthy automated tools in clinician workflow. Therefore, accurate uncertainty quantification and its downstream utilization is critical.

METHODS

Here we propose uncertainty-aware deep learning for OPC GTVp segmentation, and illustrate the utility of uncertainty in multiple applications. We examine two Bayesian deep learning (BDL) models and eight uncertainty measures, and utilize a large multi-institute dataset of 292 PET/CT scans to systematically analyze our approach.

RESULTS

We show that our uncertainty-based approach accurately predicts the quality of the deep learning segmentation in 86.6% of cases, identifies low performance cases for semi-automated correction, and visualizes regions of the scans where the segmentations likely fail.

CONCLUSIONS

Our BDL-based analysis provides a first-step towards more widespread implementation of uncertainty quantification in OPC GTVp segmentation.

Multi-sequence generative adversarial network: better generation for enhanced magnetic resonance imaging images

Introduction

MRI is one of the commonly used diagnostic methods in clinical practice, especially in brain diseases. There are many sequences in MRI, but T1CE images can only be obtained by using contrast agents. Many patients (such as cancer patients) must undergo alignment of multiple MRI sequences for diagnosis, especially the contrast-enhanced magnetic resonance sequence. However, some patients such as pregnant women, children, etc. find it difficult to use contrast agents to obtain enhanced sequences, and contrast agents have many adverse reactions, which can pose a significant risk. With the continuous development of deep learning, the emergence of generative adversarial networks makes it possible to extract features from one type of image to generate another type of image.

Methods

We propose a generative adversarial network model with multimodal inputs and end-to-end decoding based on the pix2pix model. For the pix2pix model, we used four evaluation metrics: NMSE, RMSE, SSIM, and PNSR to assess the effectiveness of our generated model.

Results

Through statistical analysis, we compared our proposed new model with pix2pix and found significant differences between the two. Our model outperformed pix2pix, with higher SSIM and PNSR, lower NMSE and RMSE. We also found that the input of T1W images and T2W images had better effects than other combinations, providing new ideas for subsequent work on generating magnetic resonance enhancement sequence images. By using our model, it is possible to generate magnetic resonance enhanced sequence images based on magnetic resonance non-enhanced sequence images.

Discussion

This has significant implications as it can greatly reduce the use of contrast agents to protect populations such as pregnant women and children who are contraindicated for contrast agents. Additionally, contrast agents are relatively expensive, and this generation method may bring about substantial economic benefits.

SC-Net: Symmetrical conical network for colorectal pathology image segmentation

BACKGROUND AND OBJECTIVE

Image segmentation of histopathology of colorectal cancer is a core task of computer aided medical image diagnosis system. Existing convolutional neural networks generally extract multi-scale information in linear flow structures by inserting multi-branch modules, which is difficult to extract heterogeneous semantic information under multi-level and different receptive field and tough to establish context dependency among different receptive field features.

METHODS

To address these issues, we propose a symmetric spiral progressive feature fusion encoder-decoder network called the Symmetric Conical Network (SC-Net). First, we design a Multi-scale Feature Extraction Block (MFEB) matching with the Symmetric Conical Network to obtain multi-branch heterogeneous semantic information under different receptive fields, so as to enrich the diversity of extracted feature information. The encoder is composed of MFEB through spiral and multi-branch arrangement to enhance context dependence between different information flow. Secondly, the information loss of contour, color and others in high-level semantic information through causally stacking MFEB, the Feature Mapping Layer (FML) is designed to map low-level features to high-level semantic features along the down-sampling branch and solve the problem of insufficient global feature extraction in deep levels.

RESULTS

The SC-Net was evaluated on our self-constructed colorectal cancer dataset, a publicly available breast cancer dataset and a polyp dataset. The results revealed that the mDice of segmentation reached 0.8611, 0.7259 and 0.7144. We compare our model with the state-of-art semantic segmentation UNet++, PSPNet, Attention U-Net, R2U-Net and other advanced segmentation networks. The experimental results demonstrate that we achieve the most advanced performance.

CONCLUSIONS

The results indicate that the proposed SC-Net excels in segmenting H&E stained pathology images, effectively preserving morphological features and spatial information even in scenarios with weak texture, poor contrast, and variations in appearance.

Semantic uncertainty Guided Cross-Transformer for enhanced macular edema segmentation in OCT images

Macular edema, a prevalent ocular complication observed in various retinal diseases, can lead to significant vision loss or blindness, necessitating accurate and timely diagnosis. Despite the potential of deep learning for segmentation of macular edema, challenges persist in accurately identifying lesion boundaries, especially in low-contrast and noisy regions, and in distinguishing between Inner Retinal Fluid (IRF), Sub-Retinal Fluid (SRF), and Pigment Epithelial Detachment (PED) lesions. To address these challenges, we present a novel approach, termed Semantic Uncertainty Guided Cross-Transformer Network (SuGCTNet), for the simultaneous segmentation of multi-class macular edema. Our proposed method comprises two key components, the semantic uncertainty guided attention module (SuGAM) and the Cross-Transformer module (CTM). The SuGAM module utilizes semantic uncertainty to allocate additional attention to regions with semantic ambiguity, improves the segmentation performance of these challenging areas. On the other hand, the CTM module capitalizes on both uncertainty information and multi-scale image features to enhance the overall continuity of the segmentation process, effectively minimizing feature confusion among different lesion types. Rigorous evaluation on public datasets and various OCT imaging device data demonstrates the superior performance of our proposed method compared to state-of-the-art approaches, highlighting its potential as a valuable tool for improving the accuracy and reproducibility of macular edema segmentation in clinical settings, and ultimately aiding in the early detection and diagnosis of macular edema-related diseases and associated retinal conditions.

The application value of LAVA-flex sequences in enhanced MRI scans of nasopharyngeal carcinoma: comparison with T1WI-IDEAL

Introduction

Magnetic resonance imaging (MRI) staging scans are critical for the diagnosis and treatment of patients with nasopharyngeal cancer (NPC). We aimed to evaluate the application value of LAVA-Flex and T1WI-IDEAL sequences in MRI staging scans.

Methods

Eighty-four newly diagnosed NPC patients underwent both LAVA-Flex and T1WI-IDEAL sequences during MRI examinations. Two radiologists independently scored the acquisitions of image quality, fat suppression quality, artifacts, vascular and nerve display. The obtained scores were compared using the Wilcoxon signed rank test. According to the signal intensity (SI) measurements, the uniformity of fat suppression, contrast between tumor lesions and subcutaneous fat tissue, and signal-to-noise ratio (SNR) were compared by the paired t-test.

Results

Compared to the T1WI-IDEAL sequence, LAVA-Flex exhibited fewer artifacts (P<0.05), better visualization of nerves and vessels (P<0.05), and performed superior in the fat contrast ratio of the primary lesion and metastatic lymph nodes (0.80 vs. 0.52, 0.81 vs. 0.56, separately, P<0.001). There was no statistically significant difference in overall image quality, tumor signal-to-noise ratio (SNR), muscle SNR, and the detection rate of lesions between the two sequences (P>0.05). T1WI-IDEAL was superior to LAVA-Flex in the evaluation of fat suppression uniformity (P<0.05).

Discussion

LAVA-Flex sequence provides satisfactory image quality and better visualization of nerves and vessels for NPC with shorter scanning times.

A deep learning-based semiautomated workflow for triaging follow-up MR scans in treated nasopharyngeal carcinoma

It is imperative to optimally utilize virtues and obviate defects of fully automated analysis and expert knowledge in new paradigms of healthcare. We present a deep learning-based semiautomated workflow (RAINMAN) with 12,809 follow-up scans among 2,172 patients with treated nasopharyngeal carcinoma from three centers (ChiCTR.org.cn, Chi-CTR2200056595). A boost of diagnostic performance and reduced workload was observed in RAINMAN compared with the original manual interpretations (internal vs. external: sensitivity, 2.5% [p = 0.500] vs. 3.2% [p = 0.031]; specificity, 2.9% [p < 0.001] vs. 0.3% [p = 0.302]; workload reduction, 79.3% vs. 76.2%). The workflow also yielded a triaging performance of 83.6%, with increases of 1.5% in sensitivity (p = 1.000) and 0.6%-1.3% (all p < 0.05) in specificity compared to three radiologists in the reader study. The semiautomated workflow shows its unique superiority in reducing radiologist's workload by eliminating negative scans while retaining the diagnostic performance of radiologists.

Graph- and transformer-guided boundary aware network for medical image segmentation

BACKGROUND AND OBJECTIVE

Despite the considerable progress achieved by U-Net-based models, medical image segmentation remains a challenging task due to complex backgrounds, irrelevant noises, and ambiguous boundaries. In this study, we present a novel approach called U-shaped Graph- and Transformer-guided Boundary Aware Network (GTBA-Net) to tackle these challenges.

METHODS

GTBA-Net uses the pre-trained ResNet34 as its basic structure, and involves Global Feature Aggregation (GFA) modules for target localization, Graph-based Dynamic Feature Fusion (GDFF) modules for effective noise suppression, and Uncertainty-based Boundary Refinement (UBR) modules for accurate delineation of ambiguous boundaries. The GFA modules employ an efficient self-attention mechanism to facilitate coarse target localization amidst complex backgrounds, without introducing additional computational complexity. The GDFF modules leverage graph attention mechanism to aggregate information hidden among high- and low-level features, effectively suppressing target-irrelevant noises while preserving valuable spatial details. The UBR modules introduce an uncertainty quantification strategy and auxiliary loss to guide the model's focus towards target regions and uncertain "ridges", gradually mitigating boundary uncertainty and ultimately achieving accurate boundary delineation.

RESULTS

Comparative experiments on five datasets encompassing diverse modalities (including X-ray, CT, endoscopic procedures, and ultrasound) demonstrate that the proposed GTBA-Net outperforms existing methods in various challenging scenarios. Subsequent ablation studies further demonstrate the efficacy of the GFA, GDFF, and UBR modules in target localization, noise suppression, and ambiguous boundary delineation, respectively.

CONCLUSIONS

GTBA-Net exhibits substantial potential for extensive application in the field of medical image segmentation, particularly in scenarios involving complex backgrounds, target-irrelevant noises, or ambiguous boundaries.

DRI-Net: segmentation of polyp in colonoscopy images using dense residual-inception network

Colorectal cancer is a common malignant tumor in the gastrointestinal tract, which usually evolves from adenomatous polyps. However, due to the similarity in color between polyps and their surrounding tissues in colonoscopy images, and their diversity in size, shape, and texture, intelligent diagnosis still remains great challenges. For this reason, we present a novel dense residual-inception network (DRI-Net) which utilizes U-Net as the backbone. Firstly, in order to increase the width of the network, a modified residual-inception block is designed to replace the traditional convolutional, thereby improving its capacity and expressiveness. Moreover, the dense connection scheme is adopted to increase the network depth so that more complex feature inputs can be fitted. Finally, an improved down-sampling module is built to reduce the loss of image feature information. For fair comparison, we validated all method on the Kvasir-SEG dataset using three popular evaluation metrics. Experimental results consistently illustrates that the values of DRI-Net on IoU, Mcc and Dice attain 77.72%, 85.94% and 86.51%, which were 1.41%, 0.66% and 0.75% higher than the suboptimal model. Similarly, through ablation studies, it also demonstrated the effectiveness of our approach in colorectal semantic segmentation.

Paced-curriculum distillation with prediction and label uncertainty for image segmentation

PURPOSE

In curriculum learning, the idea is to train on easier samples first and gradually increase the difficulty, while in self-paced learning, a pacing function defines the speed to adapt the training progress. While both methods heavily rely on the ability to score the difficulty of data samples, an optimal scoring function is still under exploration.

METHODOLOGY

Distillation is a knowledge transfer approach where a teacher network guides a student network by feeding a sequence of random samples. We argue that guiding student networks with an efficient curriculum strategy can improve model generalization and robustness. For this purpose, we design an uncertainty-based paced curriculum learning in self-distillation for medical image segmentation. We fuse the prediction uncertainty and annotation boundary uncertainty to develop a novel paced-curriculum distillation (P-CD). We utilize the teacher model to obtain prediction uncertainty and spatially varying label smoothing with Gaussian kernel to generate segmentation boundary uncertainty from the annotation. We also investigate the robustness of our method by applying various types and severity of image perturbation and corruption.

RESULTS

The proposed technique is validated on two medical datasets of breast ultrasound image segmentation and robot-assisted surgical scene segmentation and achieved significantly better performance in terms of segmentation and robustness.

CONCLUSION

P-CD improves the performance and obtains better generalization and robustness over the dataset shift. While curriculum learning requires extensive tuning of hyper-parameters for pacing function, the level of performance improvement suppresses this limitation.

MHL-Net: A Multistage Hierarchical Learning Network for Head and Neck Multiorgan Segmentation

Accurate segmentation of head and neck organs at risk is crucial in radiotherapy. However, the existing methods suffer from incomplete feature mining, insufficient information utilization, and difficulty in simultaneously improving the performance of small and large organ segmentation. In this paper, a multistage hierarchical learning network is designed to fully extract multidimensional features, combined with anatomical prior information and imaging features, using multistage subnetworks to improve the segmentation performance. First, multilevel subnetworks are constructed for primary segmentation, localization, and fine segmentation by dividing organs into two levels-large and small. Different networks both have their own learning focuses and feature reuse and information sharing among each other, which comprehensively improved the segmentation performance of all organs. Second, an anatomical prior probability map and a boundary contour attention mechanism are developed to address the problem of complex anatomical shapes. Prior information and boundary contour features effectively assist in detecting and segmenting special shapes. Finally, a multidimensional combination attention mechanism is proposed to analyze axial, coronal, and sagittal information, capture spatial and channel features, and maximize the use of structural information and semantic features of 3D medical images. Experimental results on several datasets showed that our method was competitive with state-of-the-art methods and improved the segmentation results for multiscale organs.

DCTR U-Net: automatic segmentation algorithm for medical images of nasopharyngeal cancer in the context of deep learning

Nasopharyngeal carcinoma (NPC) is a malignant tumor that occurs in the wall of the nasopharyngeal cavity and is prevalent in Southern China, Southeast Asia, North Africa, and the Middle East. According to studies, NPC is one of the most common malignant tumors in Hainan, China, and it has the highest incidence rate among otorhinolaryngological malignancies. We proposed a new deep learning network model to improve the segmentation accuracy of the target region of nasopharyngeal cancer. Our model is based on the U-Net-based network, to which we add Dilated Convolution Module, Transformer Module, and Residual Module. The new deep learning network model can effectively solve the problem of restricted convolutional fields of perception and achieve global and local multi-scale feature fusion. In our experiments, the proposed network was trained and validated using 10-fold cross-validation based on the records of 300 clinical patients. The results of our network were evaluated using the dice similarity coefficient (DSC) and the average symmetric surface distance (ASSD). The DSC and ASSD values are 0.852 and 0.544 mm, respectively. With the effective combination of the Dilated Convolution Module, Transformer Module, and Residual Module, we significantly improved the segmentation performance of the target region of the NPC.

MFA-Net: Multiple Feature Association Network for medical image segmentation

Medical image segmentation plays a crucial role in computer-aided diagnosis. However, due to the large variability of medical images, accurate segmentation is a highly challenging task. In this paper, we present a novel medical image segmentation network named the Multiple Feature Association Network (MFA-Net), which is based on deep learning techniques. The MFA-Net utilizes an encoder-decoder architecture with skip connections as its backbone network, and a parallelly dilated convolutions arrangement (PDCA) module is integrated between the encoder and the decoder to capture more representative deep features. Furthermore, a multi-scale feature restructuring module (MFRM) is introduced to restructure and fuse the deep features of the encoder. To enhance global attention perception, the proposed global attention stacking (GAS) modules are cascaded on the decoder. The proposed MFA-Net leverages novel global attention mechanisms to improve the segmentation performance at different feature scales. We evaluated our MFA-Net on four segmentation tasks, including lesions in intestinal polyp, liver tumor, prostate cancer, and skin lesion. Our experimental results and ablation study demonstrate that the proposed MFA-Net outperforms state-of-the-art methods in terms of global positioning and local edge recognition.

Application of simultaneous uncertainty quantification for image segmentation with probabilistic deep learning: Performance benchmarking of oropharyngeal cancer target delineation as a use-case

Background

Oropharyngeal cancer (OPC) is a widespread disease, with radiotherapy being a core treatment modality. Manual segmentation of the primary gross tumor volume (GTVp) is currently employed for OPC radiotherapy planning, but is subject to significant interobserver variability. Deep learning (DL) approaches have shown promise in automating GTVp segmentation, but comparative (auto)confidence metrics of these models predictions has not been well-explored. Quantifying instance-specific DL model uncertainty is crucial to improving clinician trust and facilitating broad clinical implementation. Therefore, in this study, probabilistic DL models for GTVp auto-segmentation were developed using large-scale PET/CT datasets, and various uncertainty auto-estimation methods were systematically investigated and benchmarked.

Methods

We utilized the publicly available 2021 HECKTOR Challenge training dataset with 224 co-registered PET/CT scans of OPC patients with corresponding GTVp segmentations as a development set. A separate set of 67 co-registered PET/CT scans of OPC patients with corresponding GTVp segmentations was used for external validation. Two approximate Bayesian deep learning methods, the MC Dropout Ensemble and Deep Ensemble, both with five submodels, were evaluated for GTVp segmentation and uncertainty performance. The segmentation performance was evaluated using the volumetric Dice similarity coefficient (DSC), mean surface distance (MSD), and Hausdorff distance at 95% (95HD). The uncertainty was evaluated using four measures from literature: coefficient of variation (CV), structure expected entropy, structure predictive entropy, and structure mutual information, and additionally with our novel Dice-risk measure. The utility of uncertainty information was evaluated with the accuracy of uncertainty-based segmentation performance prediction using the Accuracy vs Uncertainty (AvU) metric, and by examining the linear correlation between uncertainty estimates and DSC. In addition, batch-based and instance-based referral processes were examined, where the patients with high uncertainty were rejected from the set. In the batch referral process, the area under the referral curve with DSC (R-DSC AUC) was used for evaluation, whereas in the instance referral process, the DSC at various uncertainty thresholds were examined.

Results

Both models behaved similarly in terms of the segmentation performance and uncertainty estimation. Specifically, the MC Dropout Ensemble had 0.776 DSC, 1.703 mm MSD, and 5.385 mm 95HD. The Deep Ensemble had 0.767 DSC, 1.717 mm MSD, and 5.477 mm 95HD. The uncertainty measure with the highest DSC correlation was structure predictive entropy with correlation coefficients of 0.699 and 0.692 for the MC Dropout Ensemble and the Deep Ensemble, respectively. The highest AvU value was 0.866 for both models. The best performing uncertainty measure for both models was the CV which had R-DSC AUC of 0.783 and 0.782 for the MC Dropout Ensemble and Deep Ensemble, respectively. With referring patients based on uncertainty thresholds from 0.85 validation DSC for all uncertainty measures, on average the DSC improved from the full dataset by 4.7% and 5.0% while referring 21.8% and 22% patients for MC Dropout Ensemble and Deep Ensemble, respectively.

Conclusion

We found that many of the investigated methods provide overall similar but distinct utility in terms of predicting segmentation quality and referral performance. These findings are a critical first-step towards more widespread implementation of uncertainty quantification in OPC GTVp segmentation.

Assessing the Impact of Image Resolution on Deep Learning for TB Lesion Segmentation on Frontal Chest X-rays

Deep learning (DL) models are state-of-the-art in segmenting anatomical and disease regions of interest (ROIs) in medical images. Particularly, a large number of DL-based techniques have been reported using chest X-rays (CXRs). However, these models are reportedly trained on reduced image resolutions for reasons related to the lack of computational resources. Literature is sparse in discussing the optimal image resolution to train these models for segmenting the tuberculosis (TB)-consistent lesions in CXRs. In this study, we investigated the performance variations with an Inception-V3 UNet model using various image resolutions with/without lung ROI cropping and aspect ratio adjustments and identified the optimal image resolution through extensive empirical evaluations to improve TB-consistent lesion segmentation performance. We used the Shenzhen CXR dataset for the study, which includes 326 normal patients and 336 TB patients. We proposed a combinatorial approach consisting of storing model snapshots, optimizing segmentation threshold and test-time augmentation (TTA), and averaging the snapshot predictions, to further improve performance with the optimal resolution. Our experimental results demonstrate that higher image resolutions are not always necessary; however, identifying the optimal image resolution is critical to achieving superior performance.

GEMA-An Automatic Segmentation Method for Real-Time Analysis of Mammalian Cell Growth in Microfluidic Devices

Nowadays, image analysis has a relevant role in most scientific and research areas. This process is used to extract and understand information from images to obtain a model, knowledge, and rules in the decision process. In the case of biological areas, images are acquired to describe the behavior of a biological agent in time such as cells using a mathematical and computational approach to generate a system with automatic control. In this paper, MCF7 cells are used to model their growth and death when they have been injected with a drug. These mammalian cells allow understanding of behavior, gene expression, and drug resistance to breast cancer. For this, an automatic segmentation method called GEMA is presented to analyze the apoptosis and confluence stages of culture by measuring the increase or decrease of the image area occupied by cells in microfluidic devices. In vitro, the biological experiments can be analyzed through a sequence of images taken at specific intervals of time. To automate the image segmentation, the proposed algorithm is based on a Gabor filter, a coefficient of variation (CV), and linear regression. This allows the processing of images in real time during the evolution of biological experiments. Moreover, GEMA has been compared with another three representative methods such as gold standard (manual segmentation), morphological gradient, and a semi-automatic algorithm using FIJI. The experiments show promising results, due to the proposed algorithm achieving an accuracy above 90% and a lower computation time because it requires on average 1 s to process each image. This makes it suitable for image-based real-time automatization of biological lab-on-a-chip experiments.

Ensemble learning via supervision augmentation for white matter hyperintensity segmentation

Since the ambiguous boundary of the lesion and inter-observer variability, white matter hyperintensity segmentation annotations are inherently noisy and uncertain. On the other hand, the high capacity of deep neural networks (DNN) enables them to overfit labels with noise and uncertainty, which may lead to biased models with weak generalization ability. This challenge has been addressed by leveraging multiple annotations per image. However, multiple annotations are often not available in a real-world scenario. To mitigate the issue, this paper proposes a supervision augmentation method (SA) and combines it with ensemble learning (SA-EN) to improve the generalization ability of the model. SA can obtain diverse supervision information by estimating the uncertainty of annotation in a real-world scenario that per image have only one ambiguous annotation. Then different base learners in EN are trained with diverse supervision information. The experimental results on two white matter hyperintensity segmentation datasets demonstrate that SA-EN gets the optimal accuracy compared with other state-of-the-art ensemble methods. SA-EN is more effective on small datasets, which is more suitable for medical image segmentation with few annotations. A quantitative study is presented to show the effect of ensemble size and the effectiveness of the ensemble model. Furthermore, SA-EN can capture two types of uncertainty, aleatoric uncertainty modeled in SA and epistemic uncertainty modeled in EN.

A Systematic Evaluation of Ensemble Learning Methods for Fine-Grained Semantic Segmentation of Tuberculosis-Consistent Lesions in Chest Radiographs

Automated segmentation of tuberculosis (TB)-consistent lesions in chest X-rays (CXRs) using deep learning (DL) methods can help reduce radiologist effort, supplement clinical decision-making, and potentially result in improved patient treatment. The majority of works in the literature discuss training automatic segmentation models using coarse bounding box annotations. However, the granularity of the bounding box annotation could result in the inclusion of a considerable fraction of false positives and negatives at the pixel level that may adversely impact overall semantic segmentation performance. This study evaluates the benefits of using fine-grained annotations of TB-consistent lesions toward training the variants of U-Net models and constructing their ensembles for semantically segmenting TB-consistent lesions in both original and bone-suppressed frontal CXRs. The segmentation performance is evaluated using several ensemble methods such as bitwise- AND, bitwise-OR, bitwise-MAX, and stacking. Extensive empirical evaluations showcased that the stacking ensemble demonstrated superior segmentation performance (Dice score: 0.5743, 95% confidence interval: (0.4055, 0.7431)) compared to the individual constituent models and other ensemble methods. To the best of our knowledge, this is the first study to apply ensemble learning to improve fine-grained TB-consistent lesion segmentation performance.

CAFS: An Attention-Based Co-Segmentation Semi-Supervised Method for Nasopharyngeal Carcinoma Segmentation

Accurate segmentation of nasopharyngeal carcinoma is essential to its treatment effect. However, there are several challenges in existing deep learning-based segmentation methods. First, the acquisition of labeled data are challenging. Second, the nasopharyngeal carcinoma is similar to the surrounding tissues. Third, the shape of nasopharyngeal carcinoma is complex. These challenges make the segmentation of nasopharyngeal carcinoma difficult. This paper proposes a novel semi-supervised method named CAFS for automatic segmentation of nasopharyngeal carcinoma. CAFS addresses the above challenges through three mechanisms: the teacher-student cooperative segmentation mechanism, the attention mechanism, and the feedback mechanism. CAFS can use only a small amount of labeled nasopharyngeal carcinoma data to segment the cancer region accurately. The average DSC value of CAFS is 0.8723 on the nasopharyngeal carcinoma segmentation task. Moreover, CAFS has outperformed the state-of-the-art nasopharyngeal carcinoma segmentation methods in the comparison experiment. Among the compared state-of-the-art methods, CAFS achieved the highest values of DSC, Jaccard, and precision. In particular, the DSC value of CAFS is 7.42% higher than the highest DSC value in the state-of-the-art methods.

Spine-GFlow: A hybrid learning framework for robust multi-tissue segmentation in lumbar MRI without manual annotation

Most learning-based magnetic resonance image (MRI) segmentation methods rely on the manual annotation to provide supervision, which is extremely tedious, especially when multiple anatomical structures are required. In this work, we aim to develop a hybrid framework named Spine-GFlow that combines the image features learned by a CNN model and anatomical priors for multi-tissue segmentation in a sagittal lumbar MRI. Our framework does not require any manual annotation and is robust against image feature variation caused by different image settings and/or underlying pathology. Our contributions include: 1) a rule-based method that automatically generates the weak annotation (initial seed area), 2) a novel proposal generation method that integrates the multi-scale image features and anatomical prior, 3) a comprehensive loss for CNN training that optimizes the pixel classification and feature distribution simultaneously. Our Spine-GFlow has been validated on 2 independent datasets: HKDDC (containing images obtained from 3 different machines) and IVDM3Seg. The segmentation results of vertebral bodies (VB), intervertebral discs (IVD), and spinal canal (SC) are evaluated quantitatively using intersection over union (IoU) and the Dice coefficient. Results show that our method, without requiring manual annotation, has achieved a segmentation performance comparable to a model trained with full supervision (mean Dice 0.914 vs 0.916).

Uncertainty Quantification in Segmenting Tuberculosis-Consistent Findings in Frontal Chest X-rays

Deep learning (DL) methods have demonstrated superior performance in medical image segmentation tasks. However, selecting a loss function that conforms to the data characteristics is critical for optimal performance. Further, the direct use of traditional DL models does not provide a measure of uncertainty in predictions. Even high-quality automated predictions for medical diagnostic applications demand uncertainty quantification to gain user trust. In this study, we aim to investigate the benefits of (i) selecting an appropriate loss function and (ii) quantifying uncertainty in predictions using a VGG16-based-U-Net model with the Monto-Carlo (MCD) Dropout method for segmenting Tuberculosis (TB)-consistent findings in frontal chest X-rays (CXRs). We determine an optimal uncertainty threshold based on several uncertainty-related metrics. This threshold is used to select and refer highly uncertain cases to an expert. Experimental results demonstrate that (i) the model trained with a modified Focal Tversky loss function delivered superior segmentation performance (mean average precision (mAP): 0.5710, 95% confidence interval (CI): (0.4021,0.7399)), (ii) the model with 30 MC forward passes during inference further improved and stabilized performance (mAP: 0.5721, 95% CI: (0.4032,0.7410), and (iii) an uncertainty threshold of 0.7 is observed to be optimal to refer highly uncertain cases.