Automated Dentate Nucleus Segmentation from QSM Images Using Deep Learning

Diogo H Shiraishi,Susmita Saha,Isaac M Adanyeguh,Sirio Cocozza,Louise A Corben,Andreas Deistung,Martin B Delatycki,Imis Dogan,William Gaetz,Nellie Georgiou-Karistianis,Simon Graf,Marina Grisoli,Pierre-Gilles Henry,Gustavo M Jarola,James M Joers,Christian Langkammer,Christophe Lenglet,Jiakun Li,Camila C Lobo,Eric F Lock,David R Lynch,Thomas M Mareci,Alberto R M Martinez,Serena Monti,Anna Nigri,Massimo Pandolfo,Kathrin Reetz,Timothy P Roberts,Sandro Romanzetti,David A Rudko,Alessandra Scaravilli,Jörg B Schulz,S H Subramony,Dagmar Timmann,Marcondes C França,Ian H Harding,Thiago J R Rezende,TRACK-FA Neuroimaging Consortium

DOI: https://doi.org/10.1101/2024.06.17.24308662

2024-06-28

Abstract:To develop a dentate nucleus (DN) segmentation tool using deep learning (DL) applied to brain quantitative susceptibility mapping (QSM) images. Brain QSM images from 132 healthy controls and 170 individuals with cerebellar ataxia or multiple sclerosis were collected from nine different datasets worldwide for this retrospective study. Manual delineation of the DN (gray matter and white matter hilus) was first undertaken by experienced raters with a robust quality control process. Performance of automated segmentation was compared following training using several DL architectures. A two-step approach was implemented, composed of a localization model followed by DN segmentation. The manual tracing protocol produced ground-truth data with high intra-rater (average ICC 0.906) and inter-rater reliability (average ICC 0.776). Initial DL architecture exploration indicated that the nnU-Net framework performed best. The two-step localization plus segmentation pipeline achieved a Dice score of 0.898±0.031 and 0.894±0.036 for left and right DN, respectively. In external validation, our algorithm outperformed the leading existing automated tool (left/right DN Dice 0.863±0.038/0.843±0.066 vs. 0.568±0.222/0.582±0.239). The model demonstrated generalizability across unseen datasets during the training step. The measures showed a superior correlation index with manual annotations and performed well in both isotropic and anisotropic QSM datasets. We provide a model that accurately and efficiently segments the DN from brain QSM images. The model can be readily deployed for use in observational, natural history, and treatment trials for biomarker discovery.

Radiology and Imaging

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to develop a deep - learning - based tool for automatically segmenting the Dentate Nucleus (DN) from brain Quantitative Susceptibility Mapping (QSM) images. Specifically, the research aims to overcome the limitations of currently relying on manual drawing for segmentation. By using a multi - center data set from healthy individuals and patients with cerebellar ataxia or multiple sclerosis, an optimized and general - purpose deep - learning analysis tool is developed to achieve efficient and accurate segmentation of the Dentate Nucleus. This tool can be conveniently applied in observational studies, natural history studies and treatment trials as part of biomarker discovery. The key points of the paper include: - **Problem background**: The Dentate Nucleus is the largest among the deep cerebellar nuclei and is mainly responsible for the main efferent functions of the cerebellum. It plays an important role in a variety of neurological diseases, such as hereditary cerebellar ataxia. However, due to the tissue characteristics of these nuclei, it is difficult to clearly define their structures using standard MRI sequences (such as T1 - weighted and T2 - weighted images). Although SWI (Susceptibility - Weighted Imaging) can partially solve this problem, its non - quantitative nature and tissue boundary distortion limit its application. The development of QSM (Quantitative Susceptibility Mapping) technology overcomes these limitations, providing more accurate anatomical structure mapping and a more direct connection with iron concentration. - **Solution**: Through deep - learning methods, especially using architectures such as 3D U - Net, Swin UNETR and nnU - Net, combined with data augmentation techniques, the generalization ability and accuracy of the model are improved. The research team adopted a two - step method. First, the Dentate Nucleus is located, and then segmentation is carried out, finally achieving a high - precision automatic segmentation effect. - **Objective**: To provide a model that can accurately and efficiently segment the Dentate Nucleus from QSM images, so as to be used as a biomarker detection tool in clinical studies, especially in observational studies, natural history studies and treatment trials. By solving this problem, the research not only promotes the development of neuroimaging technology, but also provides a new tool for understanding cerebellar function and its role in various diseases.

Automated Dentate Nucleus Segmentation from QSM Images Using Deep Learning

Automated 3D Fetal Brain Segmentation Using an Optimized Deep Learning Approach

Automated Segmentation of Brain Gray Matter Nuclei on Quantitative Susceptibility Mapping Using Deep Convolutional Neural Network

Automated deep learning segmentation of high-resolution 7 Tesla postmortem MRI for quantitative analysis of structure-pathology correlations in neurodegenerative diseases

Automated deep learning segmentation of high-resolution 7 T postmortem MRI for quantitative analysis of structure-pathology correlations in neurodegenerative diseases

MoDL-QSM: Model-based Deep Learning for Quantitative Susceptibility Mapping

Accurate, rapid and reliable, fully automated MRI brainstem segmentation for application in multiple sclerosis and neurodegenerative diseases

The Effect of Deep Learning-Based QSM Magnetic Resonance Imaging on the Subthalamic Nucleus

MoG-QSM: Model-based Generative Adversarial Deep Learning Network for Quantitative Susceptibility Mapping.

msQSM: Morphology-based self-supervised deep learning for quantitative susceptibility mapping

Deep Cerebellar Nuclei Segmentation via Semi-Supervised Deep Context-Aware Learning from 7T Diffusion MRI

Brain gray matter nuclei segmentation on quantitative susceptibility mapping using dual-branch convolutional neural network

Fully automatic segmentation of brain lacunas resulting from resective surgery using a 3D deep learning model

Quantitative susceptibility mapping based basal ganglia segmentation via AGSeg: leveraging active gradient guiding mechanism in deep learning

Automatic Segmentation and Quantitative Assessment of Stroke Lesions on MR Images

DARTS: DenseUnet-based Automatic Rapid Tool for brain Segmentation

Deep Learning with Anatomical Attention Mechanism for Distinguishing Parkinson’s Disease from Normal Controls in MR Imaging

IR2QSM: Quantitative Susceptibility Mapping via Deep Neural Networks with Iterative Reverse Concatenations and Recurrent Modules

xQSM: Quantitative Susceptibility Mapping with Octave Convolutional and Noise Regularized Neural Networks

DeepBrainSeg: Automated Brain Region Segmentation for Micro-Optical Images with a Convolutional Neural Network

Deep grey matter quantitative susceptibility mapping from small spatial coverages using deep learning