Optimal Segmentation Research Articles

Features extraction of point clouds based on Otsu’s algorithm

Currently, a point cloud extraction method based on geometric features requires the configuration of two essential parameters: the neighborhood radius within the point cloud and the criterion for feature threshold selection. This article addresses the issue of manual selection of feature thresholds and proposes a feature extraction method for 3D point clouds based on the Otsu algorithm. Firstly, the curvature value of each point cloud is calculated based on the r-neighborhood of the point cloud data. Secondly, the Otsu algorithm is improved by taking the curvature values as input for the maximum inter-class variance method. The optimal segmentation threshold is obtained based on the Otsu algorithm to divide the point cloud data into two parts. Point cloud data whose curvature is greater than or equal to the threshold is extracted as feature point data. In order to verify the reliability of the algorithm presented in this paper, a method for accuracy assessment of regular point cloud data is proposed. Additionally, comparative analysis was conducted on data with varying point cloud densities and on data contaminated with Gaussian white noise using multiple methods. Experimental results show that the proposed algorithm achieves good extraction results for data with 90% simplification rate and low noise.

Advancing Borehole Imaging: A Classification Database Developed via Adaptive Ring Segmentation

The use of in-hole imaging to investigate geological structure characteristics is one of the crucial methods for the study of rock mass stability and rock engineering design. The in-hole images are usually influenced by the lighting and imaging characteristics, resulting in the presence of interference noise regions in the images and consequently impacting the classification accuracy. To enhance the analytical efficacy of in-hole images, this paper employs the proposed optimal non-concentric ring segmentation method to establish a new database. This method establishes the transformation function based on the Ansel Adams Zone System and the fluctuation values of the grayscale mean, adjusting the gray-level distribution of images to extract two visual blind spots of different scales. Thus, the inner and outer circles are located with these blind spots to achieve the adaptive acquisition of the optimal ring. Finally, we use the optimal non-concentric ring segmentation method to traverse all original images to obtain the borehole image classification database. To validate the effectiveness of this method, we conduct experiments using various segmentation and classification evaluation metrics. The results show that the Jaccard and Dice of the optimal non-concentric ring segmentation approach are 88.43% and 98.55%, respectively, indicating superior segmentation performance compared to other methods. Furthermore, after employing four commonly used classification models to validate the performance of the new classification database, the results demonstrate a significant improvement in accuracy and macro-average compared to the original database, with the highest increase in accuracy reaching 4.2%. These results fully demonstrate the effectiveness of the proposed optimal non-concentric ring segmentation method.

Improving Skin Lesion Segmentation with Self-Training.

Skin lesion segmentation plays a key role in the diagnosis of skin cancer; it can be a component in both traditional algorithms and end-to-end approaches. The quality of segmentation directly impacts the accuracy of classification; however, attaining optimal segmentation necessitates a substantial amount of labeled data. Semi-supervised learning allows for employing unlabeled data to enhance the results of the machine learning model. In the case of medical image segmentation, acquiring detailed annotation is time-consuming and costly and requires skilled individuals so the utilization of unlabeled data allows for a significant mitigation of manual segmentation efforts. This study proposes a novel approach to semi-supervised skin lesion segmentation using self-training with a Noisy Student. This approach allows for utilizing large amounts of available unlabeled images. It consists of four steps-first, training the teacher model on labeled data only, then generating pseudo-labels with the teacher model, training the student model on both labeled and pseudo-labeled data, and lastly, training the student* model on pseudo-labels generated with the student model. In this work, we implemented DeepLabV3 architecture as both teacher and student models. As a final result, we achieved a mIoU of 88.0% on the ISIC 2018 dataset and a mIoU of 87.54% on the PH2 dataset. The evaluation of the proposed approach shows that Noisy Student training improves the segmentation performance of neural networks in a skin lesion segmentation task while using only small amounts of labeled data.

An enhanced chameleon swarm algorithm for global optimization and multi-level thresholding medical image segmentation

An enhanced chameleon swarm algorithm for global optimization and multi-level thresholding medical image segmentation

Rock mass quality prediction on tunnel faces with incomplete multi-source dataset via tree-augmented naive Bayesian network

Rock mass quality serves as a vital index for predicting the stability and safety status of rock tunnel faces. In tunneling practice, the rock mass quality is often assessed via a combination of qualitative and quantitative parameters. However, due to the harsh on-site construction conditions, it is rather difficult to obtain some of the evaluation parameters which are essential for the rock mass quality prediction. In this study, a novel improved Swin Transformer is proposed to detect, segment, and quantify rock mass characteristic parameters such as water leakage, fractures, weak interlayers. The site experiment results demonstrate that the improved Swin Transformer achieves optimal segmentation results and achieving accuracies of 92%, 81%, and 86% for water leakage, fractures, and weak interlayers, respectively. A multi-source rock tunnel face characteristic (RTFC) dataset includes 11 parameters for predicting rock mass quality is established. Considering the limitations in predictive performance of incomplete evaluation parameters exist in this dataset, a novel tree-augmented naive Bayesian network (BN) is proposed to address the challenge of the incomplete dataset and achieved a prediction accuracy of 88%. In comparison with other commonly used Machine Learning models the proposed BN-based approach proved an improved performance on predicting the rock mass quality with the incomplete dataset. By utilizing the established BN, a further sensitivity analysis is conducted to quantitatively evaluate the importance of the various parameters, results indicate that the rock strength and fractures parameter exert the most significant influence on rock mass quality.

A simulation framework to determine optimal strength training and musculoskeletal geometry for sprinting and distance running.

Musculoskeletal geometry and muscle volumes vary widely in the population and are intricately linked to the performance of tasks ranging from walking and running to jumping and sprinting. As an alternative to experimental approaches, where it is difficult to isolate factors and establish causal relationships, simulations can be used to independently vary musculoskeletal geometry and muscle volumes, and develop a fundamental understanding. However, our ability to understand how these parameters affect task performance has been limited due to the high computational cost of modelling the necessary complexity of the musculoskeletal system and solving the requisite multi-dimensional optimization problem. For example, sprinting and running are fundamental to many forms of sport, but past research on the relationships between musculoskeletal geometry, muscle volumes, and running performance has been limited to observational studies, which have not established cause-effect relationships, and simulation studies with simplified representations of musculoskeletal geometry. In this study, we developed a novel musculoskeletal simulator that is differentiable with respect to musculoskeletal geometry and muscle volumes. This simulator enabled us to find the optimal body segment dimensions and optimal distribution of added muscle volume for sprinting and marathon running. Our simulation results replicate experimental observations, such as increased muscle mass in sprinters, as well as a mass in the lower end of the healthy BMI range and a higher leg-length-to-height ratio in marathon runners. The simulations also reveal new relationships, for example showing that hip musculature is vital to both sprinting and marathon running. We found hip flexor and extensor moment arms were maximized to optimize sprint and marathon running performance, and hip muscles the main target when we simulated strength training for sprinters. Our simulation results provide insight to inspire future studies to examine optimal strength training. Our simulator can be extended to other athletic tasks, such as jumping, or to non-athletic applications, such as designing interventions to improve mobility in older adults or individuals with movement disorders.

Evaluation of optimal interpolation and segmentation of the optic nerves on magnetic resonance images for cross‐sectional area measurement

AbstractThis study investigated nine combination methods produced from three interpolation (Lanczos, iterative curvature based interpolation, and contrast‐guided) and three segmentation (spatial fuzzy C‐means, modified fuzzy C‐means [mFCM], and level set method) models from optic nerve magnetic resonance images (MRI). The aim was to produce sharp edges of the optic nerves for cross‐sectional area measurement. Lanczos‐mFCM was identified as the best combination in terms of: image quality; Dice similarity coefficient; a similar area compared with measurements from raw images but with higher reproducibility; and the highest signal‐to‐noise and contrast‐to‐noise ratios. It was then applied to ten normal datasets. The mean cross‐sectional areas were 12.57 ± 1.92mm2 from proton density images, 12.98 ± 2.18mm2 from T2‐weighted (T2W) images including the optic nerve sheath, and 1.68 ± 0.69mm2 from the T2W images of the optic nerve only. The Lanczos‐mFCM method is recommended for future quantitative studies on optic nerve MRI.

Recurrent segmentation meets block models in temporal networks

A popular approach to model interactions is to represent them as a network with nodes being the agents and the interactions being the edges. Interactions are often timestamped, which leads to having timestamped edges. Many real-world temporal networks have a recurrent or possibly cyclic behaviour. In this paper, our main interest is to model recurrent activity in such temporal networks. As a starting point we use stochastic block model, a popular choice for modelling static networks, where nodes are split into R groups. We extend the block model to temporal networks by modelling the edges with a Poisson process. We make the parameters of the process dependent on time by segmenting the time line into K segments. We require that only H≤K\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$H \\le K$$\\end{document} different set of parameters can be used. If H<K\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$H < K$$\\end{document}, then several, not necessarily consecutive, segments must share their parameters, modelling repeating behaviour. We propose two variants where a group membership of a node is fixed over the course of entire time line and group memberships are allowed to vary from segment to segment. We prove that searching for optimal groups and segmentation in both variants is NP-hard. Consequently, we split the problem into 3 subproblems where we optimize groups, model parameters, and segmentation in turn while keeping the remaining structures fixed. We propose an iterative algorithm that requires OKHm+Rn+R2H\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\mathcal {O} \\left( KHm + Rn + R^2\\,H\\right)$$\\end{document} time per iteration, where n and m are the number of nodes and edges in the network. We demonstrate experimentally that the number of required iterations is typically low, the algorithm is able to discover the ground truth from synthetic datasets, and show that certain real-world networks exhibit recurrent behaviour as the likelihood does not deteriorate when H is lowered.

A data-driven active learning approach to reusing ML solutions in scientific applications

Artificial intelligence can revolutionize scientific projects, but scientists face challenges in reusing, integrating, and deploying cost-effective and high-quality machine learning solutions. Determining suitable algorithms and parameters is difficult, especially for non-programmer scientists. Some algorithms, like deep learning-based methods, offer flexibility but require extensive training on annotated data. This poses a hurdle in labor-intensive tasks like biological image segmentation that relies on expert annotations.In this paper, we present a data-driven framework designed to assist scientists in selecting, reusing, and training machine learning solutions for microscopy image segmentation. The framework is based on establishing a mapping between object morphology features and the optimal segmentation algorithms and settings for individual objects. This mapping is iteratively refined through a combination of unsupervised learning and active learning iterations. To expedite convergence, objects are initially clustered based on their morphology. In each active learning iteration, the most informative and uncertain samples are selected and queried within a specific cluster. Through a biological case study, we demonstrate that our method enables the selection and training of segmentation algorithms specific to object types. Additionally, the selective requests for user input significantly reduce the number of user interactions required for this task.

Current and future on definitive concurrent chemoradiotherapy for inoperable locally advanced esophageal squamous cell carcinoma.

Esophageal squamous cell carcinoma (ESCC) is an aggressive and fatal disease that is usually diagnosed when the chances for surgical intervention has been missed. Definitive concurrent chemoradiotherapy (dCRT) is the first choice of treatment for inoperable locally advanced esophageal squamous cell carcinoma (LA-ESCC). Nevertheless, the local recurrence rate for esophageal cancer patients undergoing dCRT remains high at 40-60%, with a 5-year overall survival rate of solely 10-30%. Immunotherapy in combination with dCRT is a promising treatment for inoperable LA-ESCC, for that improved long-term survival is expected. The present review provides a comprehensive overview of the evolutionary trajectory of dCRT for LA-ESCC, delineates notable relevant clinical studies, addresses unresolved concerns regarding the combination of dCRT with immunotherapy, and highlights promising directions for future research. When dCRT is combined with immunotherapy, the following aspects should be carefully explored in the future studies, including the optimal irradiation dose, segmentation scheme, radiotherapy technique, timing, sequence and duration of radiotherapy, and the selection of chemotherapeutic and immunologic drugs. In addition, further investigations on the mechanisms of how dCRT combined with immunotherapy exerts synergistic anti-tumor effects and molecular biomarkers ensuring precise screening of ESCC patients are needed.

ETUNet:Exploring efficient transformer enhanced UNet for 3D brain tumor segmentation

Medical image segmentation is a crucial topic in medical image processing. Accurately segmenting brain tumor regions from multimodal MRI scans is essential for clinical diagnosis and survival prediction. However, similar intensity distributions, variable tumor shapes, and fuzzy boundaries pose severe challenges for brain tumor segmentation. Traditional segmentation networks based on UNet struggle to establish explicit long-range dependencies from the feature space due to the limitations of the CNN receptive field. This is particularly crucial for dense prediction tasks such as brain tumor segmentation. Recent works have incorporated the powerful global modeling capability of Transformer into UNet to achieve more precise segmentation results. Nevertheless, these methods encounter some issues: (1) the global information is often modeled by simply stacking Transformer layers for a specific module, resulting in high computational complexity and underutilization of the potential of the UNet architecture; (2) the rich boundary information of tumor subregions in multi-scale features is often overlooked. Motivated by these challenges, we propose an advanced fusion of Transformer with UNet by reexamining the core three parts (encoder, bottleneck, and skip connections). Firstly, we introduce a CNN-Transformer module in the encoder to replace the traditional CNN module, enabling the capture of deep spatial dependencies from input images. To address high-level semantic information, we incorporate a computationally efficient spatial-channel attention layer in the bottleneck for global interaction, highlighting important semantic features from the encoder path output. For irregular lesions, we fuse the multi-scale features from the encoder output and the decoder features in the skip connections by calculating cross-attention. This adaptive querying of valuable information from multi-scale features enhances the boundary localization ability of the decoder path and suppresses redundant features with low correlation. Compared to existing methods, our model further enhances the learning capacity of the overall UNet architecture while maintaining low computational complexity. Experimental results on the BraTS2018 and BraTS2020 datasets for brain tumor segmentation tasks demonstrate that our model achieves comparable or superior results compared to recent CNN or Transformer-based models. The average DSC and HD95 on the two datasets are 0.854, 6.688, and 0.862, 5.455 respectively. At the same time, our model achieves optimal segmentation of Enhancing tumors, showcasing the effectiveness of our method. Our code will be made publicly available at https://github.com/wzhangck/ETUnet.

Deep Learning Auto-Segmentation Network for Pediatric Computed Tomography Data Sets: Can We Extrapolate From Adults?

PurposeAI-based auto-segmentation models hold promise for enhanced efficiency and consistency in organ contouring for adaptive radiotherapy and radiotherapy planning. However, their performance on paediatric CT data and cross-scanner compatibility remains unclear. This study aims to evaluate the performance of AI-based auto-segmentation models trained on adult CT data when applied to paediatric datasets and explore the improvement in performance gained by including paediatric training data. It also examines their ability to accurately segment CT data acquired from different scanners. Methods and MaterialsUsing the nnU-Net framework, segmentation models were trained on datasets of adult, paediatric, and combined CT scans for seven pelvic/thoracic organs. Each model was trained on 290-300 cases per category and organ. Training datasets included a combination of clinical data and several open repositories. The study incorporated a database of 459 paediatric (0-16 years) CT scans and 950 adults (>18 years), ensuring all scans had human expert ground-truth contours of the selected organs. Performance was evaluated based on Dice similarity coefficients (DSC) of the model-generated contours. ResultsAI-models trained exclusively on adult data underperformed on paediatric data, especially for the 0-2 age group: mean DSC was below 0.5 for the bladder and spleen. The addition of paediatric training data demonstrated significant improvement for all age groups, achieving a mean DSC of above 0.85 for all organs in every age group. Larger organs like the liver and kidneys maintained consistent performance for all models across age groups. No significant difference emerged in the cross-scanner performance evaluation, suggesting robust cross-scanner generalization. ConclusionFor optimal segmentation across age groups, it is important to include paediatric data in the training of segmentation models. The successful cross-scanner generalization also supports the real-world clinical applicability of these AI models. This study emphasizes the significance of dataset diversity in training robust AI systems for medical image interpretation tasks.

Selection of Optimal Segmentation Algorithm for Satellite Images by Intuitionistic Fuzzy PROMETHEE Method

The combination of MCDM and fuzzy sets offers new potential ways to solve the challenges posed by complex image contents, such as selecting the optimal segmentation algorithm or evaluating the segmentation quality based on various parameters. Since no single segmentation algorithm can achieve the best results on satellite image datasets, it is essential to determine the most appropriate segmentation algorithm for each satellite image, the content of which can be characterized by relevant visual features. In this research, we proposed a set of visual criteria representing the fundamental aspects of satellite image segmentation. The segmentation algorithms chosen for testing were evaluated for their performance against each criterion. We introduced a new method to create a decision matrix for each image using fuzzy fusion, which combines the image content vector and the evaluation matrix of the studied segmentation algorithms. An extension of the Preference Ranking Organization Method Enrichment Evaluation (PROMETHEE) using intuitive fuzzy sets (IFSs) was applied to solve this problem. The results acquired by the proposed methodology were validated by comparing them with those obtained in expert ratings and by performing a sensitivity analysis.

Quantitative anatomical analysis of lumbar interspaces based on 3D CT imaging: optimized segment selection for lumbar puncture in different age groups

BackgroundOptimal lumbar puncture segment selection remains controversial. This study aims to analyze anatomical differences among L3-4, L4-5, and L5-S1 segments across age groups and provide quantitative evidence for optimized selection.Methods80 cases of CT images were collected with patients aged 10–80 years old. Threedimensional models containing L3-S1 vertebrae, dural sac, and nerve roots were reconstructed. Computer simulation determined the optimal puncture angles for the L3-4, L4-5, and L5-S1 segments. The effective dural sac area (ALDS), traversing nerve root area (ATNR), and area of the lumbar inter-laminar space (ALILS) were measured. Puncture efficacy ratio (ALDS/ALILS) and nerve injury risk ratio (ATNR/ALILS) were calculated. Cases were divided into four groups: A (10–20 years), B (21–40 years), C (41–60 years), and D (61–80 years). Statistical analysis was performed using SPSS.Results1) ALDS was similar among segments; 2) ATNR was greatest at L5-S1; 3) ALILS was greatest at L5-S1; 4) Puncture efficacy ratio was highest at L3-4 and lowest at L5-S1; 5) Nerve injury risk was highest at L5-S1. In group D, L5-S1 ALDS was larger than L3-4 and L4-5. ALDS decreased after age 40. Age variations were minimal across parameters.ConclusionThe comprehensive analysis demonstrated L3-4 as the optimal first-choice segment for ages 10–60 years, conferring maximal efficacy and safety. L5-S1 can serve as an alternative option for ages 61–80 years when upper interspaces narrow. This study provides quantitative imaging evidence supporting age-specific, optimized lumbar puncture segment selection.

Advances in Slime Mould Algorithm: A Comprehensive Survey.

The slime mould algorithm (SMA) is a new swarm intelligence algorithm inspired by the oscillatory behavior of slime moulds during foraging. Numerous researchers have widely applied the SMA and its variants in various domains in the field and proved its value by conducting various literatures. In this paper, a comprehensive review of the SMA is introduced, which is based on 130 articles obtained from Google Scholar between 2022 and 2023. In this study, firstly, the SMA theory is described. Secondly, the improved SMA variants are provided and categorized according to the approach used to apply them. Finally, we also discuss the main applications domains of the SMA, such as engineering optimization, energy optimization, machine learning, network, scheduling optimization, and image segmentation. This review presents some research suggestions for researchers interested in this algorithm, such as conducting additional research on multi-objective and discrete SMAs and extending this to neural networks and extreme learning machining.

Accurate multilevel thresholding image segmentation via oppositional Snake Optimization algorithm: Real cases with liver disease

Liver-related diseases significantly contribute to global mortality rates. Accurate segmentation of liver disease from CT scans is essential for early diagnosis and treatment selection, particularly in computer-aided diagnosis (CAD) systems. To address challenges posed by inconsistent liver presence and unclear boundaries, an enhanced Snake Optimization (SO) algorithm is proposed that integrates with opposition-based learning (OBL) called (SO-OBL), proving effective in global optimization and multilevel image segmentation. Experiments using CEC’2022 test functions compare SO-OBL with eleven recent and state-of-the-art metaheuristic algorithms, demonstrating its superior performance. Additionally, an advanced liver disease segmentation model based on SO-OBL incorporates an optimized multilevel thresholding technique, leveraging Otsu’s function. Notable segmentation metric results, including FSIM = 0.947, SSIM = 0.941, PSNR = 24.876, MSE = 236.88, and execution time = 0.281, underscore the model’s efficiency and potential for accurate diagnosis in CAD systems.

Designing Post-Disaster Restoration Schemes for Traffic Networks based on Network Topological Indicators

Identifying key road segments and prioritizing their protection to enhance the resilience of road networks against crisis events remains a pressing issue. In this paper, complex network theory is adopted to determine the importance ranking of road segments and analyse how to select the road segments that need to be rehabilitated in the case of insufficient budget. This study will provide three different methods to assess the importance of road segments, and with the help of user equilibrium model for traffic demand allocation, calculate the system travel time after rehabilitation to evaluate, compare, and analyse different rehabilitation schemes, and then get the rehabilitation scheme with better performance. Meanwhile, it is investigated whether the optimal road segment rehabilitation combination will change under different budget levels. The results of the case study finally show that the rehabilitation set determined by the Clustering Coefficient and the betweenness is better, which can provide practical guidance on how to select the prioritized rehabilitation segments.

Optimal green product segmentation and differential pricing in a platform supply chain

Optimal green product segmentation and differential pricing in a platform supply chain

Exploring the effects of cooperative adaptive cruise control-based transit bus operation on signalized corridors

The world’s transportation system is overburdened by ever-growing travel demand, which brings mobility, safety, and pollution problems. To combat these issues and make better use of existing road capacity on urban arterials, public transit buses are designed to carry more people in fewer vehicles. However, transit buses travel slowly and make frequent intermittent stops, resulting in unreliable travel times and inconvenient riding experiences; this hinders people from choosing to travel on public transit buses. However, a research gap exists in using connected and automated vehicle (CAV) technologies specifically for transit bus operation optimization. To bridge this gap, the present study extended CAV applications to transit bus operation and developed a transit bus control method based on cooperative adaptive cruise control (CACC). The proposed model first optimized bus segment speeds to minimize schedule deviations and fuel consumption. Then, a CACC algorithm was integrated with optimal segment speed. The proposed control method was implemented through micro-simulations of an actual corridor in Jinan, Shandong, China. The evaluation results indicate that the proposed control method reduced the total arrival deviation up to 65.1% and total fuel consumption up to 6.8%. The study adapts CAV technologies to transit bus operation. The findings in this study validate CAV applications in transit bus operation.

RECIST-Induced Reliable Learning: Geometry-Driven Label Propagation for Universal Lesion Segmentation.

Automatic universal lesion segmentation (ULS) from Computed Tomography (CT) images can ease the burden of radiologists and provide a more accurate assessment than the current Response Evaluation Criteria In Solid Tumors (RECIST) guideline measurement. However, this task is underdeveloped due to the absence of large-scale pixel-wise labeled data. This paper presents a weakly-supervised learning framework to utilize the large-scale existing lesion databases in hospital Picture Archiving and Communication Systems (PACS) for ULS. Unlike previous methods to construct pseudo surrogate masks for fully supervised training through shallow interactive segmentation techniques, we propose to unearth the implicit information from RECIST annotations and thus design a unified RECIST-induced reliable learning (RiRL) framework. Particularly, we introduce a novel label generation procedure and an on-the-fly soft label propagation strategy to avoid noisy training and poor generalization problems. The former, named RECIST-induced geometric labeling, uses clinical characteristics of RECIST to preliminarily and reliably propagate the label. With the labeling process, a trimap divides the lesion slices into three regions, including certain foreground, background, and unclear regions, which consequently enables a strong and reliable supervision signal on a wide region. A topological knowledge-driven graph is built to conduct the on-the-fly label propagation for the optimal segmentation boundary to further optimize the segmentation boundary. Experimental results on a public benchmark dataset demonstrate that the proposed method surpasses the SOTA RECIST-based ULS methods by a large margin. Our approach surpasses SOTA approaches over 2.0%, 1.5%, 1.4%, and 1.6% Dice with ResNet101, ResNet50, HRNet, and ResNest50 backbones.