Classification of Pathological Cases of Myocardial Infarction Using Convolutional Neural Network and Random Forest

Attri-VAE: attribute-based, disentangled and interpretable representations of medical images with variational autoencoders

“…The proposed Attri-VAE model also achieved excellent classification performance (healthy vs. myocardial infarction), outperforming the other VAE-based approaches, with slightly better results when trained with radiomics. When evaluated in the EMIDEC training dataset with ground-truth labels, the Attri-VAE approach provided accuracy results (0.98) equivalent to the best challenge participants reporting their performance on the same dataset (1.0 (Lourenc ¸o et al, 2021), 0.95 (Shi et al, 2021), 0.94 (Ivantsits et al, 2021) and 0.90 (Sharma et al, 2021)). For the testing EMIDEC dataset (Lalande et al, 2021), the best participant method obtained a decreased accuracy (0.82, (Lourenc ¸o et al, 2021;Girum et al, 2021)), increasing to 0.92 for the challenge organizers (Shi et al, 2021).…”

“…When evaluated in the EMIDEC training dataset with ground-truth labels, the Attri-VAE approach provided accuracy results (0.98) equivalent to the best challenge participants reporting their performance on the same dataset (1.0 (Lourenc ¸o et al, 2021), 0.95 (Shi et al, 2021), 0.94 (Ivantsits et al, 2021) and 0.90 (Sharma et al, 2021)). For the testing EMIDEC dataset (Lalande et al, 2021), the best participant method obtained a decreased accuracy (0.82, (Lourenc ¸o et al, 2021;Girum et al, 2021)), increasing to 0.92 for the challenge organizers (Shi et al, 2021). As for the ACDC dataset, which was tested as an external database (i.e., without considering it in training), classification accuracy was substantially reduced (0.59), being worst than results reported by challenge participants (Bernard et al, 2018) (0.96) to classify between the different pathologies (not only between healthy and myocardial infarction).…”

Deep Learning methods for automatic evaluation of delayed enhancement-MRI. The results of the EMIDEC challenge

“…Challengers employed a variety of machine learning-based algorithms to interpret the DE-MRI and the clinical features. Provided with the MRI, a simple down-sampling CNN as AlexNet ( [26]) encodes the images to regression or classification outputs ( [40,41,30]), or optionally U-Net based downsampling up-sampling models yield the segmentation of different myocardial tissues so that the volume of each tissue can be quantified ( [30,12]).…”

“…Inputs are passed through multiple layers in which data are mapped with non-linear activation functions in the forward stage ( [22,40]). The decision tree ( [33]) and the random forest ( [18]) are flow-chart-like decision models that consist of nodes ( [40,41,22]). The random forest corrects the overfitting habit of the decision trees by training uncorrelated trees and the final decision is made by individual trees.…”

“…Data fusion is therefore a challenging issue to achieve the maximum semantic information. [30], [12] and [41] Apart from the surface regression methods, the concatenation of the surface information to other textual features was also variable. Lourenço et al added the volumes of all myocardial tissues as four additional textual features.…”