Medical Image Segmentation Techniques

This study evaluates image registration accuracy in portable low-field 3D brain MRI, finding that synthesis-based methods perform best despite acquisition artifacts.

This paper presents a Hybrid-Fusion Transformer model that refines myocardium segmentation in cardiac CT by combining CT images with semi-automatic labels to remove papillary muscles.

This paper evaluates Graph Neural Networks for brain cortex segmentation using structural MRI and diffusion MRI data, finding GAT models perform competitively with non-graph methods.

This paper develops BrainMLSR, an automated framework to reconstruct three intracortical signal-defined layers from 5T 3D FLAIR MRI, showing high agreement with manual annotations and revealing regional morphometric differences in epilepsy.

The paper develops a harmonized segmentation protocol for hippocampal subfields in MRI, improving cross-study comparisons and understanding of subfield structure and function.

The paper develops a multiscale segmentation pipeline using HiP-CT and deep learning to identify glomeruli in human kidneys across different resolutions.

FibroTrack automates fibrosis quantification in muscle and cardiac histology using deep learning, achieving high precision and agreement with pathologists.

This paper introduces a 3D U-Net with an anatomy-aware loss function for automated segmentation of human vagus nerve fascicles and epineurium from microCT images, achieving high accuracy and reducing anatomical errors.

The paper introduces a fast segmentation tool for brain MRI scans using a hybrid method, achieving comparable accuracy to a Bayesian approach but with significantly reduced runtime.

SUITPy improves cerebellar imaging analysis by isolating the cerebellum with a U-Net model, enhancing alignment and reducing contamination from surrounding cortical structures.

The paper develops a deep learning model to automate electrode detection in tDCS-fMRI studies, outperforming manual and semi-automated methods in precision and reducing human bias.

This study uses shape analysis to detect and prioritize errors in deep learning-based glomerular segmentation, enabling more efficient correction by pathologists.

The paper introduces the MTL200, a surface-based probabilistic atlas of the medial temporal lobe regions, derived from anatomical MRI data of 200 participants.

This paper introduces a deep learning model for automated joint space detection in micro-CT images of mouse paws, significantly improving segmentation accuracy compared to semi-automated methods.

The paper characterizes hippocampal subfield border variability using postmortem MRI and histology, finding consistent subfield appearance/disappearance orders and no clear diagnosis-related differences.

The paper presents a harmonized segmentation protocol for the hippocampal tail in high-resolution MRI, improving cross-study comparability through standardized definitions and consensus-based guidelines.

MORPH2DIAG automates MRI preprocessing and segmentation, using machine and deep learning to classify neuroanatomical subtypes with high accuracy.

This paper develops a deep learning framework for automated segmentation of 3D CT data and validates its biomechanical accuracy using zebrafish mandibles.

This paper develops a multi-task deep learning method to segment and parcellate fetal brain structures from diffusion MRI, achieving high accuracy in tissue segmentation, tract segmentation, and parcellation.

This paper develops a multi-modality MRI segmentation model to improve robustness of medial temporal lobe subregion segmentation in poor-quality images.

PELICAN is a longitudinal MRI processing pipeline that improves accuracy and reliability in analyzing brain atrophy and abnormalities in aging and neurodegenerative disease populations.

This paper introduces a 2D segmentation framework for neonatal HIE lesions using DWI data, achieving competitive performance against 3D methods with a Dice Similarity Coefficient of 0.6060.

This paper develops an automated method for segmenting the locus coeruleus in MRI scans and examines age-related signal changes across different imaging modalities.

The paper uses X-ray micro-tomography and deep-learning segmentation to estimate brain region volumes in wild lizards, showing that a deep-learning model can be trained with few datasets.

The paper develops a segmentation protocol for anterior medial temporal lobe regions and integrates it into an automated approach to improve imaging biomarkers for neurodegenerative diseases.

The paper develops Macaca-Star, a pipeline for automated macaque brain registration across modalities, addressing deformation and variability challenges in ex vivo samples.

This paper presents an automatic system to evaluate semantic segmentation quality in laryngeal endoscopy images, showing its performance matches human raters using intersection over union metric.

SurfNet jointly reconstructs cortical surfaces from MRIs using diffeomorphic deformations, achieving state-of-the-art accuracy and topological correctness.

This paper introduces a user-assisted nonrigid registration method that improves accuracy by incorporating fiducial points, achieving sufficient biomedical application accuracy within minutes.

This study compares ATRA to FSL, FreeSurfer, and ANTS for longitudinal brain MRI alignment, finding ATRA more accurate in most cases.