Segmentation Of Structures Research Articles

DSA-Former: A Network of Hybrid Variable Structures for Liver and Liver Tumour Segmentation.

Accurately annotated CT images of liver tumours can effectively assist doctors in diagnosing and treating liver cancer. However, due to the relatively low density of the liver, its tumours, and surrounding tissues, as well as the existence of multi-scale problems, accurate automatic segmentation still faces challenges. We propose a segmentation network DSA-Former that combines convolutional kernels and attention. By combining the morphological and edge features of liver tumour images, capture global/local features and key inter-layer information, and integrate attention mechanisms obtaining detailed information to improve segmentation accuracy. Compared to other methods, our approach demonstrates significant advantages in evaluation metrics such as the Dice coefficient, IOU, VOE, and HD95. Specifically, we achieve Dice coefficients of 96.8% for liver segmentation and 72.2% for liver tumour segmentation. Our method offers enhanced precision in segmenting liver and liver tumour images, laying a robust foundation for liver cancer diagnosis and treatment.

Multiple paralogues and recombination mechanisms contribute to the high incidence of 22q11.2 Deletion Syndrome.

The 22q11.2 deletion syndrome (22q11.2DS) is the most common microdeletion disorder. Why the incidence of 22q11.2DS is much greater than that of other genomic disorders remains unknown. Short read sequencing cannot resolve the complex segmental duplicon structure to provide direct confirmation of the hypothesis that the rearrangements are caused by nonallelic homologous recombination between the low copy repeats on Chromosome 22 (LCR22s). To enable haplotype-specific assembly and rearrangement mapping in LCR22 regions, we combined fiber-FISH optical mapping with whole genome (ultra-)long read sequencing or rearrangement-specific long-range PCR on 24 duos (22q11.2DS patient and parent-of-origin) comprising several different LCR22-mediated rearrangements. Unexpectedly, we demonstrate that not only different paralogous segmental duplicon but also palindromic AT-rich repeats (PATRR) are driving 22q11.2 rearrangements. In addition, we show the existence of two different inversion polymorphisms preceding rearrangement, and somatic mosaicism. The existence of different recombination sites and mechanisms in paralogues and PATRRs which are copy number expanding in the human population are a likely explanation for the high 22q11.2DS incidence.

Hyperspectral imaging analysis for early detection of tomato bacterial leaf spot disease.

Recent advancements in hyperspectral imaging (HSI) for early disease detection have shown promising results, yet there is a lack of validated high-resolution (spatial and spectral) HSI data representing the responses of plants at different stages of leaf disease progression. To address these gaps, we used bacterial leaf spot (Xanthomonas perforans) of tomato as a model system. Hyperspectral images of tomato leaves, validated against in planta pathogen populations for seven consecutive days, were analyzed to reveal differences between infected and healthy leaves. Machine learning models were trained using leaf-level full spectra data, leaf-level Vegetation index (VI) data, and pixel-level full spectra data at four disease progression stages. The results suggest that HSI can detect disease on tomato leaves at pre-symptomatic stages and differentiate bacterial disease spots from abiotic leaf spots. Using VI data as features for machine learning improved overall classification performance by 26-37% compared to the direct use of raw data. Critical wavelength bands and VIs varied across disease progression stages, suggesting that pre-symptomatic disease detection relied more on changes in leaf water content (1400nm) and plant defense hormone-mediated responses (750nm) rather than changes in leaf pigments or internal structure (800-900nm), which may become more crucial during symptomatic stages. In conclusion, this study provides valuable insights into the dynamics of bacterial spot disease, revealing the potential benefits of leaf structure segmentation and VI group pattern analysis in HSI studies for the early detection of leaf diseases.

Shufflenetv2UNet: An improved neural network model for grassland sample coverage extraction

AbstractAccurate extraction of grassland sample coverage is crucial for regional ecological environment monitoring. Due to the strong feature learning capability, high flexibility, and scalability of deep learning methods, they have great potential in grassland sample extraction modelling. However, we still lack a model that can achieve both lightweight structure and effective performance for small object segmentation to considering the small target characteristics of grassland vegetation and the requirements for model deployment in later stages. Here, we combined the UNet model, which performs well in small target segmentation, with the lightweight network Shufflenetv2 model, proposing an improved UNet neural network, Shufflenetv2UNet, for grassland sample coverage extraction. The core of Shufflenetv2UNet is the removal of maximum pooling and double‐layer convolution modules from downsampling in the UNet neural network. In addition, the Inverted Residual Block structure module from Shufflenetv2 was added to achieve a lightweight model and improved extraction accuracy. The Shufflenetv2UNet achieves an accuracy of 98.23%, with a parameter size of 50.74 M, and a model inference speed of 0.004 s. Compared to existing extraction methods, this model has advantages in prediction accuracy, parameter size, and model inference speed. Moreover, Shufflenetv2UNet achieved different types of grassland sample coverage extractions, with good robustness, generalization, and universality, enabling investigators to quickly and accurately obtain grassland sample coverage. This allows more dynamic and accurate ground measurement data for regional grassland environmental monitoring.

FA-HRNet: A New Fusion Attention Approach for Vegetation Semantic Segmentation and Analysis

Semantic segmentation of vegetation in aerial remote sensing images is a critical aspect of vegetation mapping. Accurate vegetation segmentation effectively informs real-world production and construction activities. However, the presence of species heterogeneity, seasonal variations, and feature disparities within remote sensing images poses significant challenges for vision tasks. Traditional machine learning-based methods often struggle to capture deep-level features for the segmentation. This work proposes a novel deep learning network named FA-HRNet that leverages the fusion of attention mechanism and a multi-branch network structure for vegetation detection and segmentation. Quantitative analysis from multiple datasets reveals that our method outperforms existing approaches, with improvements in MIoU and PA by 2.17% and 4.85%, respectively, compared with the baseline network. Our approach exhibits significant advantages over the other methods regarding cross-region and cross-scale capabilities, providing a reliable vegetation coverage ratio for ecological analysis.

Visualizing the human olfactory projection and ancillary structures in a 3D reconstruction

Visualizing in 3D the histological microanatomy of the human olfactory projection from the olfactory mucosa in the nasal cavity to the olfactory bulbs in the cranial cavity necessitates a workflow for handling a great many sections. Here, we assembled a 3D reconstruction of a 7.45 cm3en-bloc specimen extracted from an embalmed human cadaver. A series of 10 µm coronal sections was stained with quadruple fluorescence histology and scanned in four channels. A trained anatomist manually segmented six structures of interest in a subset of the sections to generate the ground truth. Six convolutional neural networks were then trained for automatic segmentation of these structures in 1234 sections. A high-performance computing solution was engineered to register the sections based on the fluorescence signal and segmented structures. The resulting 3D visualization offers several novel didactic opportunities of interactive exploration and virtual manipulation. By extrapolating manual counts of OSNs in a subset of sections to the calculated volume of the envelope of the entire olfactory epithelium, we computed a total of ~2.7 million OSNs in the specimen. Such empirically derived information helps assess the extent to which the organizational principles of the human olfactory projection may differ from those in mice.

Unsupervised Image Segmentation on 2D Echocardiogram

Echocardiography is a widely used, non-invasive imaging technique for diagnosing and monitoring heart conditions. However, accurate segmentation of cardiac structures, particularly the left ventricle, remains a complex task due to the inherent variability and noise in echocardiographic images. Current supervised models have achieved state-of-the-art results but are highly dependent on large, annotated datasets, which are costly and time-consuming to obtain and depend on the quality of the annotated data. These limitations motivate the need for unsupervised methods that can generalize across different image conditions without relying on annotated data. In this study, we propose an unsupervised approach for segmenting 2D echocardiographic images. By combining customized objective functions with convolutional neural networks (CNNs), our method effectively segments cardiac structures, addressing the challenges posed by low-resolution and gray-scale images. Our approach leverages techniques traditionally used outside of medical imaging, optimizing feature extraction through CNNs in a data-driven manner and with a new and smaller network design. Another key contribution of this work is the introduction of a post-processing algorithm that refines the segmentation to isolate the left ventricle in both diastolic and systolic positions, enabling the calculation of the ejection fraction (EF). This calculation serves as a benchmark for evaluating the performance of our unsupervised method. Our results demonstrate the potential of unsupervised learning to improve echocardiogram analysis by overcoming the limitations of supervised approaches, particularly in settings where labeled data are scarce or unavailable.

Primary Congenital Glaucoma presenting with Acute Corneal Hydrops: a challenging approach to amblyopia prevention

Introduction: Acute hydrops is a rare initial manifestation of Primary Congenital Glaucoma (PCG). It can result in significant corneal opacity, posing an additional risk for amblyopia development in PCG patients. We report a case of a newborn with PCG presenting with acute corneal hydrops and significant corneal scarring, discussing the management and challenges of amblyopia prevention. Case Presentation: A 3-day-old female newborn was referred to our department due to bilateral corneal clouding. Anterior segment examination revealed bilateral corneal edema, with severe corneal prolapse in the right eye consistent with acute hydrops. The anterior chamber structures of the right eye were indiscernible due to significant corneal edema. The left eye presented a deep, quiet anterior chamber with no dysgenesis of anterior segment structures. Intraocular pressure (IOP) measurements were >40 mmHg in both eyes. Corneal diameters were 10 mm OD and 12 mm OS. Fundoscopic posterior segment assessment was not possible due to unclear media. B-scan ultrasonography revealed no significant changes. The patient underwent urgent bilateral trabeculotomy. Postoperatively, intense mechanical stretching induced by severely increased IOP was visible, with the right cornea exhibiting a central opacity managed with topical corticosteroids and showing improvement. Right corneal opacity spared the superior cornea, and atropine drops combined with left eye occlusion were employed to mitigate right eye deprivation amblyopia. Eighteen months later, the patient can fixate and follow objects with the right eye. Left eye corneal transparency was restored, and bilateral IOP control was achieved. Conclusion: Despite prompt surgical treatment, acute hydrops presents a postoperative management challenge and may result in irreversible corneal scarring. Timely management of corneal opacities is crucial to prevent amblyopia. A multidisciplinary approach is essential for the best visual outcome.

Structural studies of ion channels: achievements, problems and perspectives

The superfamily of membrane proteins known as P-loop channels encompasses potassium, sodium, and calcium channels, as well as TRP channels and ionotropic glutamate receptors. An increasing number of crystal and cryo-EM structures are uncovering both general and specific features of these channels. Fundamental folding principles, the arrangement of structural segments, key residues that influence ionic selectivity, gating, and binding sites for toxins and medically relevant ligands have now been firmly established. The advent of AlphaFold2 (AF2) models represents another significant step in computationally predicting protein structures. Comparison of experimental P-loop channel structures with their corresponding AF2 models shows consistent folding patterns in experimentally resolved regions. Despite this remarkable progress, many crucial structural details, particularly important for predicting the outcomes of mutations and designing new medically relevant ligands, remain unresolved. Certain methodological challenges currently hinder the direct assessment of such details. Until the next methodological breakthrough occurs, a promising approach to analyzing ion channel structures in greater depth involves integrating various experimental and theoretical methods.

Exploring Inherent Consistency for Semi-Supervised Anatomical Structure Segmentation in Medical Imaging.

Due to the exorbitant expense of obtaining labeled data in the field of medical image analysis, semi-supervised learning has emerged as a favorable method for the segmentation of anatomical structures. Although semi-supervised learning techniques have shown great potential in this field, existing methods only utilize image-level spatial consistency to impose unsupervised regularization on data in label space. Considering that anatomical structures often possess inherent anatomical properties that have not been focused on in previous works, this study introduces the inherent consistency into semi-supervised anatomical structure segmentation. First, the prediction and the ground-truth are projected into an embedding space to obtain latent representations that encapsulate the inherent anatomical properties of the structures. Then, two inherent consistency constraints are designed to leverage these inherent properties by aligning these latent representations. The proposed method is plug-and-play and can be seamlessly integrated with existing methods, thereby collaborating to improve segmentation performance and enhance the anatomical plausibility of the results. To evaluate the effectiveness of the proposed method, experiments are conducted on three public datasets (ACDC, LA, and Pancreas). Extensive experimental results demonstrate that the proposed method exhibits good generalizability and outperforms several state-of-the-art methods.

Segmental-porosity regulation of nanochannel membranes with anomalously improved ion selectivity for enhanced osmotic power generation

Nanochannel-based osmotic energy conversion can directly convert salinity gradient energy between solutions with different ion concentrations into electric energy. In the energy conversion process, the high-concentration side of the nanochannel membrane has high ion permeability but low ion selectivity, whereas the low-concentration side exhibits opposite properties, which is the inherent limitation of osmotic power. However, the relationship between ion selectivity and permeability is unclear, which hinders the improvement of the output power density. This paper reports a segmental-porosity regulation method to break through the inherent constraints of the ion selectivity–permeability trade-off. The nanochannel membrane was divided into three regions along the ion transport direction as upstream, midstream, and downstream. The porosities of these three regions were separately regulated to optimize the structures for ion selectivity–permeability matching. When a high porosity was adopted for the downstream region and a small and mediate porosity combination was used for the upstream and midstream regions, the highest output power density was achieved, which was 78.44% higher than that of a straight nanochannel with homogeneous porosity under a 100-fold concentration ratio. The outstanding osmotic power generation performance of the optimized segmental structure was attributed to the alleviated ion concentration polarization and unexpectedly improved ion selectivity. Finally, the process and underlying mechanisms of segmental-porosity regulation are also proposed. This paper reports a novel methodology for designing the structure and porosity of nanochannel membranes to enhance osmotic power generation.

Distinct Effects of SARS-CoV-2 Protein Segments on Structural Stability, Amyloidogenic Potential, and α-Synuclein Aggregation.

Amyloidosis is characterized by the abnormal accumulation of misfolded proteins, called amyloid fibrils, leading to diverse clinical manifestations. Recent studies on the amyloidogenesis of SARS-CoV‑2 protein segments have raised concerns on their potential link to post-infection neurodegeneration, however, the mechanisms remain unclear. Herein, we investigated the structure, stability, and amyloidogenic propensity of a nine-residue segment (SK9) of the SARS-CoV-2 envelope protein and their impact on neuronal protein α-synuclein (αSyn) aggregation. Specifically, the amino acid sequence of the SK9 wildtype has been modified from a basic and positively charged peptide (SFYVYSRVK), to a nearly neutral and more hydrophobic peptide (SAAVASAVK, labelled as SK9 var1), and to an acidic and positively charged peptide (SFYVYSRVK, labelled as SK9 var2). Our findings reveal that the SK9 wildtype exhibited a pronounced amyloidogenic propensity due to its disordered and unstable nature, while the SK9 variants possessed more ordered and stable structures preventing the amyloid formation. Significantly, the SK9 wildtype demonstrated distinct effect on αSyn aggregation kinetics and aggregate morphology to facilitate the formation of αSyn aggregates with enhanced resistance against enzymatic degradation. This study highlights the potential of modifying short peptide sequences to fine-tune their properties, providing insights into understanding and regulating viral-induced amyloid aggregations.

Axon Initial Segment Structure and Function in Health and Disease.

At the simplest level, neurons are structured to integrate synaptic input and perform computational transforms on that input, converting it into an action potential (AP) code. This process-converting synaptic input into AP output-typically occurs in a specialized region of axon termed the axon initial segment (AIS). The AIS, as its name implies, is often contained to the first section of axon abutted to the soma and is home to a dizzying array of ion channels, attendant scaffolding proteins, intracellular organelles, extracellular proteins, and, in some cases, synapses. The AIS serves multiple roles as the final arbiter for determining if inputs are sufficient to evoke APs, as a gatekeeper that physically separates the somatodendritic domain from the axon proper, and as a regulator of overall neuronal excitability, dynamically tuning its size to best suit the needs of parent neurons. These complex roles have received considerable attention from experimentalists and theoreticians alike. Here, we review recent advances in our understanding of the AIS and its role in neuronal integration and polarity in health and disease.

Calcium Carbonate as an Ionic Molecular Lock for Ultrastrong Fluorescence of Single Organic Molecules

Locking molecular conformation are widely applied in molecular engineering for improved performance. However, locking via organic functional groups often changes the original molecular properties. Following the rigidity and stability of ionic interaction in ionic compounds, we suggested the use of a molecular‐scale ionic compound, calcium carbonate oligomer, as a robust molecular segment to functionalize organic molecules. The rigid structure of the ionic molecular segments locked the organic molecules, which could remarkably limit the intramolecular motion and intermolecular interactions. This ensured a stable and ultrastrong fluorescence of the single organic molecule while preserving its original maximum emission wavelength. The locking strategy was general and extendable to multiple organic molecules. Additionally, the ultrastrong single‐molecular fluorescence can be maintained in inorganic solids with even higher quantum yields and almost unchanged maximum emission wavelength. The highest quantum yield of the investigated molecules reached 99.9%, superior to all reported organic‒inorganic fluorescent composite under air conditions. This work demonstrates a general strategy to restrict intramolecular motion and intermolecular interactions by using ionic oligomers as molecular locks, providing an alternative method for realizing ultraemissive molecules. This further demonstrates a fascinating example of molecular engineering in the presence of inorganic ionic molecules.

Calcium Carbonate as an Ionic Molecular Lock for Ultrastrong Fluorescence of Single Organic Molecules.

Locking molecular conformation are widely applied in molecular engineering for improved performance. However, locking via organic functional groups often changes the original molecular properties. Following the rigidity and stability of ionic interaction in ionic compounds, we suggested the use of a molecular-scale ionic compound, calcium carbonate oligomer, as a robust molecular segment to functionalize organic molecules. The rigid structure of the ionic molecular segments locked the organic molecules, which could remarkably limit the intramolecular motion and intermolecular interactions. This ensured a stable and ultrastrong fluorescence of the single organic molecule while preserving its original maximum emission wavelength. The locking strategy was general and extendable to multiple organic molecules. Additionally, the ultrastrong single-molecular fluorescence can be maintained in inorganic solids with even higher quantum yields and almost unchanged maximum emission wavelength. The highest quantum yield of the investigated molecules reached 99.9%, superior to all reported organic‒inorganic fluorescent composite under air conditions. This work demonstrates a general strategy to restrict intramolecular motion and intermolecular interactions by using ionic oligomers as molecular locks, providing an alternative method for realizing ultraemissive molecules. This further demonstrates a fascinating example of molecular engineering in the presence of inorganic ionic molecules.

Development of data-efficient training techniques for detection and segmentation models in atrial septum defect analysis

The object of this research is to develop a data-efficient pipeline for the detection of atrial septal defects (ASDs) using echocardiographic images. ASDs are common congenital heart defects that can lead to serious health issues if not diagnosed early. Rising mortality rates due to undetected ASDs highlight the urgent need for improved diagnostic methods. To address the problem of limited annotated medical data hindering accurate detection models, this study fine-tuned the SegFormer model for precise segmentation of cardiac structures in echocardiography images, focusing on the four-chamber heart view essential for ASD detection. By integrating SegFormer with the YOLOv7 detection model, known for real-time object detection, the ASD regions within the segmented heart structures were accurately identified. This cross-referencing ensures anatomically accurate diagnoses and reduces false positives. The study results demonstrate that despite limited data, the integrated method achieves high accuracy and speed, outperforming traditional models. This improvement is explained by the synergy between SegFormer’s transformer-based segmentation and YOLOv7’s efficient detection capabilities. The distinctive feature of our approach is the successful integration of these models in a data-efficient manner, enabling effective ASD detection even with scarce data. The scope of practical use includes deployment in clinical settings with limited resources, requiring only echocardiographic equipment and basic computational resources. By providing clinicians with a reliable tool for ASD detection, the study supports timely interventions in pediatric cardiology, ultimately improving patient outcomes and enhancing care consistency

Multi-task OCTA image segmentation with innovative dimension compression

Optical Coherence Tomography Angiography (OCTA) plays a crucial role in the early detection and continuous monitoring of ocular diseases, which relies on accurate multi-tissue segmentation of retinal images. Existing OCTA segmentation methods typically focus on single-task designs that do not fully utilize the information of volume data in these images. To bridge this gap, our study introduces H2C-Net, a novel network architecture engineered for simultaneous and precise segmentation of various retinal structures, including capillaries, arteries, veins, and the fovea avascular zone (FAZ). At its core, H2C-Net consists of a plug-and-play Height-Channel Module (H2C) and an Enhanced U-shaped Network (GPC-Net). The H2C module cleverly converts the height information of the OCTA volume data into channel information through the Squeeze operation, realizes the lossless dimensionality reduction from 3D to 2D, and provides the "Soft layering" information by unidirectional pooling. Meanwhile, in order to guide the network to focus on channels for training, U-Net is enhanced with group normalization, channel attention mechanism, and Parametric Rectified Linear Unit (PReLU), which reduces the dependence on batch size and enhances the network's ability to extract salient features. Extensive experiments on two subsets of the publicly available OCTA-500 dataset have shown that H2C-Net outperforms existing state-of-the-art methods. It achieves average Intersection over Union (IoU) scores of 82.84 % and 88.48 %, marking improvements of 0.81 % and 1.59 %, respectively. Similarly, the average Dice scores are elevated to 90.40 % and 93.76 %, exceeding previous benchmarks by 0.42 % and 0.94 %. The proposed H2C-Net exhibits excellent performance in OCTA image segmentation, providing an efficient and accurate multi-task segmentation solution in ophthalmic diagnostics. The code is publicly available at: https://github.com/IAAI-SIT/H2C-Net.

A road map to manual segmentation of cerebral structures.

Manual segmentation is an essential tool in the researcher's technical arsenal. It is a frequent practice necessary for image analysis in many protocols, especially in neuroimaging and comparative brain anatomy. In the framework of emergence of studies focusing on alternative animal models, manual segmentation procedures play a critical role. Nevertheless, this critical task is often assigned to students, a process that, unfortunately, tends to be time-consuming and repetitive. Well-conducted and well-described segmentation procedures can potentially guide novice and even expert operators and enhance research works' internal and external validity, making it possible to harmonize studies and facilitate data sharing. Furthermore, recent advances in neuroimaging, such as exvivo imaging or ultra-high-field MRI, enable new acquisition modalities and the identification of minute structures that are barely visible with typical approaches. In this context of increasingly detailed and multimodal brain studies, reflecting on methodology is relevant and necessary. Because it is crucial to implement good practices in manual segmentation per se but also in the description of the segmentation procedures in research papers, we propose a general roadmap for optimizing the technique, its process and the reporting of manual segmentation. For each of them, the relevant elements of the literature have been collected and cited. The article is accompanied by a checklist that the reader can use to verify that the critical steps are being followed.

Detection and interpretation of central type structures within the territory of southeastern Transbaikalia for prediction of ore-forming systems

Extremely little attention is paid to the issues of detecting and interpreting of central type structures (CTS) when conducting remote structural-geological and structural-geomorphological studies. At the same time, in the 70–80s of the 20th century, the important role of CTSs in the localization of deposits and ore fields was proven. The position of these structures must necessarily be taken into account when solving problems of searching for and predicting mineral resources in the context of metallogenic analysis and reconstruction of the geological history of development of the studied areas. The almost absence of results of mass detecting and interpreting of CTSs can be explained by the still poorly developed methodology for identifying and analyzing this type of structure. In the present study for the territory of southeastern Transbaikalia, based on modern geoinformation technologies, the use of remote sensing data (radar topographic survey) of high resolution, the creation of a digital elevation model and the application of an integrated structural-spatial analysis, an author’s approach to detection and interpretation of the CTSs is presented, including in connection with the localization of ore objects of various geological-industrial (geological-genetic) types within the framework of the concept of the formation of mineral systems. A statistical analysis of the CTSs identified in the area was carried out, which made it possible to establish a smooth increase in the number of structures with a decrease in their diameter. It is shown that the spatial maxima of ore mineralization extent within the territory are concentrated on the periphery of large CTSs and in their immediate vicinity. Most of the known large ore objects are confined to the internal areas of structures less than 10 km in diameter. Based on the approach of constructing model sections, it was possible to reconstruct the deep position of magma chambers associated with the identified CTSs, and, thereby, to determine the probable sources of metal-bearing fluids. A close spatial relationship between the identified magma chambers and deep faults has been established. To determine the most favorable sites for the deposition of ore mineralization, based on structural-spatial criteria, which include not only structural elements of the CTSs, but also segments of known fault structures, weight of evidence models of the territory have been created. The accuracy of the complex model is 89%. Thus, in accordance with the concept of mineral systems, the sources, migration pathways and sites of the most probable deposition of ore mineralization have been reconstructed.

Necessity and impact of specialization of large foundation model for medical segmentation tasks.

Large foundation models, such as the Segment Anything Model (SAM), have shown remarkable performance in image segmentation tasks. However, the optimal approach to achieve true utility of these models for domain-specific applications, such as medical image segmentation, remains an open question. Recent studies have released a medical version of the foundation model MedSAM by training on vast medical data, who promised SOTA medical segmentation. Independent community inspection and dissection is needed. Foundation models are developed for general purposes. On the other hand, stable delivery of reliable performance is key to clinical utility. This study aims at elucidating the potential advantage and limitations of landing the foundation models in clinical use by assessing the performance of off-the-shelf medical foundation model MedSAM for the segmentation of anatomical structures in pelvic MR images. We also explore the simple remedies by evaluating the dependency on prompting scheme. Finally, we demonstrate the need and performance gain of further specialized fine-tuning. MedSAM and its lightweight version LiteMedSAM were evaluated out-of-the-box on a public MR dataset consisting of 589 pelvic images split 80:20 for training and testing. An nnU-Net model was trained from scratch to serve as a benchmark and to provide bounding box prompts for MedSAM. MedSAM was evaluated using different quality bounding boxes, those derived from ground truth labels, those derived from nnU-Net, and those derived from the former two but with 5-pixel isometric expansion. Lastly, LiteMedSAM was refined on the training set and reevaluated on this task. Out-of-the-box MedSAM and LiteMedSAM both performed poorly across the structure set, especially for disjoint or non-convex structures. Varying prompt with different bounding box inputs had minimal effect. For example, the mean Dice score and mean Hausdorff distances (in mm) for obturator internus using MedSAM and LiteMedSAM were {0.251±0.110, 0.101±0.079} and {34.142±5.196, 33.688±5.306}, respectively. Fine-tuning of LiteMedSAM led to significant performance gain, improving Dice score and Hausdorff distance for the obturator internus to 0.864±0.123 and 5.022±10.684, on par with nnU-Net with no significant difference in evaluation of most structures. All segmentation structures benefited significantly from specialized refinement, at varying improvement margin. While our study alludes to the potential of deep learning models like MedSAM and LiteMedSAM for medical segmentation, it highlights the need for specialized refinement and adjudication. Off-the-shelf use of such large foundation models is highly likely to be suboptimal, and specialized fine-tuning is often necessary to achieve clinical desired accuracy and stability.