CoChat
Sign Up
Cited by 29Functional Brain Connectivity StudiesMental Health Research TopicsEmotion and Mood Recognition

Predicting Brain Functional Connectivity Using Mobile Sensing

Open access
Obuchi, Mikio, Huckins, Jeremy F., Wang, Weichen, daSilva, Alex, Rogers, Courtney, Murphy, Eilis, Hedlund, Elin, Holtzheimer, Paul +2 more·2020·Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
DOI 10.1145/3381001PMC PMC9762691
Open this paper in CoChat — free account

Abstract

Brain circuit functioning and connectivity between specific regions allow us to learn, remember, recognize and think as humans. In this paper, we ask the question if mobile sensing from phones can predict brain functional connectivity. We study the brain resting-state functional connectivity (RSFC) between the ventromedial prefrontal cortex (vmPFC) and the amygdala, which has been shown by neuroscientists to be associated with mental illness such as anxiety and depression. We discuss initial results and insights from the NeuroSence study, an exploratory study of 105 first year college students using neuroimaging and mobile sensing across one semester. We observe correlations between several behavioral features from students' mobile phones and connectivity between vmPFC and amygdala, including conversation duration (r=0.365, p<0.001), sleep onset time (r=0.299, p<0.001) and the number of phone unlocks (r=0.253, p=0.029). We use a support vector classifier and 10-fold cross validation and show that we can classify whether students have higher (i.e., stronger) or lower (i.e., weaker) vmPFC-amygdala RSFC purely based on mobile sensing data with an F1 score of 0.793. To the best of our knowledge, this is the first paper to report that resting-state brain functional connectivity can be predicted using passive sensing data from mobile phones.

AI snapshot

Written from the paper’s full text.

TL;DR

This exploratory NeuroSence study asks whether passive mobile-phone sensing can predict resting-state functional connectivity (RSFC) between the vmPFC and amygdala, a circuit linked to anxiety and depression. Using data from first-year college students across one semester, it finds modest correlations between phone-derived behaviors and RSFC and reports that a support vector classifier can classify higher versus lower RSFC with an F1 score of 0.793.

Key findings

  • Mobile sensing features correlated with vmPFC-amygdala RSFC, including conversation duration (r=0.365, p<0.001), sleep onset time (r=0.299, p<0.001), and number of phone unlocks (abstract reports r=0.253, p=0.029; contributions section lists r=0.299, p=0.029).
  • Physical activity regularity, conversation features, time at social places, and regularity of visiting other students’ dorms were associated with higher RSFC; sleep duration and sleep variability were not significantly correlated.
  • A support vector classifier with RBF kernel using 23 features achieved the best classification of higher versus lower RSFC: F1=0.793, accuracy=0.791, precision=0.798, recall=0.788, AUROC=0.811.
  • The authors state the maximum correlation coefficient obtained was 0.368, so they cannot claim the correlations are strong.

Methods

  • NeuroSence recruited 105 first-year undergraduates (75 female, 30 male; mean age 18.2 years) at the start of fall semester; 92 were scanned, and 75 met mobile sensing data-quality criteria (≥18 hours/day for ≥14 days). The final median-split groups were 38 higher-RSFC and 37 lower-RSFC students.
  • Ground truth was fMRI-based resting-state functional connectivity between vmPFC and amygdala, computed with Pearson correlation and Fisher’s Z transform; mobile sensing data were collected over one academic term (79 days) via a smartphone app.
  • Analysis included Spearman correlations with Benjamini-Hochberg FDR correction and binary classification of higher versus lower RSFC using eight machine-learning algorithms, nested 10-fold outer and 3-fold inner cross-validation, 20 random seeds, and sequential forward feature selection.

Limitations

  • Stated: participants were first-year undergraduates at a US university, mostly female, so results may not generalize across age, gender, first-generation status, socioeconomics, or nationality; students with mental health conditions were not screened.
  • Stated: the study assumes vmPFC-amygdala RSFC relates to mental health, especially anxiety, but the physiological mechanism of anxiety is not fully elucidated; other brain-region pairs may yield different accuracy.
  • Evident from the text: correlations are modest (maximum r=0.368), the sample is relatively small, the sleep classifier has about ±30 minutes error, conversation sensing only infers being “around” conversation rather than active involvement, and the study lacks app-level and contextual information.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1145/3381001

Cites

99 of 99 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 36 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    The support vector classifier with 10-fold cross validation (nested with hyperparameter tuning) achieves an F1 score of 0.793 in classifying higher vs lower vmPFC-amygdala RSFC from mobile sensing data.

    “We use a support vector classifier and 10-fold cross validation and show that we can classify whether students have higher (i.e., stronger) or lower (i.e., weaker) vmPFC-amygdala RSFC purely based on mobile sensing data with an F1 score of 0.793.”
    • Also stated in: § INTRODUCTION · plain
      “After applying a 10-fold nested cross validation with hyperparameter tuning, the support vector classifier achieves an F1 score of 0.793.”
    Quote in the paperSubstantial claim
  • Findingown data

    Sleep onset time from mobile phones correlates with vmPFC-amygdala RSFC in first-year college students (r=0.299, p<0.001).

    “sleep onset time (r=0.299, p<0.001)”
    • Also stated in: § INTRODUCTION · plain
      “their sleep onset time (r=0.299, p<0.001)”
    Quote in the paperSubstantial claim
  • Findingown data

    Conversation duration from mobile phones correlates with vmPFC-amygdala RSFC in first-year college students (r=0.365, p<0.001).

    “including conversation duration (r=0.365, p<0.001)”
    • Also stated in: § INTRODUCTION · plain
      “including the conversation duration a student is around (r=0.365, p<0.001)”
    Quote in the paperSubstantial claim
  • Findingown data

    Number of phone unlocks correlates with vmPFC-amygdala RSFC in first-year college students (reported r=0.253, p=0.029 in Abstract; r=0.299, p=0.029 in Contributions).

    “the number of phone unlocks (r=0.253, p=0.029)”
    • Also stated in: § INTRODUCTION · plain
      “the number of phone unlocks (r=0.299, p=0.029) they initiate”
    Quote in the paperSubstantial claim
  • Findingown data

    The study identified a set of behaviors that correlate with vmPFC-amygdala connectivity.

    “We identified a set of behaviors that correlate with vmPFC-amygdala connectivity.”
    Quote in the paperSubstantial claim
  • Classified as probably not a substantial claim. 33
  • May combine two claims. 25

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    The average vmPFC-amygdala RSFC is 0.12 with a median of 0.14.

    “The average value of RSFC is 0.12 with the median of 0.14.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Four regularity features correlate with vmPFC-amygdala RSFC, and students who regularly walk or run tend to have higher RSFC.

    “Also, four regularity features correlate with vmPFC-amygdala RSFC in our dataset. Students who regularly walk or run prone to have higher RSFC.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    vmPFC-amygdala RSFC correlates with number of phone unlocks and unlock duration between 0 AM and 9 AM, but not with other time epochs.

    “Regarding phone usage, vmPFC-amygdala RSFC correlates with the number of phone unlocks and unlock duration between 0 AM and 9 AM, but not with the other time epochs.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    For the best SVC, the standard deviation of the F1 score across 20 models is 0.03 and the mean standard deviation of 10-fold cross validation is 0.17.

    “Regarding the variance of the 20 models of the best SVC, the standard deviation of the F1 score is 0.03 and the mean of the standard deviation of 10-Fold cross validation is 0.17.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    All subjects are first-year undergraduate students at a US university, with the majority being female.

    “First, all the subjects that participated in our study are first-year undergraduate students at a US university, with the majority being female students.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The distribution of vmPFC-amygdala RSFC approximates a normal distribution (Shapiro-Wilk p=0.515).

    “The distribution approximates a normal distribution as the null hypothesis is not rejected (p=0.515) in the Shapiro-Wilk test of normality [84].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [84] Shapiro Samuel Sanford and Wilk Martin B. 1965. An analysis of variance test for normality (complete samples). Biometrika 52, 3/4 (1965), 591– 611. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The higher and lower RSFC groups are well balanced, with 38 students in the higher RSFC group and 37 in the lower group.

    “Although we applied the median split for RSFC before filtering the participants, the number of higher RSFC and lower RSFC subjects is well balanced (i.e., specifically, 38 students in the higher RSFC group and 37 in the lower group).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Sleep onset time is the only sleep feature significantly correlated with vmPFC-amygdala RSFC, with later sleep onset associated with higher RSFC.

    “Finally, sleep onset time shows the most significant and only correlation among sleep features, implying RSFC is higher for students that go to sleep later at night.”
    • Also stated in: § Correlated Mobile Sensing Features · hedged
      “Our analysis indicates that vmPFC-amygdala RSFC is higher (i.e., stronger) when students go to sleep later.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Sleep duration (r=−0.161, p=0.168) and sleep variability features are not significantly correlated with vmPFC-amygdala RSFC.

    “Surprisingly, we do not find significant correlations for sleep duration (r=−0.161, p=0.168) and sleep variability features in our study.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    More frequent conversation and longer conversation duration (inferred from microphone) are associated with higher vmPFC-amygdala RSFC.

    “The result suggest that RSFC increases when subjects are around more frequent conversation and conversation of increasing duration; that is, our inference using the microphone infers that a subject is “around” conversation.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Of 105 consenting students, 92 opted to be scanned, and 75 of the 92 scanned subjects met the mobile sensing data quality criteria of 18 hours per day for at least 14 days.

    “of the 105 students consenting to be part of the NeuroSence study only 92 opted to be scanned at the start of the semester. In addition, only 75 of the 92 scanned subjects met the mobile sensing data quality criteria [100] of 18 hours of sensing data per day for a minimum of 14 days across the term.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [100] Wang Weichen, Harari Gabriella M, Wang Rui, Müller Sandrine R, Mirjafari Shayan, Masaba Kizito, and Campbell Andrew T. 2018. Sensing Behavioral Change over Time: Using Within-Person Variability Features from Mobile Sensing to Predict Personality Traits. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 2, 3 (2018), 141. source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    SVC results in the highest F1 score (0.793) using 23 features, whereas using 6 or 7 features results in F1 scores of 0.710 and 0.775, respectively.

    “In terms of performance, SVC results in the highest F1 score (0.793) using 23 features, whereas using the 6 or 7 features results in an F1 score of 0.710 and 0.775, respectively.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    The study does not screen for students with mental health conditions.

    “We do not screen for students with mental health conditions.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The SVC model's accuracy is 0.791, precision 0.798, recall 0.788, and AUROC 0.811.

    “Note that the accuracy is 0.791, precision is 0.798, recall is 0.788 and AUROC is 0.811.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The SVC with RBF kernel using the top two features has an F1 score of 0.655.

    “Figure 7 presents the boundary of SVC with RBF kernel using the top two features (F1 score is 0.655 in this case).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    During cross validation, 17 out of 20 models selected C=1.0 and gamma=0.1 as the best hyper-parameters.

    “During cross validation, 17 out of 20 models selected the pair of 1.0 for C and 0.1 for gamma as the best hyper-parameter.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Smartphone unlock duration and number of unlocks between 0 AM and 9 AM positively correlate with vmPFC-amygdala RSFC.

    “We find that correlation associated with smartphone usage relates to the unlock duration between time epoch of 0 AM - 9 AM. During the same epoch, we also observe that the number of unlocks positively correlates with RSFC.”
    • Also stated in: § Correlated Mobile Sensing Features · plain
      “Regarding phone usage, vmPFC-amygdala RSFC correlates with the number of phone unlocks and unlock duration between 0 AM and 9 AM, but not with the other time epochs.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingreasoning

    The linear SVC provided lower performance but would be more interpretable.

    “The linear SVC provided lower performance but would be more interpretable.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Huckins et al. found connections between mobile sensing and fMRI scans, including correlations between smartphone usage (unlock duration) and RSFC for the subgenual cingulate cortex.

    “Huckins et al. [44] present connections between mobile sensing technology and fMRI scans.”
    • Also stated in: § Neuroimaging and Mental Health · plain
      “Furthermore, they conduct correlation analysis between each functional connectivity value and smartphone usage (specifically unlock duration) in order to find which brain region is associated with depression and unlock duration.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [44] Huckins Jeremy F 2019. Fusing Mobile Phone Sensing and Brain Imaging to Assess Depression in College Students. Frontiers in Neuroscience 13 (2019), 248. 30949024 PMC6437560 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    The study's findings are applicable to a first-year college student cohort at a 4-year university.

    “While our study is exploratory and results preliminary, and our findings applicable to a first year college student cohort at a 4 year university”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Depressed adolescents have lower vmPFC-amygdala RSFC compared to healthy controls.

    “Depressed adolescents (N=48) have lower vmPFC-amygdala RSFC compared to healthy controls (N=53).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [20] Connolly Colm G, Ho Tiffany C, Blom Eva Henje, LeWinn Kaja Z, Sacchet Matthew D, Tymofiyeva Olga, Simmons Alan N, and Yang Tony T. 2017. Resting-state functional connectivity of the amygdala and longitudinal changes in depression severity in adolescent depression. Journal of affective disorders 207 (2017), 86– 94. 27716542 PMC5149416 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    RSFC is associated with individual differences such as anxiety and depression.

    “Researchers have found that RSFC is associated with a variety of individual differences, such as, personality [3], depression [47, 48], anxiety [50, 107], autism [18], dementia [6] and Alzheimer’s disease [93].”
    Quote in the paperNot a substantial claim0 of 4 sources checked
    • cites [47] Kaiser Roselinde H, Andrews-Hanna Jessica R, Wager Tor D, and Pizzagalli Diego A. 2015. Large-scale network dysfunction in major depressive disorder: a meta-analysis of resting-state functional connectivity. JAMA psychiatry 72, 6 (2015), 603– 611. 25785575 PMC4456260 source

      Not checked in the free run.

    • cites [48] Kaiser Roselinde H, Whitfield-Gabrieli Susan, Dillon Daniel G, Goer Franziska, Beltzer Miranda, Minkel Jared, Smoski Moria, Dichter Gabriel, and Pizzagalli Diego A. 2016. Dynamic resting-state functional connectivity in major depression. Neuropsychopharmacology 41, 7 (2016), 1822. 26632990 PMC4869051 source

      Not checked in the free run.

    • cites [50] Kim M Justin, Gee Dylan G, Loucks Rebecca A, Davis F Caroline, and Whalen Paul J. 2010. Anxiety dissociates dorsal and ventral medial prefrontal cortex functional connectivity with the amygdala at rest. Cerebral cortex 21, 7 (2010), 1667– 1673. 21127016 PMC3116741

      Not checked in the free run.

    • cites [107] Zhao Xiao-Hu, Wang Pei-Jun, Li Chun-Bo, Hu Zheng-Hui, Xi Qian, Wu Wen-Yuan, and Tang Xiao-Wei. 2007. Altered default mode network activity in patient with anxiety disorders: an fMRI study. European journal of radiology 63, 3 (2007), 373– 378. 17400412 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Sleep deprivation is associated with lower vmPFC-amygdala functional connectivity.

    “Prior research links vmPFC-amygdala functional connectivity with sleep, suggesting that sleep deprivation (a common condition reported by students during the college years) is associated with lower vmPFC-amygdala functional connectivity [24, 105].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [24] Feng Pan, Becker Benjamin, Feng Tingyong, and Zheng Yong. 2018. Alter spontaneous activity in amygdala and vmPFC during fear consolidation following 24 h sleep deprivation. NeuroImage 172 (2018), 461– 469. 29378319 source

      Not checked in the free run.

    • cites [105] Yoo Seung-Schik, Gujar Ninad, Hu Peter, Jolesz Ferenc A, and Walker Matthew P. 2007. The human emotional brain without sleep—a prefrontal amygdala disconnect. Current Biology 17, 20 (2007), R877– R878. 17956744 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Kim et al. found that RSFC between the right amygdala and vmPFC is negatively correlated with state anxiety.

    “Kim et al. [50] conduct a study involving 28 students and find that the RSFC between the right amygdala and vmPFC regions is negatively correlated with state anxiety.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [50] Kim M Justin, Gee Dylan G, Loucks Rebecca A, Davis F Caroline, and Whalen Paul J. 2010. Anxiety dissociates dorsal and ventral medial prefrontal cortex functional connectivity with the amygdala at rest. Cerebral cortex 21, 7 (2010), 1667– 1673. 21127016 PMC3116741

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductreasoning

    The conversation inference indicates the subject is around conversation, not necessarily actively involved, and is used as a proxy for social engagement or isolation.

    “We can not be completely sure if the subject is actively involved in the conversation, the inference of conversation is associated with the subject being around conversation. We use this as a proxy for social engagement or isolation.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    The analysis makes best effort to prevent over-fitting or bias in training data.

    “In our analysis, we make our best effort to prevent over-fitting or to bias training data.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    Twenty models with different random seeds are trained and averaged to prevent bias from initial seed selection.

    “Furthermore, we evaluate 20 models using different random seeds when training each model and average the score to prevent bias from the initial seed selection.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Visits to religious places have been shown to have an association with mental wellbeing (i.e., anxiety).

    “visits to religious places have been shown to have an association with mental wellbeing (i.e., anxiety) [43, 80].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [43] Huang Yu, Xiong Haoyi, Leach Kevin, Zhang Yuyan, Chow Philip, Fua Karl, Teachman Bethany A., and Barnes Laura E.. 2016. Assessing Social Anxiety Using Gps Trajectories and Point-of-interest Data. In Proceedings of the 2016 ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp ‘16). ACM, New York, NY, USA, 898– 903. 10.1145/2971648.2971761

      Not checked in the free run.

    • cites [80] Saeb Sohrab, Lattie Emily G, Kording Konrad P, and Mohr David C. 2017. Mobile phone detection of semantic location and its relationship to depression and anxiety. JMIR mHealth and uHealth 5, 8 (2017), e112. 28798010 PMC5571235

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    The model exceeds the 50% baseline, reasoning that mobile sensing contains a signal for predicting brain activity.

    “Our model exceeds the 50% baseline, which is relevant to random prediction, reasoning that mobile sensing contains a signal for predicting brain activity.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Study design and conductnone given

    The NeuroSence sensing platform is a tailored version of a mobile phone sensing platform that has been validated by a number of published studies.

    “The NeuroSence sensing platform is a tailored version of a mobile phone sensing platform that has been validated by a number of published studies.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The NeuroSence study was approved by the study institution's Institutional Review Board.

    “The NeuroSence study is approved by the study institution’s Institutional Review Board”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Previous studies find that sleep deprivation is related to weaker vmPFC-amygdala RSFC.

    “In terms of sleep, previous studies [24, 105] find that sleep deprivation is related to weaker vmPFC-amygdala RSFC.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [24] Feng Pan, Becker Benjamin, Feng Tingyong, and Zheng Yong. 2018. Alter spontaneous activity in amygdala and vmPFC during fear consolidation following 24 h sleep deprivation. NeuroImage 172 (2018), 461– 469. 29378319 source

      Not checked in the free run.

    • cites [105] Yoo Seung-Schik, Gujar Ninad, Hu Peter, Jolesz Ferenc A, and Walker Matthew P. 2007. The human emotional brain without sleep—a prefrontal amygdala disconnect. Current Biology 17, 20 (2007), R877– R878. 17956744 source

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    The F1 score of 0.793 (0.791 accuracy) is a strong result because it achieves similar accuracy to binary classification performance reported in mobile mental health sensing studies.

    “The F1 score of 0.793 (0.791 accuracy) is a strong result as it achieves a similar accuracy of binary classification performance reported in mobile mental health sensing studies, as discussed in Section 3.1.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The study considers it has good power.

    “We consider we have good power but increasing the number would be the next step as well as studying the replicability of our results at higher scale.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductcitation

    The sleep classifier used has an error of +/− 30 minutes, but the trends remain accurate.

    “While there is an error of +/− 30 minutes in the sleep classifier used in [17], the trends remain accurate.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [17] Chen Zhenyu, Lin Mu, Chen Fanglin, Lane Nicholas D, Cardone Giuseppe, Wang Rui, Li Tianxing, Chen Yiqiang, Choudhury Tanzeem, and Campbell Andrew T. 2013. Unobtrusive sleep monitoring using smartphones. In Proceedings of the 7th International Conference on Pervasive Computing Technologies for Healthcare. ICST (Institute for Computer Sciences, Social-Informatics and …, 145– 152. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationown data

    Considering the dataset size, a smaller number of features may be more robust and generalizable.

    “Considering our dataset size, a smaller number of features may be more robust and generalizable.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductreasoning

    A nested cross validation scheme (outer 10-fold, inner 3-fold) is used to prevent over-fitting, build a generalized model, and restrict hyper-parameters from being solely adapted to training data.

    “We use a nested cross validation (CV) scheme to prevent over-fitting, to build a generalized model and to restrict hyper-parameters from being solely adapted to the training data (see Figure 3). The outer cross validation for evaluating the performance of the machine learning model is 10-Fold, and the inner CV for hyper-parameter tuning is 3-Fold.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    The same individual's data are not included in both training and testing sets during record-wise cross validation.

    “We do not include the same individual’s data on both training and testing set while doing record-wise cross validation.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Mobile sensing has the potential to bring new insights to neuroscience by capturing user behavior in-the-wild.

    “Mobile sensing, which allows us to capture the user’s behavior in-the-wild, has the potential to brings new insights to neuroscience, where today research studies typical rely on self-report surveys.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Only subjects with 5 minutes or more of low-motion data are included in the analysis to minimize the influence of motion.

    “Only subjects with 5 minutes or more of low-motion data are included in the analysis to minimize the influence of motion.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductcitation

    Mobile sensing data quality criteria exclude data with less than 18 hours of data per day and less than 14 days of data during the term.

    “Specifically, in terms of mobile sensing data we exclude data with less than 18 hours [100] of data per day and less than 14 days of data during the term.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [100] Wang Weichen, Harari Gabriella M, Wang Rui, Müller Sandrine R, Mirjafari Shayan, Masaba Kizito, and Campbell Andrew T. 2018. Sensing Behavioral Change over Time: Using Within-Person Variability Features from Mobile Sensing to Predict Personality Traits. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 2, 3 (2018), 141. source

      Not checked in the free run.

  • May combine two claims.

    Study design and conductown data

    The correlations are modest at best, with a maximum r-value of 0.368, so the correlations cannot be claimed to be strong.

    “Specifically, we find many pairs of variables having significant p-values; however, the correlation coefficient (r) is modest at best. In fact, the maximum r-value we obtain is 0.368. Therefore, we cannot claim that our correlations are strong [5].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [5] Akoglu Haldun. 2018. User’s guide to correlation coefficients. Turkish journal of emergency medicine 18, 3 (2018), 91– 93. 30191186 PMC6107969

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    A number of neuroscientists find that higher vmPFC-amygdala functional connectivity is correlated with lower depression and anxiety.

    “As discussed in the related work section, a number of neuroscientists find that higher vmPFC-amygdala functional connectivity is correlated with lower depression and anxiety [20, 50].”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [20] Connolly Colm G, Ho Tiffany C, Blom Eva Henje, LeWinn Kaja Z, Sacchet Matthew D, Tymofiyeva Olga, Simmons Alan N, and Yang Tony T. 2017. Resting-state functional connectivity of the amygdala and longitudinal changes in depression severity in adolescent depression. Journal of affective disorders 207 (2017), 86– 94. 27716542 PMC5149416 source

      Not checked in the free run.

    • cites [50] Kim M Justin, Gee Dylan G, Loucks Rebecca A, Davis F Caroline, and Whalen Paul J. 2010. Anxiety dissociates dorsal and ventral medial prefrontal cortex functional connectivity with the amygdala at rest. Cerebral cortex 21, 7 (2010), 1667– 1673. 21127016 PMC3116741

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    The Benjamini-Hochberg procedure is used to regulate the false discovery rate in the exploratory correlation analysis.

    “To avoid this issue, we use the Benjamini-Hochberg procedure [11], which regulates the false discovery rate (FDR) in our exploratory analysis.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [11] Benjamini Yoav and Hochberg Yosef. 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal statistical society: series B (Methodological) 57, 1 (1995), 289– 300. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Study design and conductnone given

    The reason the SVC selects specific features cannot be easily comprehended.

    “but we cannot easily comprehend why the model selects those specific features.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductown data

    The study used 10-fold nested cross validation with hyperparameter tuning to evaluate the machine learning models.

    “After applying a 10-fold nested cross validation with hyperparameter tuning”
    • Also stated in: § Abstract · plain
      “We use a support vector classifier and 10-fold cross validation”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationown data

    vmPFC-amygdala RSFC is stronger when students are around conversation or spend more time in social or working locations where they typically engage with other students.

    “Our results indicate that vmPFC-amygdala RSFC increases (i.e., is stronger) when students are around a conversation or spend more time in social or working locations where they typically engage with other students (e.g., spend time at friends’ dorms).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Frequent aerobic exercise (i.e., running) plays a vital role in improving mental wellbeing.

    “It is widely accepted that frequent aerobic exercise (i.e., running) plays a vital role in improving mental wellbeing [85].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [85] Sharma Ashish, Madaan Vishal, and Petty Frederick D. 2006. Exercise for mental health. Primary care companion to the Journal of clinical psychiatry 8, 2 (2006), 106– 106. PMC1470658 16862239

      Not checked in the free run.

  • May combine two claims.

    Study design and conductcitation

    The NeuroSence application detects sleep and speech/conversation data using pre-trained classifiers on accelerometer and audio data.

    “Furthermore, our application detects sleep [17] and speech/conversation data [54, 76] by incorporating pre-trained classifiers using accelerometer and audio data from a microphone, respectively.”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [17] Chen Zhenyu, Lin Mu, Chen Fanglin, Lane Nicholas D, Cardone Giuseppe, Wang Rui, Li Tianxing, Chen Yiqiang, Choudhury Tanzeem, and Campbell Andrew T. 2013. Unobtrusive sleep monitoring using smartphones. In Proceedings of the 7th International Conference on Pervasive Computing Technologies for Healthcare. ICST (Institute for Computer Sciences, Social-Informatics and …, 145– 152. source

      Not checked in the free run.

    • cites [54] Lane Nicholas D, Mohammod Mashfiqui, Lin Mu, Yang Xiaochao, Lu Hong, Ali Shahid, Doryab Afsaneh, Berke Ethan, Choudhury Tanzeem, and Campbell Andrew. 2011. Bewell: A smartphone application to monitor, model and promote wellbeing. In 5th international ICST conference on pervasive computing technologies for healthcare. 23– 26. source

      Not checked in the free run.

    • cites [76] Rabbi Mashfiqui, Ali Shahid, Choudhury Tanzeem, and Berke Ethan. 2011. Passive and in-situ assessment of mental and physical well-being using mobile sensors. In Proceedings of the 13th international conference on Ubiquitous computing. ACM, 385– 394. PMC4180507 25285324 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The two most common reasons students go to sleep later, discussed in the StudentLife study, are academic demands and social events.

    “We do not have any further data on why students go to sleep later but the two most common reasons discussed in the StudentLife study [96] are academic demands such as assignment due dates and exams and social events such as hanging out with friends and parties.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [96] Wang Rui, Chen Fanglin, Chen Zhenyu, Li Tianxing, Harari Gabriella, Tignor Stefanie, Zhou Xia, Ben-Zeev Dror, and Campbell Andrew T. 2014. StudentLife: assessing mental health, academic performance and behavioral trends of college students using smartphones. In Proceedings of the 2014 ACM international joint conference on pervasive and ubiquitous computing. ACM, 3– 14. source

      Not checked in the free run.

50 of 50

  • The paper relies on this without support.

    Mobile sensing data can provide unbiased accounts of behaviors that participants would typically self-report.

    “From the examples above, it is clear how mobile sensing could provide unbiased accounts of the behaviors that participants would typically have to self-report.”

  • The paper relies on this without support.

    The mobile sensing features used (conversation duration, sleep onset time, number of phone unlocks) are valid measures of the corresponding behaviors.

    “We use a smartphone application to continuously collect students’ sensing and behavioral data from their phones over one semester as predictive variables for the study.”

  • The authors state this as an assumption.

    The performance of the machine learning model (SVC) is reliable.

    “We assume that the performance of our machine learning model (i.e., SVC) is reliable.”

  • The authors state this as an assumption.

    vmPFC-amygdala RSFC is associated with mental health, particularly anxiety, in the current study.

    “Based on prior research in neuroscience, we assume that vmPFC-amygdala RSFC is associated with mental health - anxiety, in particular - in the current study.”
4 of 4

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Not all consenting students were scanned

    Minor

    The paper says all consenting students took part in an fMRI session, but later states that only 92 of the 105 consenting students opted to be scanned.

  • Conflicting correlation for phone unlocks

    Minor

    The abstract reports r=0.253 for the number of phone unlocks, while the contributions section reports r=0.299 for the same relationship; both give p=0.029.

  • Reference [98] is described as two different studies

    Minor

    The text cites [98] for a study predicting depression in college students, but the listed title of [98] identifies it as a study predicting schizophrenia symptom trajectories.

3 of 3

Cites

Flagged first, then by citations

Research works printed but not linked: 8

  1. Jessica R Andrews-Hanna , Matthew D Grilli , and Muireann Irish . 2019. A review and reappraisal of the default network in normal aging and dementia . In Oxford Research Encyclopedia of Psychology. Jessica R Andrews-Hanna, Matthew D Grilli, and Muireann Irish. 2019. A review and reappraisal of the default network in normal aging and dementia. In Oxford Research Encyclopedia of Psychology.
  2. Anxiety and Depression Association of America , ADAA 2019 . https://adaa.org/. ( 2019 ). Anxiety and Depression Association of America, ADAA 2019. https://adaa.org/. (2019).
  3. Despina E Ganella , Marjolein EA Barendse , Jee H Kim, and Sarah Whittle. 2017 . Prefrontal-amygdala connectivity and state anxiety during fear extinction recall in adolescents. Frontiers in human neuroscience 11 (2017), 587. Despina E Ganella, Marjolein EA Barendse, Jee H Kim, and Sarah Whittle. 2017. Prefrontal-amygdala connectivity and state anxiety during fear extinction recall in adolescents. Frontiers in human neuroscience 11 (2017), 587.
  4. Yu Huang , Haoyi Xiong , Kevin Leach , Yuyan Zhang , Philip Chow , Karl Fua , Bethany A. Teachman , and Laura E. Barnes . 2016. Assessing Social Anxiety Using Gps Trajectories and Point-of-interest Data . In Proceedings of the 2016 ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp '16) . ACM, New York, NY, USA, 898--903. https://doi.org/10.1145/2971648.2971761 10.1145/2971648.2971761 Yu Huang, Haoyi Xiong, Kevin Leach, Yuyan Zhang, Philip Chow, Karl Fua, Bethany A. Teachman, and Laura E. Barnes. 2016. Assessing Social Anxiety Using Gps Trajectories and Point-of-interest Data. In Proceedings of the 2016 ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp '16). ACM, New York, NY, USA, 898--903. https://doi.org/10.1145/2971648.2971761 · DOI · Paper page
  5. Richard P Mattick and J Christopher Clarke . 1998. Development and validation of measures of social phobia scrutiny fear and social interaction anxiety. Behaviour research and therapy 36, 4 ( 1998 ), 455--470. Richard P Mattick and J Christopher Clarke. 1998. Development and validation of measures of social phobia scrutiny fear and social interaction anxiety. Behaviour research and therapy 36, 4 (1998), 455--470.
  6. David Premack and Guy Woodruff . 1978. Does the chimpanzee have a theory of mind? Behavioral and brain sciences 1, 4 ( 1978 ), 515--526. David Premack and Guy Woodruff. 1978. Does the chimpanzee have a theory of mind? Behavioral and brain sciences 1, 4 (1978), 515--526.
  7. David R Reetz , Brian Krylowicz , and Brian Mistler . 2014 . The association for university and college counseling center directors annual survey . Aurora 51 (2014), 60506 . David R Reetz, Brian Krylowicz, and Brian Mistler. 2014. The association for university and college counseling center directors annual survey. Aurora 51 (2014), 60506.
  8. Charles D Spielberger . 2010. State-Trait anxiety inventory. The Corsini encyclopedia of psychology ( 2010 ), 1--1. Charles D Spielberger. 2010. State-Trait anxiety inventory. The Corsini encyclopedia of psychology (2010), 1--1.

Cited by

Sorted by citations

Related

Authors
Obuchi, M., Huckins, J. F., Wang, W., daSilva, A., Rogers, C., Murphy, E., Hedlund, E., Holtzheimer, P., Mirjafari, S., & Campbell, A.
Affiliations
  • Dartmouth College (US)
  • Dartmouth–Hitchcock Medical Center (US)
ORCID
Venue
Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Publisher
Association for Computing Machinery (ACM)
Published
2020-03-18
Volume
4
Issue
1
Pages
1-22
ISSN
2474-9567 (ISSN-L)
DOI
10.1145/3381001
PMC
PMC9762691
OpenAlex
W3012180041
Semantic Scholar
48ef63b5ea589796b0198b33d383c5d25a9ca419
Cited by
29
References
99
Citations by year
2026: 1 · 2025: 1 · 2024: 5 · 2023: 6 · 2022: 6 · 2021: 7 · 2020: 3
FWCI
2.96
Type
journal article
Language
en
Open access
Licence
www.acm.org/publications/policies/copyright_policy#Background (vor)
Locations
  • Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies (publishedVersion)
  • PubMed (publishedVersion)
  • PubMed Central (acceptedVersion)
Funders
National Institute of Mental Health (5R01MH059282-12)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Functional Brain Connectivity Studies
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Mental Health Research Topics
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Emotion and Mood Recognition
Keywords
resting-state functional connectivity, mobile sensing, ventromedial prefrontal cortex, amygdala, passive sensing, sleep latency, first-year college students, mental illness, anxiety, depression

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review