An Efficient 1DCNN–LSTM Deep Learning Model for Assessment and Classification of fMRI-Based Autism Spectrum Disorder

Qayyum, Abdul; Khan, Mohamed K.; Benzinou, Abdesslam; Mazher, Moona; Ramasamy, Manickam; Aramugam, Kalaiselvi; Deisy, C.; Sridevi, S.; Suresh, M.

doi:10.1007/978-981-16-7167-8_77

“…1) LSTM-based methods: This line of methods utilizes LSTM network to capture the temporal dynamics in rs-fMRI, two typical examples are included for comparison: i) cGC-NLSTM [35], which constructs FBNs by Pearson correlation matrices and feeds them into GCN and LSTM for feature extraction. ii) 1DCNN-LSTM [6], which develops a hybrid model of CNN and LSTM for temporal information extraction from fMRI time series.…”

Section: B Methods For Comparisonmentioning

confidence: 99%

“…The resulting sequence of subFBNs from the same scan is expected to reveal more temporal dynamics of the brain functioning from an FBN perspective across windows and they can be jointly analyzed, e.g., via long short-term memory (LSTM) for disease identification. Besides, LSTM can also be directly applied to the raw or transformed rs-fMRI time series for temporal information extraction, e.g., in [6]; however, it does not attempt to construct FBN, which can be used for interpretation purpose, as in this paper.…”

mentioning

confidence: 99%

Asynchronous Functional Brain Network Construction with Spatiotemporal Transformer for MCI Classification

Tang

2023

Preprint

0

View full text Add to dashboard Cite

<p>Construction and analysis of functional brain networks (FBNs) with rs-fMRI is a promising method to diagnose functional brain diseases. Nevertheless, the existing methods suffer from several limitations. First, the functional connectivities (FCs) of the FBN are usually measured by the temporal co-activation level between rs-fMRI time series from regions of interest (ROIs). While enjoying simplicity, the existing approach implicitly assumes simultaneous co-activation of all the ROIs, and models only their synchronous dependencies. However, the functional coactivation is not necessarily always synchronous due to the time lag of information flow and cross-time interactions between ROIs. Therefore, it is desirable to model the asynchronous functional interactions. Second, the traditional methods usually construct FBNs at individual level for feature extraction and classification, leading to large variability and degraded diagnosis accuracy when modeling asynchronous FBN. Third, the FBN construction and analysis are conducted in two independent steps without joint alignment for the target diagnosis task. To address the first limitation, this paper proposes an effective sliding-window-based method to model spatiotemporal FCs in Transformer. Regarding the second limitation, we propose to learn common and individual FBNs adaptively with the common FBN as prior knowledge, thus alleviating the variability and enabling the network to focus on the individual disease-specific asynchronous FCs. To address the third limitation, the common and individual asynchronous FBNs are built and analyzed by an integrated network, enabling end-to-end training and improving the flexibility and discriminativity. The effectiveness of the proposed method is consistently demonstrated on three data sets for MCI diagnosis.</p>

show abstract

“…1) LSTM-based methods: This line of methods utilizes LSTM network to capture the temporal dynamics in rs-fMRI, two typical examples are included for comparison: i) cGC-NLSTM [35], which constructs FBNs by Pearson correlation matrices and feeds them into GCN and LSTM for feature extraction. ii) 1DCNN-LSTM [6], which develops a hybrid model of CNN and LSTM for temporal information extraction from fMRI time series.…”

Section: B Methods For Comparisonmentioning

confidence: 99%

“…The resulting sequence of subFBNs from the same scan is expected to reveal more temporal dynamics of the brain functioning from an FBN perspective across windows and they can be jointly analyzed, e.g., via long short-term memory (LSTM) for disease identification. Besides, LSTM can also be directly applied to the raw or transformed rs-fMRI time series for temporal information extraction, e.g., in [6]; however, it does not attempt to construct FBN, which can be used for interpretation purpose, as in this paper.…”

mentioning

confidence: 99%

Asynchronous Functional Brain Network Construction with Spatiotemporal Transformer for MCI Classification

Tang

2023

Preprint

0

View full text Add to dashboard Cite

<p>Construction and analysis of functional brain networks (FBNs) with rs-fMRI is a promising method to diagnose functional brain diseases. Nevertheless, the existing methods suffer from several limitations. First, the functional connectivities (FCs) of the FBN are usually measured by the temporal co-activation level between rs-fMRI time series from regions of interest (ROIs). While enjoying simplicity, the existing approach implicitly assumes simultaneous co-activation of all the ROIs, and models only their synchronous dependencies. However, the functional coactivation is not necessarily always synchronous due to the time lag of information flow and cross-time interactions between ROIs. Therefore, it is desirable to model the asynchronous functional interactions. Second, the traditional methods usually construct FBNs at individual level for feature extraction and classification, leading to large variability and degraded diagnosis accuracy when modeling asynchronous FBN. Third, the FBN construction and analysis are conducted in two independent steps without joint alignment for the target diagnosis task. To address the first limitation, this paper proposes an effective sliding-window-based method to model spatiotemporal FCs in Transformer. Regarding the second limitation, we propose to learn common and individual FBNs adaptively with the common FBN as prior knowledge, thus alleviating the variability and enabling the network to focus on the individual disease-specific asynchronous FCs. To address the third limitation, the common and individual asynchronous FBNs are built and analyzed by an integrated network, enabling end-to-end training and improving the flexibility and discriminativity. The effectiveness of the proposed method is consistently demonstrated on three data sets for MCI diagnosis.</p>

show abstract

An Exploration of Autism Spectrum Disorder Classification from Structural and Functional MRI Images

Krajevski¹,

Ivanoska²,

Trivodaliev³

et al. 2022

Communications in Computer and Information Science

0

View full text Add to dashboard Cite

An Efficient 1DCNN–LSTM Deep Learning Model for Assessment and Classification of fMRI-Based Autism Spectrum Disorder

Cited by 7 publications

References 18 publications

Asynchronous Functional Brain Network Construction with Spatiotemporal Transformer for MCI Classification

Asynchronous Functional Brain Network Construction with Spatiotemporal Transformer for MCI Classification

Asynchronous Functional Brain Network Construction with Spatiotemporal Transformer for MCI Classification

An Exploration of Autism Spectrum Disorder Classification from Structural and Functional MRI Images

Contact Info

Product

Resources

About