Functional connectome fingerprinting using shallow feedforward neural networks

Sarar, Gokce; Rao, Bhaskar D.; Liu, Thomas T.

doi:10.1073/pnas.2021852118

Cited by 12 publications

(8 citation statements)

References 8 publications

(13 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Functional connectome fingerprinting has been an active area of research (Abbas et al, 2020; Amico & Goñi, 2018; Finn et al, 2015; Griffa et al, 2022), but these studies have primarily involved modest numbers of individuals from single high-quality datasets. Neural network applications in this area are also not new (Misra et al, 2021; Sarar et al, 2021; Wang et al, 2019), but again, these studies have not looked at repurposing the intermediate representation of these models for new data. We achieve near ceiling accuracy on multiple datasets, even when scan duration is severely limited.…”

Section: Discussionmentioning

confidence: 99%

“…These are derived using a given dataset’s subject identifiers and not anyone’s real-world identities. Functional connectome fingerprinting in particular has a long history as a research topic and has proven to be a powerful representation of brain function that quantifies a person’s brain activity by correlating brain region responses over time (Abbas et al, 2020; Amico & Goñi, 2018; Griffa et al, 2022; Sarar et al, 2021; Van De Ville et al, 2021; Wang et al, 2019). Functional connectome fingerprinting has achieved impressive accuracy for recognizing individuals across tasks (Finn et al, 2015) and can be used to predict cognitive performance (Rosenberg et al, 2016; 2020) and traits of individuals (Greene et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Large-Scale Functional Connectome Fingerprinting for Generalization and Transfer Learning in Neuroimaging

Ogg,

Kitchell

2024

Preprint

View full text Add to dashboard Cite

Functional MRI currently supports a limited application space stemming from modest dataset sizes, large interindividual variability and heterogeneity among scanning protocols. These constraints have made it difficult for fMRI researchers to take advantage of modern deep-learning tools that have revolutionized other fields such as NLP, speech transcription, and image recognition. To address these issues, we scaled up functional connectome fingerprinting as a neural network pre-training task, drawing inspiration from speaker recognition research, to learn a generalizable representation of brain function. This approach sets a new high-water mark for neural fingerprinting on a previously unseen scale, across many popular public fMRI datasets (individual recognition over held out scan sessions: 94% on MPI-Leipzig, 94% on NKI-Rockland, 73% on OASIS-3, and 99% on HCP). We show that this representation can also generalize to support accurate neural fingerprinting for completely new datasets and participants not used in training. Finally, we demonstrate that the representation learned by the network encodes features related to individual variability that supports some transfer learning to new tasks. These results open the door for a new generation of clinical applications based on functional imaging data.SIGNIFICANCE STATEMENTDeep learning models that leverage the increasing scale of available fMRI data could address fundamental generalization roadblocks. We drew inspiration from other domains that have successfully used AI to address these problems, namely human language technology, to guide our exploration of the potential for this approach in neuroimaging. Our pre-training approach sets a new high-watermark for functional connectome fingerprinting, achieving very high recognition accuracy across different tasks, scanning sessions, and acquisition parameters, even when the duration of a scan is limited to less than two minutes. We showed that we could re-purpose the representation learned by our model to recognize new individuals from new datasets and to predict new participants’ cognitive performance and traits.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Large-Scale Functional Connectome Fingerprinting for Generalization and Transfer Learning in Neuroimaging

Ogg,

Kitchell

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…The Fingerprint Property of Children's FNC Brain FC and its network analog, FNC, are believed to provide a window into brain function and intrinsic brain organization 2,37−39 . Neuroimaging studies have successfully established that adults' FC pro le shows substantial inter-subject variability, and such variability can distinguish individuals from another scan 7,12 . Unlike adults, children might show more intra-subject variability in FC due to the developments in the brain 40 .…”

Section: Discussionmentioning

confidence: 99%

“…FC heterogeneity has long been appreciated in fMRI studies, even within the same population [8][9][10] . By using a multi-condition fMRI dataset from the Human Connectome Project, studies have shown that FC pro le can distinguish adult subjects across scan sessions and even between distinct task conditions 7,11,12 , acting as a " ngerprint". A recent study has captured ve different functional clusters in the pulvinar with speci c connectivity ngerprints, which are associated with distinct components of cognition 13 .…”

Section: Introductionmentioning

confidence: 99%

Functional Connectivity Uniqueness and Stability? A Signature of Cognitive and Psychiatric Problems in Children

Liu

Salman

et al. 2022

Preprint

View full text Add to dashboard Cite

Brain functional connectivity (FC) derived from functional magnetic resonance imaging (fMRI) has been serving as a potential “fingerprint” for adults. However, intra-subject variation of FC can be substantial and carry biologically meaningful information, especially during adolescence, a time of intense brain changes. Here, for the first time, we performed a large-scale analysis on cross-scan FC stability and its association with a diverse range of health measures in children from the Adolescent Brain Cognitive Development (ABCD) data. Functional network connectivity (FNC), a network analog of FC, was extracted via an automated independent component analysis framework on 9071 subjects (age: 9 ~ 11 years) and compared across 4 scans. We found that the FNC profile can identify a given child from a large group with high accuracy (> 80%), though intra-subject variations exist cross-scan. The robustness of this finding is verified by replicating the child identification using the second-year scans and between longitudinal scans with a two-year interval. Moreover, cross-scan FNC stability was predictive of sleep condition, cognitive performance, and psychiatric problems in children, with higher stability correlated with better cognitive performance, longer sleep duration, and less psychotic expression. Meanwhile, parental psychopathology and prenatal exposure were associated with the FNC stability in their children. Overall, our findings show that a child’s connectivity profile is not only intrinsic but also exhibits reliable variability cross-scan, regardless of brain growth and development. The intra-subject connectivity stability may serve as a valuable biomarker to draw inferences on early cognitive and psychiatric behaviors in children.

show abstract

“…Because resting‐state FC is thought to characterize intrinsic functions (Biswal et al, 1995 ), it has been frequently employed in various studies. For example, previous reports have identified individuals using their FC information obtained from rfMRI (Cai et al, 2020 ; Chen & Hu, 2018 ; Demeter et al, 2020 ; Finn et al, 2015 ; Finn et al, 2017 ; Horien et al, 2019 ; Pallarés et al, 2018 ; Sarar et al, 2021 ). In this context, there have been attempts to uncover the distinguishing interindividual differences of FCs.…”

Section: Introductionmentioning

confidence: 99%

Discovering individual fingerprints in resting‐state functional connectivity using deep neural networks

Lee,

Lee

2023

Human Brain Mapping

View full text Add to dashboard Cite

Non‐negligible idiosyncrasy due to interindividual differences is an ongoing issue in resting‐state functional MRI (rfMRI) analysis. We show that a deep neural network (DNN) can be employed for individual identification by learning important features from the time‐varying functional connectivity (FC) of rfMRI in the Human Connectome Project. We employed the trained DNN to identify individuals from an independent dataset acquired at our institution. The results revealed that the DNN could successfully identify 300 individuals with an error rate of 2.9% using 15 s time‐window and 870 individuals with an error rate of 6.7%. A trained DNN with nonlinear hidden layers led to the proposal of the “fingerprint of FC” (fpFC) as representative edges of individual FC. The fpFCs for individuals exhibited commonly important and individual‐specific edges across time‐window lengths (from 5 min to 15 s). Furthermore, the utility of our model for another group of subjects was validated, supporting the feasibility of our technique in the context of transfer learning. In conclusion, our study offers an insight into the discovery of the intrinsic mode of the human brain using whole‐brain resting‐state FC and DNNs.

show abstract

Functional connectome fingerprinting using shallow feedforward neural networks

Cited by 12 publications

References 8 publications

Large-Scale Functional Connectome Fingerprinting for Generalization and Transfer Learning in Neuroimaging

Large-Scale Functional Connectome Fingerprinting for Generalization and Transfer Learning in Neuroimaging

Functional Connectivity Uniqueness and Stability? A Signature of Cognitive and Psychiatric Problems in Children

Discovering individual fingerprints in resting‐state functional connectivity using deep neural networks

Contact Info

Product

Resources

About