Convolutional neural networks for classification of Alzheimer's disease: Overview and reproducible evaluation

Wen, Junhao; Thibeau–Sutre, Elina; Diaz-Melo, Mauricio; Samper‐Gonzàlez, Jorge; Routier, Alexandre; Bottani, Simona; Dormont, Didier; Durrleman, Stanley; Burgos, Ninon; Colliot, Olivier

doi:10.1016/j.media.2020.101694

Cited by 493 publications

(482 citation statements)

References 136 publications

Supporting

Mentioning

401

Contrasting

Unclassified

Order By: Relevance

“…Still, many papers published in the area are suffering from biased results most probably caused by limited experience with medical data. While working on this paper, we became aware of the recent work by Wen et al[41] that illustrated the presence of data leakage across various studies which use ML in AD classification. They performed a rigorous literature search on AD and grouped the studies into three categorize: (a) studies without data leakage; (b) studies with potential data leakage and (c) studies with clear data leakage.…”

mentioning

confidence: 99%

Generalization Performance of Deep Learning Models in Neurodegenerative Disease Classification

Yağış

García

Citi

2019

2019 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)

View full text Add to dashboard Cite

Over the past decade, machine learning gained considerable attention from the scientific community and has progressed rapidly as a result. Given its ability to detect subtle and complicated patterns, deep learning (DL) has been utilized widely in neuroimaging studies for medical data analysis and automated diagnostics with varying degrees of success. In this paper, we question the remarkable accuracies of the best performing models by assessing generalization performance of the stateof-the-art convolutional neural network (CNN) models on the classification of two most common neurodegenerative diseases, namely Alzheimer's Disease (AD) and Parkinson's Disease (PD) using MRI. We demonstrate the impact of the data division strategy on the model performances by comparing the results derived from two different split approaches. We first evaluated the performance of the CNN models by dividing the dataset at the subject level in which all of the MRI slices of a patient are put into either training or test set. We then observed that pooling together all slices prior to applying cross-validation, as erroneously done in a number of previous studies, leads to inflated accuracies by as much as 26% for the classification of the diseases.

show abstract

mentioning

confidence: 99%

Generalization Performance of Deep Learning Models in Neurodegenerative Disease Classification

Yağış

García

Citi

2019

2019 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)

View full text Add to dashboard Cite

show abstract

“…This is largely due to the much smaller amount of training data in clinical applications, but also to the particularity of medical images which often have poor contrast and low resolution. A recent survey on deep learning for Alzheimer's prediction [60] found that most studies reporting high accuracy suffered from some form of data leakage (e.g., using images from the same subject in both training and testing).…”

Section: Discussionmentioning

confidence: 99%

Deep Radiomic Analysis to Predict Gleason Score in Prostate Cancer

et al. 2020

View full text Add to dashboard Cite

Convolutional neural networks (CNNs) require large amounts of data for training, beyond what can be acquired for current radiomics models. We hypothesize that deep entropy features (DEFs) derived from existing CNNs can be applied to MRI images of prostate cancers (PCa) to reliably predict the Gleason score (GS) of PCa lesions. In this study, we analyzed 112 lesions acquired from 99 PCa patients, either pre-biopsy or pre-treatment, their associated GS, and multi-parametric MRI (mpMRI) sequences. Our approach is based on the extraction of DEF features produced in individual layers of 9 pre-trained CNN models. We first analyze DEFs from separate CNNs using the Wilcoxon test and Spearman correlation to find significant features associated with GS. In a multivariate analysis, we then use the combined DEFs of all CNNs as input to a random forest (RF) classifier for predicting the Gleason grade group of patients. Among the 9 pre-trained CNNs, the NASNet-mobile architecture offered the features most correlated to GS (ρ=0.47; p<0.05). From the 7,857 combined features, 11 DEFs could differentiate GS < 8 from GS ≥ 8 (corrected p < 0.05). Moreover, the RF classifier discerned GS of 6, 3+4, 4+3, 8 and ≥ 9 with an AUC (%) of 80.08, 85.77, 97.30, 98.20, and 86.51, respectively. Our results suggest that the DEFs can be used to differentiate GS of PCa lesions with the highest accuracy of GS ≥ 8 based on mpMRI. DEFs could improve diagnosis accuracy, reduce the risks of misclassification, help to better assess prognosis, and individualize patient care approaches.

show abstract

“…However, we decide to discard this modality because not every subject has information of all modalities and the number of patients with all modalities available is too small for reasonable classification. The second limitation is the lack of methods that separate MCI groups (EMCI and LMCI) with directed graphs in their experiments (see [ 14 , 40 , 45 ]). Moreover, other limitations include the cross-sectional nature of this database and the absence of longitudinal RS-fMRI data.…”

Section: Discussionmentioning

confidence: 99%

“…This methodology offers an attractive approach since it provides useful and effective tools for characterizing network structures together with their intrinsic complexity. There are several approaches that use global and local structures encoded by undirected graphs [ 4 , 5 , 6 , 7 , 8 , 9 , 10 ], and machine learning [ 11 , 12 , 13 , 14 ].…”

Section: Introductionmentioning

confidence: 99%

Early Detection of Alzheimer’s Disease: Detecting Asymmetries with a Return Random Walk Link Predictor

Curado

Escolano

Lozano

et al. 2020

Entropy

View full text Add to dashboard Cite

Alzheimer’s disease has been extensively studied using undirected graphs to represent the correlations of BOLD signals in different anatomical regions through functional magnetic resonance imaging (fMRI). However, there has been relatively little analysis of this kind of data using directed graphs, which potentially offer the potential to capture asymmetries in the interactions between different anatomical brain regions. The detection of these asymmetries is relevant to detect the disease in an early stage. For this reason, in this paper, we analyze data extracted from fMRI images using the net4Lap algorithm to infer a directed graph from the available BOLD signals, and then seek to determine asymmetries between the left and right hemispheres of the brain using a directed version of the Return Random Walk (RRW). Experimental evaluation of this method reveals that it leads to the identification of anatomical brain regions known to be implicated in the early development of Alzheimer’s disease in clinical studies.

show abstract

Convolutional neural networks for classification of Alzheimer's disease: Overview and reproducible evaluation

Cited by 493 publications

References 136 publications

Generalization Performance of Deep Learning Models in Neurodegenerative Disease Classification

Generalization Performance of Deep Learning Models in Neurodegenerative Disease Classification

Deep Radiomic Analysis to Predict Gleason Score in Prostate Cancer

Early Detection of Alzheimer’s Disease: Detecting Asymmetries with a Return Random Walk Link Predictor

Contact Info

Product

Resources

About