Contrastive machine learning reveals the structure of neuroanatomical variation within autism

Aglinskas, Aidas; Hartshorne, Joshua K.; Anzellotti, Stefano

doi:10.1126/science.abm2461

Cited by 53 publications

(45 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, we applied these models to a large database of neuroanatomical scans (ABIDE I, 6 512 ASD, 470 TD) extracting ASD-specific features of neuroanatomy that vary within the ASD population. 7 We show that the CVAE approach improves over previous methods of studying individual variation in ASD in several key areas.…”

Section: The Techniquementioning

confidence: 83%

“…Contrastive variational autoencoders (CVAEs 5 ) are unsupervised deep learning models that take in samples from two populations, such as typical controls (TC) and ASD, and can be trained to isolate features that capture variation specific to one population (features that are ASD‐specific) from features that are common to both (features that are shared). Recently, we applied these models to a large database of neuroanatomical scans (ABIDE I, 6 512 ASD, 470 TD) extracting ASD‐specific features of neuroanatomy that vary within the ASD population 7 . We show that the CVAE approach improves over previous methods of studying individual variation in ASD in several key areas.…”

Section: The Techniquementioning

confidence: 98%

See 1 more Smart Citation

Precision psychiatry requires disentangling disorder‐specific variation: The case of ASD

Aglinskas

Anzellotti

2022

Clinical & Translational Med

Self Cite

View full text Add to dashboard Cite

Individual variation within autism spectrum disorder (ASD) is an established fact. Interventions that are effective for some individuals are not effective for others. 1 Symptom severity, adaptive behavior scores and intelligence quotient partly correlate with differences in intervention outcomes, but a large amount of variation remains unexplained. 2 Predicting how an individual would respond to different interventions is a key goal for clinical research. Such information could be used to recommend personalized care, curbing the process of trial and error and leading to benefits for the patients who would be able to access optimized interventions early. However, to make this possible, we need to gain a deeper understanding of individual variation within autism.Neural data are a valuable source of information to make sense of individual variation. The brain mediates the relationship between genetics and behavior. Therefore, we can expect to observe individual variation within autism at the level of neural measures. Consistent with this, neuroimaging investigations have shown highly variable neuroanatomy among ASD participants. 3 However, studying this variation has proven challenging: all brains (both with and without ASD) are unique and differ from another due to numerous genetic and environmental causes not related to ASD. 4 Additional variation in common to ASD participants and controls is introduced by the process of measurement (for example, in large datasets different participants might be measured at different sites). ASD-specific variation, therefore, is entangled with a large amount of variation from sources that are shared between typically developing individuals and those with ASD.Traditional approaches, such as case-control matching, can help to address these challenges when the factors that need to be matched are few and known. If some factors are unknown, however, it is impossible to match cases to controls along those factors. Furthermore, evenThis 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.

show abstract

Section: The Techniquementioning

confidence: 83%

Section: The Techniquementioning

confidence: 98%

Precision psychiatry requires disentangling disorder‐specific variation: The case of ASD

Aglinskas

Anzellotti

2022

Clinical & Translational Med

Self Cite

View full text Add to dashboard Cite

show abstract

“…Recently, a deep contrast variational autoencoder was used to extract neuroanatomical features from MRI data to identify brain dysfunction that can be attributed to ASD and not to other causes of individual variation. 146 …”

Section: How Can ML Help Psychiatry?mentioning

confidence: 99%

Modern views of machine learning for precision psychiatry

Chen¹,

Kulkarni²,

Galatzer‐Levy³

et al. 2022

Patterns

View full text Add to dashboard Cite

“…The existing methods in screening ASD mainly rely on scale screening, such as M-CHAT and social communication questionnaire (SCQ) by parent’s report, with some subjective errors and long time-consuming. Compared with the disadvantages of traditional methods, machine learning has risen to be a promising alternative in the screening and diagnosis of ASD ( 15 , 16 ). Machine learning aims to construct predictive models from the datasets, which encompasses search methods, artificial intelligence, and mathematical modeling.…”

Section: Introductionmentioning

confidence: 99%

A multimodal machine learning system in early screening for toddlers with autism spectrum disorders based on the response to name

et al. 2023

View full text Add to dashboard Cite

BackgroundReduced or absence of the response to name (RTN) has been widely reported as an early specific indicator for autism spectrum disorder (ASD), while few studies have quantified the RTN of toddlers with ASD in an automatic way. The present study aims to apply a multimodal machine learning system (MMLS) in early screening for toddlers with ASD based on the RTN.MethodsA total of 125 toddlers were recruited, including ASD (n = 61), developmental delay (DD, n = 31), and typical developmental (TD, n = 33). Procedures of RTN were, respectively, performed by the evaluator and caregiver. Behavioral data were collected by eight-definition tripod-mounted cameras and coded by the MMLS. Response score, response time, and response duration time were accurately calculated to evaluate RTN.ResultsTotal accuracy of RTN scores rated by computers was 0.92. In both evaluator and caregiver procedures, toddlers with ASD had significant differences in response score, response time, and response duration time, compared to toddlers with DD and TD (all P-values < 0.05). The area under the curve (AUC) was 0.81 for the computer-rated results, and the AUC was 0.91 for the human-rated results. The accuracy in the identification of ASD based on the computer- and human-rated results was, respectively, 74.8 and 82.9%. There was a significant difference between the AUC of the human-rated results and computer-rated results (Z = 2.71, P-value = 0.007).ConclusionThe multimodal machine learning system can accurately quantify behaviors in RTN procedures and may effectively distinguish toddlers with ASD from the non-ASD group. This novel system may provide a low-cost approach to early screening and identifying toddlers with ASD. However, machine learning is not as accurate as a human observer, and the detection of a single symptom like RTN is not sufficient enough to detect ASD.

show abstract

Contrastive machine learning reveals the structure of neuroanatomical variation within autism

Cited by 53 publications

References 31 publications

Precision psychiatry requires disentangling disorder‐specific variation: The case of ASD

Precision psychiatry requires disentangling disorder‐specific variation: The case of ASD

Modern views of machine learning for precision psychiatry

A multimodal machine learning system in early screening for toddlers with autism spectrum disorders based on the response to name

Contact Info

Product

Resources

About