Medical Image Segmentation Techniques

STAMP trains and prunes a UNet for medical image segmentation with online pruning, achieving higher performance and reducing parameters by over 85% with fewer labels.

This paper develops an automated framework for analyzing low-field MRI scans to detect biomarkers of severe brain swelling in cerebral malaria.

This paper proposes an automated pipeline for reconstructing fetal thorax MRI from motion-corrupted stacks across 21-36 weeks GA, using 3D CNNs and landmark-guided pose estimation.

This paper evaluates the feasibility of using FreeSurfer for processing T1-weighted brain images of 5-year-olds, showing minor effects of manual edits and recommending quality assessment for reliable segmentation.

This paper introduces ATRA, a symmetric registration algorithm for longitudinal brain MRI scans, achieving precise alignment without interpolation bias or reference volume.

The paper develops a CNN for automated segmentation of fetal brain structures in 3D ultrasound using few-shot learning, achieving performance close to intra-observer variability with minimal manual annotations.

This paper develops a multi-step algorithm for registering perinatal mouse brain images across developmental stages, achieving accurate alignment and synchronization.

This paper introduces a Bayesian 3D U-Net for WMH segmentation that provides uncertainty estimates and is robust to imaging variations, achieving high accuracy and efficiency.

This paper develops deep learning models to automate segmentation of swallowing and chewing structures in H&N CT scans, achieving high accuracy and clinical utility.

The paper compares three WMH segmentation algorithms on MRI data, finding BIANCA and UBO Detector outperform FreeSurfer in accuracy and correlation with Fazekas scores.

This study compares manual and automatic amygdala segmentation methods in children and adolescents, finding systematic volume differences and method-specific growth patterns.

This study applies the anatomical fiducials framework to Parkinson's Disease clinical MRI data, showing it can measure registration errors and detect morphometric biomarkers of the disease.

CEREBRUM-7T is a fast, fully-volumetric brain segmentation method for 7T MRI that outperforms its training labels and is suitable for neuroimaging studies.

This study compares domain adaptation techniques to improve white matter hyperintensity segmentation across different MRI datasets, finding that unsupervised domain adversarial training performs best.

This paper proposes a triplanar ensemble U-Net model for WMH segmentation on MR images, achieving robust performance comparable to existing methods.

This paper develops a deep learning algorithm for 3D segmentation of perivascular spaces in MRI data, achieving high accuracy and demonstrating interoperability across scanners.

This paper introduces a radius-optimized template matching method for brain lesion detection, reducing computational complexity while maintaining performance comparable to existing techniques.

This paper develops an automated pipeline for fetal brain MRI analysis, including extraction, super-resolution, and 4D atlasing, to map fetal brain development in a Chinese population.

This study compares manual and automated quality control strategies for brain MRI segmentation, finding that Euler numbers and MRIQC provide reliable and efficient methods to reduce measurement error.

This paper compares two whole-brain segmentation methods, ASEG and SAMSEG, in FreeSurfer, finding SAMSEG provides more reliable and sensitive measurements for detecting longitudinal changes and Alzheimer's disease-related atrophy.

The paper proposes a machine learning framework for automatically checking the quality of rigid and affine brain MRI registrations using similarity functions as features, achieving high F1-scores on a dataset of 215 autism subjects.

This paper creates an average brain template for male Yucatan minipigs using non-linear registration and evaluates its performance compared to linear methods and another minipig species template.

The paper develops a deep learning segmentation method for analyzing calcified cartilage thickness using micro-CT and compares it to histology across rabbit knee regions.

This paper introduces GP-net, a deep-learning model for automatic segmentation of the globus pallidus interna and externa using 7 Tesla MRI, showing improved results over atlas-based methods.

This paper introduces a robust T1 MRI cortical surface pipeline for neonatal brains and evaluates it across multiple datasets, showing high accuracy and reproducibility.

The paper develops a computational method to construct 3D atlases from 2D data, improving accuracy and reducing artifacts in mouse brain development studies.

The paper introduces CNN-based models for segmenting intracranial and ventricular volumes in neuroimaging data, showing improved accuracy and efficiency compared to existing methods.

The paper introduces a variational method for aligning 1D signals with non-flat displacements, achieving state-of-the-art alignment quality and speed in neuroimaging applications.

This paper develops a deep learning method for segmenting brain vasculature from two-photon microscopy images, achieving faster computation and better generalization across imaging systems.

The paper introduces MU-Net, a convolutional neural network for automated skull-stripping and segmentation in mouse brain MRI, achieving high accuracy and speed.