PolypMixNet: Enhancing semi-supervised polyp segmentation with polyp-aware augmentation.

BACKGROUND: AI-assisted polyp segmentation in colonoscopy plays a crucial role in enabling prompt diagnosis and treatment of colorectal cancer. However, the lack of sufficient annotated data poses a significant challenge for supervised learning approaches. Existing semi-supervised learning methods also suffer from performance degradation, mainly due to task-specific characteristics, such as class imbalance in polyp segmentation. PURPOSE: The purpose of this work is to develop an effective semi-supervised learning framework for accurate polyp segmentation in colonoscopy, addressing limited annotated data and class imbalance challenges. METHODS: We proposed PolypMixNet, a semi-supervised framework, for colorectal polyp segmentation, utilizing novel augmentation techniques and a Mean Teacher architecture to improve model performance. PolypMixNet introduces the polyp-aware mixup (PolypMix) algorithm and incorporates dual-level consistency regularization. PolypMix addresses the class imbalance in colonoscopy datasets and enhances the diversity of training data. By performing a polyp-aware mixup on unlabeled samples, it generates mixed images with polyp context along with their artificial labels. A polyp-directed soft pseudo-labeling (PDSPL) mechanism was proposed to generate high-quality pseudo labels and eliminate the dilution of lesion features caused by mixup operations. To ensure consistency in the training phase, we introduce the PolypMix prediction consistency (PMPC) loss and PolypMix attention consistency (PMAC) loss, enforcing consistency at both image and feature levels. Code is available at https://github.com/YChienHung/PolypMix. RESULTS: PolypMixNet was evaluated on four public colonoscopy datasets, achieving 88.97% Dice and 88.85% mIoU on the benchmark dataset of Kvasir-SEG. In scenarios where the labeled training data is limited to 15%, PolypMixNet outperforms the state-of-the-art semi-supervised approaches with a 2.88-point improvement in Dice. It also shows the ability to reach performance comparable to the fully supervised counterpart. Additionally, we conducted extensive ablation studies to validate the effectiveness of each module and highlight the superiority of our proposed approach. CONCLUSION: PolypMixNet effectively addresses the challenges posed by limited annotated data and unbalanced class distributions in polyp segmentation. By leveraging unlabeled data and incorporating novel augmentation and consistency regularization techniques, our method achieves state-of-the-art performance. We believe that the insights and contributions presented in this work will pave the way for further advancements in semi-supervised polyp segmentation and inspire future research in the medical imaging domain.

Boundary-enhanced semi-supervised retinal layer segmentation in optical coherence tomography images using fewer labels.

Automatic segmentation of multiple layers in retinal optical coherence tomography (OCT) images is crucial for eye disease diagnosis and treatment. Despite the success of deep learning algorithms, it still remains a challenge due to the blurry layer boundaries and lack of adequate pixel-wise annotations. To tackle these issues, we propose a Boundary-Enhanced Semi-supervised Network (BE-SemiNet) that exploits an auxiliary distance map regression task to improve retinal layer segmentation with scarce labeled data and abundant unlabeled data. Specifically, a novel Unilaterally Truncated Distance Map (UTDM) is firstly introduced to alleviate the class imbalance problem and enhance the layer boundary learning in the regression task. Then for the pixel-wise segmentation and UTDM regression branches, we impose task-level and data-level consistency regularization on unlabeled data to enrich the diversity of unsupervised information and improve the regularization effects. Pseudo supervision is incorporated in consistency regularization to bridge the task prediction spaces for consistency and expand training labeled data. Experiments on two public retinal OCT datasets show that our method can greatly improve the supervised baseline performance with only 5 annotations and outperform the state-of-the-art methods. Since it is difficult and labor-expensive to obtain adequate pixel-wise annotations in practice, our method has a promising application future in clinical retinal OCT image analysis.

A photoacoustics-enhanced drilling probe for radiation-free pedicle screw implantation in spinal surgery.

This article proposes a novel intra-operative navigation and sensing system that optimizes the functional accuracy of spinal pedicle screw implantation. It does so by incorporating radiation-free and multi-scale macroscopic 3D ultrasound (US) imaging and local tissue-awareness from in situ photoacoustic (PA) sensing at a clinically relevant mesoscopic scale. More specifically, 3D US imaging is employed for online status updates of spinal segment posture to determine the appropriate entry point and coarse drilling path once non-negligible or relative patient motion occurs between inter-vertebral segments in the intra-operative phase. Furthermore, a sophisticated sensor-enhanced drilling probe has been developed to facilitate fine-grained local navigation that integrates a PA endoscopic imaging component for in situ tissue sensing. The PA signals from a sideways direction to differentiate cancellous bone from harder cortical bone, or to indicate weakened osteoporotic bone within the vertebrae. In so doing it prevents cortical breaches, strengthens implant stability, and mitigates iatrogenic injuries of the neighboring artery and nerves. To optimize this PA-enhanced endoscopic probe design, the light absorption spectrum of cortical bone and cancellous bone are measured in vitro, and the associated PA signals are characterized. Ultimately, a pilot study is performed on an ex vivo bovine spine to validate our developed multi-scale navigation and sensing system. The experimental results demonstrate the clinical feasibility, and hence the great potential, for functionally accurate screw implantation in complex spinal stabilization interventions.

Design and Control of a Magnetically-Actuated Capsule Robot With Biopsy Function.

OBJECTIVE: Wireless capsule endoscopy has been well used for gastrointestinal (GI) tract diagnosis. However, it can only obtain images and cannot take samples of GI tract tissues. In this study, we designed a magnetically-actuated biopsy capsule (MABC) robot for GI tract diagnosis. METHODS: The proposed robot can achieve locomotion and biopsy functions under the control of external electromagnetic actuation (EMA) system. Two types of active locomotion can be achieved, plane motion refers to the robot rolling on the surface of the GI tract with a rotating uniform magnetic field. 3D motion refers to the robot moving in 3D space under the control of the EMA system. After reaching the target position, the biopsy needle can be sprung out for sampling and then retracted under a gradient magnetic field. RESULTS: A pill-shaped robot prototype ( ϕ15 mm×32 mm) has been fabricated and tested with phantom experiments. The average motion control error is 0.32 mm in vertical direction, 3.3 mm in horizontal direction, and the maximum sampling error is about 5.0 mm. The average volume of the sampled tissue is about 0.35 mm 3. CONCLUSION: We designed a MABC robot and proposed a control framework which enables planar and 3D spatial locomotion and biopsy sampling. SIGNIFICANCE: The untethered MABC robot can be remotely controlled to achieve accurate sampling in multiple directions without internal power sources, paving the way towards precision sampling techniques for GI diseases in clinical procedures.

APRNet: Alternative Prediction Refinement Network for Polyp Segmentation.

Colorectal cancer has become the second leading cause of cancer-related death, attracting considerable interest for automatic polyp segmentation in polyp screening system. Accurate segmentation of polyps from colonoscopy is a challenging task as the polyps diverse in color, size and texture while the boundary between polyp and background is sometimes ambiguous. We propose a novel alternative prediction refinement network (APRNet) to more accurately segment polyps. Based on the UNet architecture, our APRNet aims at exploiting all-level features by alternatively leveraging features from encoder and decoder branch. Specifically, a series of prediction residual refinement modules (PRR) learn the residual and progressively refine the segmentation at various resolution. The proposed APRNet is evaluated on two benchmark datasets and achieves new state-of-the-art performance with a dice of 91.33% and an accuracy of 97.31% on the Kvasir-SEG dataset, and a dice of 86.33% and an accuracy of 97.12% on the EndoScene dataset.Clinical relevance- This work proposes an automatic and accurate polyp segmentation algorithm that achieves new state- of-the-art performance, which can potentially act as an observer pointing out polyps in colonoscopy procedure.

A Virtual Scanning Framework for Robotic Spinal Sonography with Automatic Real-time Recognition of Standard Views.

Ultrasound (US) imaging is widely used to assist in the diagnosis and intervention of the spine, but the manual scanning process would bring heavy physical and cognitive burdens on the sonographers. Robotic US acquisitions can provide an alternative to the standard handheld technique to reduce operator workload and avoid direct patient contact. However, the real-time interpretation of the acquired images is rarely addressed in existing robotic US systems. Therefore, we envision a robotic system that can automatically scan the spine and search for the standard views like an expert sonographer. In this work, we propose a virtual scanning framework based on real-world US data acquired by a robotic system to simulate the autonomous robotic spinal sonography, and incorporate automatic real-time recognition of the standard views of the spine based on a multi-scale fusion approach and deep convolutional neural networks. Our method can accurately classify 96.71% of the standard views of the spine in the test set, and the simulated clinical application preliminarily demonstrates the potential of our method.

Development of a Compact Photoacoustic Tomography Imaging System with Dual Single-Element Transducers for Image Enhancement.

OBJECTIVE: This paper proposes a new photoacoustic computed tomography (PACT) imaging system employing dual ultrasonic transducers with different frequencies. When imaging complex biological tissues, photoacoustic (PA) signals with multiple frequencies are produced simultaneously; however, due to the limited bandwidth of a single-frequency transducer, the received PA signals with specific frequencies may be missing, leading to a low imaging quality. METHODS: In contrast to our previous work, the proposed system has a compact volume as well as specific selection of the detection center frequency of the transducer, which can provide a comprehensive range for the detection of PA signals. In this study, a series of numerical simulation and phantom experiments were performed to validate the efficacy of the developed PACT system. RESULTS: The images generated by our system combined the advantages of both high resolution and ideal brightness/contrast. CONCLUSION: The interchangeability of transducers with different frequencies provides potential for clinical deployment under the circumstance where a single frequency transducer cannot perform well.

Tip estimation approach for concentric tube robots using 2D ultrasound images and kinematic model.

Concentric tube robot (CTR) is an efficient approach for minimally invasive surgery (MIS) and diagnosis due to its small size and high dexterity. To manipulate the robot accurately and safely inside the human body, tip position and shape information need to be well measured. In this paper, we propose a tip estimation method based on 2D ultrasound images with the help of the forward kinematic model of CTR. The forward kinematic model can help to provide a fast ultrasound scanning path and narrow the region of interest in ultrasound images. For each tube, only three scan positions are needed by combining the kinematic model prediction as prior knowledge. After that, the curve fitting method is used for its shape reconstruction, while its tip position can be estimated based on the constraints of its structure and length.7 This method provides the advantage that only three scan positions are needed for estimating the tip of each telescoping section. Moreover, no structure modification is needed on the robot, which makes it an appropriate approach for existing flexible surgical robots. Experimental results verified the feasibility of the proposed method and the tip estimation error is 0.59 mm. Graphical abstract In this paper, we propose a tip estimation method based on 2D Ultrasound images with the help of the forward kinematic model of CTR. The forward kinematic model can help to provide a fast Ultrasound scanning path and narrow the region of interest in Ultrasound images. For each tube, only three scan positions are needed by combining the kinematic model prediction as prior knowledge. After that, the curve fitting method is used for its shape reconstruction, while its tip position can be estimated based on the constraints of its structure and length.

A Novel Approach for Automatic State Detection of A Magnetically Actuated Capsule.

In recent years, the Simultaneous Magnetic Actuation and Localization (SMAL) technology has been developed to accelerate and locate the wireless capsule endoscopy (WCE) in the intestine. In this paper, we propose a novel approach to detect the state of the capsule for improving the localization results. By creating a function to fit the relationship between the theoretical values of the actuating magnetic field and the measurement results, we present an algorithm for automatic estimation of the capsule state according to the fitting parameters. Experiment results on phantoms demonstrate the feasibility of the proposed method for detecting different states of the capsule during magnetic actuation.

General first-order target registration error model considering a coordinate reference frame in an image-guided surgical system.

Point-based rigid registration (PBRR) techniques are widely used in many aspects of image-guided surgery (IGS). Accurately estimating target registration error (TRE) statistics is of essential value for medical applications such as optically surgical tool-tip tracking and image registration. For example, knowing the TRE distribution statistics of surgical tool tip can help the surgeon make right decisions during surgery. In the meantime, the pose of a surgical tool is usually reported relative to a second rigid body whose local frame is called coordinate reference frame (CRF). In an n-ocular tracking system, fiducial localization error (FLE) should be considered inhomogeneous, that means FLE is different between fiducials, and anisotropic that indicates FLE is different in all directions. In this paper, we extend the TRE estimation algorithm relative to a CRF from homogeneous and anisotropic to heterogeneous FLE cases. Arbitrary weightings can be assumed in solving the registration problems in the proposed TRE estimation algorithm. Monte Carlo simulation results demonstrate the proposed algorithm's effectiveness for both homogeneous and inhomogeneous FLE distributions. The results are further compared with those using the other two algorithms. When FLE distribution is anisotropic and homogeneous, the proposed TRE estimation algorithm's performance is comparable with that of the first one. When FLE distribution is heterogeneous, proposed TRE estimation algorithm outperforms the other two classical algorithms in all test cases when ideal weighting scheme is adopted in solving two registrations. Possible clinical applications include the online estimation of surgical tool-tip tracking error with respect to a CRF in IGS. Graphical Abstract This paper provides the target registration error model considering a coordinate reference frame in surgical navigation.

Zoom in Lesions for Better Diagnosis: Attention Guided Deformation Network for WCE Image Classification.

Wireless capsule endoscopy (WCE) is a novel imaging tool that allows noninvasive visualization of the entire gastrointestinal (GI) tract without causing discomfort to patients. Convolutional neural networks (CNNs), though perform favorably against traditional machine learning methods, show limited capacity in WCE image classification due to the small lesions and background interference. To overcome these limits, we propose a two-branch Attention Guided Deformation Network (AGDN) for WCE image classification. Specifically, the attention maps of branch1 are utilized to guide the amplification of lesion regions on the input images of branch2, thus leading to better representation and inspection of the small lesions. What's more, we devise and insert Third-order Long-range Feature Aggregation (TLFA) modules into the network. By capturing long-range dependencies and aggregating contextual features, TLFAs endow the network with a global contextual view and stronger feature representation and discrimination capability. Furthermore, we propose a novel Deformation based Attention Consistency (DAC) loss to refine the attention maps and achieve the mutual promotion of the two branches. Finally, the global feature embeddings from the two branches are fused to make image label predictions. Extensive experiments show that the proposed AGDN outperforms state-of-the-art methods with an overall classification accuracy of 91.29% on two public WCE datasets. The source code is available at https://github.com/hathawayxxh/WCE-AGDN.

Design of a multi-arm concentric-tube robot system for transnasal surgery.

Concentric tube robot (CTR) has gradually attracted the attention of researchers on the basis of its small size and curved shape control ability. However, most of current experimental prototypes of CTR are single-arm structure, which can only carry out simple operation such as drug delivery or monitoring. In this paper, design and analysis of a three-arm CTR system is proposed. It has a four-DOF vision arm and two six-DOF manipulator arms, which equipped with special end effectors to achieve different surgical operations. Finally, a mean motion accuracy of 0.33 mm has been obtained quantitatively through teleoperation experiments. Moreover, tissue excision experiment in skull model is carried out to prove the effectiveness and feasibility of the proposed CTR system in nasopharyngeal carcinoma surgery. Graphical Abstract Platform of the proposed Multi-Arm Concentric Tube Robot system. (a) Configuration of the end-effectors with the CTR system. (b) The setup of the tissue removal experiment in a skull model.

Pilot Study of Trans-oral Robotic-Assisted Needle Direct Tracheostomy Puncture in Patients Requiring Prolonged Mechanical Ventilation.

COVID-19 can induce severe respiratory problems that need prolonged mechanical ventilation in the intensive care unit. While Open Tracheostomy (OT) is the preferred technique due to the excellent visualization of the surgical field and structures, Percutaneous Tracheostomy (PT) has proven to be a feasible minimally invasive alternative. However, PT's limitation relates to the inability to precisely enter the cervical trachea at the exact spot since the puncture is often performed based on crude estimation from anatomical laryngeal surface landmarks. Besides, there is no absolute control of the trajectory and force required to make the percutaneous puncture into the trachea, resulting in inadvertent injury to the cricoid ring, cervical esophagus, and vessels in the neck. Therefore, we hypothesize that a flexible mini-robotic system, incorporating the robotic needling technology, can overcome these challenges by allowing the trans-oral robotic instrument of the cervical trachea. This approach promises to improve current PT technology by making the initial trachea puncture from an "inside-out" approach, rather than an "outside-in" manner, fraught with several technical uncertainties.

Augmented Bladder Tumor Detection Using Deep Learning.

Adequate tumor detection is critical in complete transurethral resection of bladder tumor (TURBT) to reduce cancer recurrence, but up to 20% of bladder tumors are missed by standard white light cystoscopy. Deep learning augmented cystoscopy may improve tumor localization, intraoperative navigation, and surgical resection of bladder cancer. We aimed to develop a deep learning algorithm for augmented cystoscopic detection of bladder cancer. Patients undergoing cystoscopy/TURBT were recruited and white light videos were recorded. Video frames containing histologically confirmed papillary urothelial carcinoma were selected and manually annotated. We constructed CystoNet, an image analysis platform based on convolutional neural networks, for automated bladder tumor detection using a development dataset of 95 patients for algorithm training and five patients for testing. Diagnostic performance of CystoNet was validated prospectively in an additional 54 patients. In the validation dataset, per-frame sensitivity and specificity were 90.9% (95% confidence interval [CI], 90.3-91.6%) and 98.6% (95% CI, 98.5-98.8%), respectively. Per-tumor sensitivity was 90.9% (95% CI, 90.3-91.6%). CystoNet detected 39 of 41 papillary and three of three flat bladder cancers. With high sensitivity and specificity, CystoNet may improve the diagnostic yield of cystoscopy and efficacy of TURBT. PATIENT SUMMARY: Conventional cystoscopy has recognized shortcomings in bladder cancer detection, with implications for recurrence. Cystoscopy augmented with artificial intelligence may improve cancer detection and resection.

Safe and Robust Mobile Robot Navigation in Uneven Indoor Environments.

Complex environments pose great challenges for autonomous mobile robot navigation. In this study, we address the problem of autonomous navigation in 3D environments with staircases and slopes. An integrated system for safe mobile robot navigation in 3D complex environments is presented and both the perception and navigation capabilities are incorporated into the modular and reusable framework. Firstly, to distinguish the slope from the staircase in the environment, the robot builds a 3D OctoMap of the environment with a novel Simultaneously Localization and Mapping (SLAM) framework using the information of wheel odometry, a 2D laser scanner, and an RGB-D camera. Then, we introduce the traversable map, which is generated by the multi-layer 2D maps extracted from the 3D OctoMap. This traversable map serves as the input for autonomous navigation when the robot faces slopes and staircases. Moreover, to enable robust robot navigation in 3D environments, a novel camera re-localization method based on regression forest towards stable 3D localization is incorporated into this framework. In addition, we utilize a variable step size Rapidly-exploring Random Tree (RRT) method which can adjust the exploring step size automatically without tuning this parameter manually according to the environment, so that the navigation efficiency is improved. The experiments are conducted in different kinds of environments and the output results demonstrate that the proposed system enables the robot to navigate efficiently and robustly in complex 3D environments.

A Novel Relative Position Estimation Method for Capsule Robot Moving in Gastrointestinal Tract.

Recently, a variety of positioning and tracking methods have been proposed for capsule robots moving in the gastrointestinal (GI) tract to provide real-time unobstructed spatial pose results. However, the current absolute position-based result cannot match the GI structure due to its unstructured environment. To overcome this disadvantage and provide a proper position description method to match the GI tract, we here present a relative position estimation method for tracking the capsule robot, which uses the moving distance of the robot along the GI tract to indicate the position result. The procedure of the proposed method is as follows: firstly, the absolute position results of the capsule robot are obtained with the magnetic tracking method; then, the moving status of the robot along the GI tract is determined according to the moving direction; and finally, the movement trajectory of the capsule robot is fitted with the Bézier curve, where the moving distance can then be evaluated using the integral method. Compared to state-of-the-art capsule tracking methods, the proposed method can directly help to guide medical instruments by providing physicians the insertion distance in patients' bodies, which cannot be done based on absolute position results. Moreover, as relative distance information was used, no reference tracking objects needed to be mounted onto the human body. The experimental results prove that the proposed method achieves a good distance estimation of the capsule robot moving in the simulation platform.

Socially Compliant Path Planning for Robotic Autonomous Luggage Trolley Collection at Airports.

This paper describes a socially compliant path planning scheme for robotic autonomous luggage trolley collection at airports. The robot is required to efficiently collect all assigned luggage trolleys in a designated area, while avoiding obstacles and not offending the pedestrians. This path planning problem is formulated in this paper as a Traveling Salesman Problem (TSP). Different from the conventional solutions to the TSP, in which the Euclidean distance between two sites is used as the metric, a high-dimensional metric including the factor of pedestrians' feelings is applied in this work. To obtain the new metric, a novel potential function is firstly proposed to model the relationship between the robot, luggage trolleys, obstacles, and pedestrians. The Social Force Model (SFM) is utilized so that the pedestrians can bring extra influence on the potential field, different from ordinary obstacles. Directed by the attractive and repulsive force generated from the potential field, a number of paths connecting the robot and the luggage trolley, or two luggage trolleys, can be obtained. The length of the generated path is considered as the new metric. The Self-Organizing Map (SOM) satisfies the job of finding a final path to connect all luggage trolleys and the robot located in the potential field, as it can find the intrinsic connection in the high dimensional space. Therefore, while incorporating the new metric, the SOM is used to find the optimal path in which the robot can collect the assigned luggage trolleys in sequence. As a demonstration, the proposed path planning method is implemented in simulation experiments, showing an increase of efficiency and efficacy.

An Improved Calibration Method for a Rotating 2D LIDAR System.

This paper presents an improved calibration method of a rotating two-dimensional light detection and ranging (R2D-LIDAR) system, which can obtain the 3D scanning map of the surroundings. The proposed R2D-LIDAR system, composed of a 2D LIDAR and a rotating unit, is pervasively used in the field of robotics owing to its low cost and dense scanning data. Nevertheless, the R2D-LIDAR system must be calibrated before building the geometric model because there are assembled deviation and abrasion between the 2D LIDAR and the rotating unit. Hence, the calibration procedures should contain both the adjustment between the two devices and the bias of 2D LIDAR itself. The main purpose of this work is to resolve the 2D LIDAR bias issue with a flat plane based on the Levenberg-Marquardt (LM) algorithm. Experimental results for the calibration of the R2D-LIDAR system prove the reliability of this strategy to accurately estimate sensor offsets with the error range from -15 mm to 15 mm for the performance of capturing scans.

Automatic Polyp Detection via a Novel Unified Bottom-Up and Top-Down Saliency Approach.

In this paper, we propose a novel automatic computer-aided method to detect polyps for colonoscopy videos. To capture perceptually and semantically meaningful salient polyp regions, we first segment images into multilevel superpixels. Each level corresponds to different sizes of superpixels. Rather than adopting hand-designed features to describe these superpixels in images, we employ sparse autoencoder (SAE) to learn discriminative features in an unsupervised way. Then, a novel unified bottom-up and top-down saliency method is proposed to detect polyps. In the first stage, we propose a weak bottom-up (WBU) saliency map by fusing the contrast-based saliency and object-center-based saliency together. The contrast-based saliency map highlights image parts that show different appearances compared with surrounding areas, whereas the object-center-based saliency map emphasizes the center of the salient object. In the second stage, a strong classifier with multiple kernel boosting is learned to calculate the strong top-down (STD) saliency map based on samples directly from the obtained multilevel WBU saliency maps. We finally integrate these two-stage saliency maps from all levels together to highlight polyps. Experiment results achieve 0.818 recall for saliency calculation, validating the effectiveness of our method. Extensive experiments on public polyp datasets demonstrate that the proposed saliency algorithm performs better compared with state-of-the-art saliency methods to detect polyps.

Discriminative Joint-Feature Topic Model With Dual Constraints for WCE Classification.

Wireless capsule endoscopy (WCE) enables clinicians to examine the digestive tract without any surgical operations, at the cost of a large amount of images to be analyzed. The main challenge for automatic computer-aided diagnosis arises from the difficulty of robust characterization of these images. To tackle this problem, a novel discriminative joint-feature topic model (DJTM) with dual constraints is proposed to classify multiple abnormalities in WCE images. We first propose a joint-feature probabilistic latent semantic analysis (PLSA) model, where color and texture descriptors extracted from same image patches are jointly modeled with their conditional distributions. Then the proposed dual constraints: visual words importance and local image manifold are embedded into the joint-feature PLSA model simultaneously to obtain discriminative latent semantic topics. The visual word importance is proposed in our DJTM to guarantee that visual words with similar importance come from close latent topics while the local image manifold constraint enforces that images within the same category share similar latent topics. Finally, each image is characterized by distribution of latent semantic topics instead of low level features. Our proposed DJTM showed an excellent overall recognition accuracy 90.78%. Comprehensive comparison results demonstrate that our method outperforms existing multiple abnormalities classification methods for WCE images.