Deep learning‐based cardiac cine segmentation: Transfer learning application to 7T ultrahigh‐field MRI

Abstract

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Machine Learning in Cardiovascular Imaging: A Scoping Review of Published Literature

“…Pretraining DL models using public medical image datasets is a more effective technique. Ankenbrand et al illustrate this strategy by transferring weights from a DL model pretrained on a public medical image dataset to train their DL model for segmenting the heart on cardiac MRI data [ 41 ].…”

A data‐driven semantic segmentation model for direct cardiac functional analysis based on undersampled radial MR cine series

“…The herein used data were accessed on July 29, 2019. 24 As the Kaggle data are subject to inconsistencies, we performed a data curation as described by Ankenbrand et al 22 Detailed information and examples regarding data curation can be found in this online repository: https://github.com/chfc-cmi/ cmr-seg-tl#data-curat ion-and-conve rsion. The number of examinations in the repository within the 15% confidence set is higher than in this article.…”

“…The Kaggle data are a compilation of real, clinical data from multiple sites and as such, subject to inconsistencies within individual examinations. A data curation was performed as described by Ankenbrand et al, 22 to remove these inconsistencies as much as possible. More detailed information and examples regarding data curation can be found in this online repository: https://github.com/chfccmi/cmr-seg-tl#data-curat ion-and-conve rsion.…”

“…In this article, we describe a 3D artificial neural network that was trained for semantic segmentation of the LV, the RV, and the myocardium with radially undersampled cine data as input. Radial acquisitions were first simulated based on a larger data set holding Cartesian cine images 22 and used for a pretraining of the model (see Figure 1). For translation into a realistic setting, the network was fine‐tuned using a smaller data set with genuine radial cine MRI data 23 and ultimately applied to unseen native radial cine MRI data with different levels of undersampling.…”

Deep convolution neural network for screening carotid calcification in dental panoramic radiographs

“…CNN and TL have been widely used in the prediction of medical conditions using different techniques (CT, MRI, panoramic images) -for example: identification of prostate cancer [12]; prediction of bladder cancer treatment response in CT [13]; detection of maxillary sinusitis on panoramic radiographs [14]; screening for osteoporosis in dental panoramic radiographs [15]; cardiac cine segmentation [16] and even COVID-19 detection from chest CT-scans [17].…”