About

Birkan Tunç, PhD, is a Research Associate Professor of Psychiatry at the University of Pennsylvania and a Research Scientist at Children’s Hospital of Philadelphia. His educational background includes a BSc in Industrial Engineering, an MSc, and a PhD in Computational Science & Engineering from Istanbul Technical University. His research expertise encompasses machine learning, computer vision, computational behavior analysis, and brain connectivity. His work involves developing automated tract extraction methods, characterizing population heterogeneity, and establishing links between sex-related differences in the structural connectome and behavior. Additionally, he focuses on individualized mapping of white matter pathways for neurosurgical planning and the inference of meso-scale structures in networks, contributing to the understanding of brain connectivity and its implications in neuropsychiatric conditions.

Research topics

• Computer science
• Artificial intelligence
• Psychology
• Computer vision
• Neuroscience

Selected publications

• Stakeholder Perspectives on Humanistic Implementation of Computer Perception in Health Care: Qualitative Study

  JMIR Mental Health · 2025-11-21

  articleOpen access

  BACKGROUND: Computer perception (CP) technologies-including digital phenotyping, affective computing, and related passive sensing approaches-offer unprecedented opportunities to personalize health care, especially mental health care, yet they also provoke concerns about privacy, bias, and the erosion of empathic, relationship-centered practice. At present, it remains elusive what stakeholders who design, deploy, and experience these tools in real-world settings perceive as the risks and benefits of CP technologies. OBJECTIVE: This study aims to explore key stakeholder perspectives on the potential benefits, risks, and concerns associated with integrating CP technologies into patient care. A better understanding of these concerns is crucial for responding to and mitigating such concerns via design implementation strategies that augment, rather than compromise, patient-centered and humanistic care and associated outcomes. METHODS: We conducted in-depth, semistructured interviews with 102 stakeholders involved at key points in CP's development and implementation: adolescent patients (n=20) and their caregivers (n=20); frontline clinicians (n=20); technology developers (n=21); and ethics, legal, policy, or philosophy scholars (n=21). Interviews (~ 45 minutes each) explored perceived benefits, risks, and implementation challenges of CP in clinical care. Transcripts underwent thematic analysis by a multidisciplinary team; reliability was enhanced through double coding and consensus adjudication. RESULTS: Stakeholders raised concerns across 7 themes: (1) Data Privacy and Protection (88/102, 86.3%); (2) Trustworthiness and Integrity of CP Technologies (72/102, 70.6%); (3) direct and indirect Patient Harms (65/102, 63.7%); (4) Utility and Implementation Challenges (60/102, 58.8%); (5) Patient-Specific Relevance (24/102, 23.5%); (6) Regulation and Governance (17/102, 16.7%); and (7) Philosophical Critiques of reductionism (13/102, 12.7%). A cross-cutting insight was the primacy of context and subjective meaning in determining whether CP outputs are clinically valid and actionable. Participants warned that without attention to these factors, algorithms risk misclassification and dehumanization of care. CONCLUSIONS: To operationalize humanistic safeguards, we propose "personalized road maps": co-designed plans that predetermine which metrics will be monitored, how and when feedback is shared, thresholds for clinical action, and procedures for reconciling discrepancies between algorithmic inferences and lived experience. Road maps embed patient education, dynamic consent, and tailored feedback, thereby aligning CP deployment with patient autonomy, therapeutic alliance, and ethical transparency. This multistakeholder study provides the first comprehensive, evidence-based account of relational, technical, and governance challenges raised by CP tools in clinical care. By translating these insights into personalized road maps, we offer a practical framework for developers, clinicians, and policy makers seeking to harness continuous behavioral data while preserving the humanistic core of care.

  PublisherDOI

• Stakeholder Perspectives on Humanistic Implementation of Computer Perception in Healthcare: A Qualitative Study

  ArXiv.org · 2025-08-04

  preprintOpen access

  Computer perception (CP) technologies (digital phenotyping, affective computing and related passive sensing approaches) offer unprecedented opportunities to personalize healthcare, but provoke concerns about privacy, bias and the erosion of empathic, relationship-centered practice. A comprehensive understanding of perceived risks, benefits, and implementation challenges from those who design, deploy and experience these tools in real-world settings remains elusive. This study provides the first evidence-based account of key stakeholder perspectives on the relational, technical, and governance challenges raised by the integration of CP technologies into patient care. We conducted in-depth, semi-structured interviews with 102 stakeholders: adolescent patients and their caregivers, frontline clinicians, technology developers, and ethics, legal, policy or philosophy scholars. Transcripts underwent thematic analysis by a multidisciplinary team; reliability was enhanced through double coding and consensus adjudication. Stakeholders articulated seven interlocking concern domains: (1) trustworthiness and data integrity; (2) patient-specific relevance; (3) utility and workflow integration; (4) regulation and governance; (5) privacy and data protection; (6) direct and indirect patient harms; and (7) philosophical critiques of reductionism. To operationalize humanistic safeguards, we propose "personalized roadmaps": co-designed plans that predetermine which metrics will be monitored, how and when feedback is shared, thresholds for clinical action, and procedures for reconciling discrepancies between algorithmic inferences and lived experience. By translating these insights into personalized roadmaps, we offer a practical framework for developers, clinicians and policymakers seeking to harness continuous behavioral data while preserving the humanistic core of care.

  PublisherOA PDFDOI

• 3D Face Reconstruction Error Decomposed: A Modular Benchmark for Fair and Fast Method Evaluation

  2025-05-26 · 1 citations

  articleOpen accessSenior author

  Computing the standard benchmark metric for 3D face reconstruction, namely geometric error, requires a number of steps, such as mesh cropping, rigid alignment, or point correspondence. Current benchmark tools are monolithic (they implement a specific combination of these steps), even though there is no consensus on the best way to measure error. We present a toolkit for a Modularized 3D Face reconstruction Benchmark (M3DFB), where the fundamental components of error computation are segregated and interchangeable, allowing one to quantify the effect of each. Furthermore, we propose a new component, namely correction, and present a computationally efficient approach that penalizes for mesh topology inconsistency. Using this toolkit, we test 16 error estimators with 10 reconstruction methods on two real and two synthetic datasets. Critically, the widely used ICP-based estimator provides the worst benchmarking performance, as it significantly alters the true ranking of the top- 5 reconstruction methods. Notably, the correlation of ICP with the true error can be as low as 0.41. Moreover, non-rigid alignment leads to significant improvement (correlation larger than 0.90), highlighting the importance of annotating 3D landmarks on datasets. Finally, the proposed correction scheme, together with non-rigid warping, leads to an accuracy on a par with the best non-rigid ICP-based estimators, but runs an order of magnitude faster. Our open-source codebase is designed for researchers to easily compare alternatives for each component, thus helping accelerating progress in benchmarking for 3D face reconstruction and, furthermore, supporting the improvement of learned reconstruction methods, which depend on accurate error estimation for effective training.

  PublisherOA PDFDOI

• Bitbox: Behavioral Imaging Toolbox for Computational Analysis of Behavior from Videos

  arXiv (Cornell University) · 2025-12-19

  preprintOpen accessSenior author

  Computational measurement of human behavior from video has recently become feasible due to major advances in AI. These advances now enable granular and precise quantification of facial expression, head movement, body action, and other behavioral modalities and are increasingly used in psychology, psychiatry, neuroscience, and mental health research. However, mainstream adoption remains slow. Most existing methods and software are developed for engineering audiences, require specialized software stacks, and fail to provide behavioral measurements at a level directly useful for hypothesis-driven research. As a result, there is a large barrier to entry for researchers who wish to use modern, AI-based tools in their work. We introduce Bitbox, an open-source toolkit designed to remove this barrier and make advanced computational analysis directly usable by behavioral scientists and clinical researchers. Bitbox is guided by principles of reproducibility, modularity, and interpretability. It provides a standardized interface for extracting high-level behavioral measurements from video, leveraging multiple face, head, and body processors. The core modules have been tested and validated on clinical samples and are designed so that new measures can be added with minimal effort. Bitbox is intended to serve both sides of the translational gap. It gives behavioral researchers access to robust, high-level behavioral metrics without requiring engineering expertise, and it provides computer scientists a practical mechanism for disseminating methods to domains where their impact is most needed. We expect that Bitbox will accelerate integration of computational behavioral measurement into behavioral, clinical, and mental health research. Bitbox has been designed from the beginning as a community-driven effort that will evolve through contributions from both method developers and domain scientists.

  PublisherDOI

• Perceived Sensitivity of Sensor-Based Digital Health Data: A Qualitative Interview Study (Preprint)

  JMIR mhealth and uhealth · 2025-06-09

  articleOpen access
  PublisherDOI

• Stakeholder Perspectives on Trust and Transparency around Digital Health Data: A Qualitative Interview Study (Preprint)

  2025-06-09 · 2 citations

  preprint

  <sec> <title>BACKGROUND</title> The integration of digital health tools and algorithms in mental healthcare brings a transformative era where technologies like digital phenotyping, affective computing, and computational behavioral analysis are used to passively collect and analyze data from patients outside clinical settings. These "computer perception" (CP) technologies offer new insights into symptom manifestation in daily life while generating large volumes of potentially sensitive data that raise significant data privacy concerns requiring high levels of patient awareness and consent. Empirical research is lacking into stakeholder understandings and perspective towards the collection, management and protection of these data types. </sec> <sec> <title>OBJECTIVE</title> This study aimed to explore key stakeholder perspectives on the sensitivity of CP data, trust in existing data protections, willingness to share CP data externally, and desire for transparency of CP data transactions outside of the clinical space. </sec> <sec> <title>METHODS</title> As part of a larger, multi-site study, we conducted qualitative interviews (n=40) via Zoom with 20 adolescents (aged 12-17 years) familiar with CP tools and their caregivers (n=20). Interviews consisted of a series of open-ended questions regarding stakeholders’ perspectives on privacy, data security, and the use and exchange of CP data. We developed a qualitative codebook to identify and label thematic patterns in responses to questions addressing the topics above, using thematic content analysis to inductively identify themes. Each interview was coded by merging work from at least two separate coders, and several team members contributed to qualitative analysis. </sec> <sec> <title>RESULTS</title> Most adolescents and caregivers viewed CP data as highly sensitive and expressed a reluctance to share these data beyond their clinical teams. While many participants expressed trust in existing data protections to protect CP data, they often misunderstood or overestimated the extent of protections like HIPAA to safeguard CP data. </sec> <sec> <title>CONCLUSIONS</title> Our findings underscore the critical need for clear and effective patient communication and education about the risks, benefits, and protections associated with CP data through informed consent protocols. To promote greater visibility, understanding, and trust (when justified) in digital tools that have strong promise to promote patient health, we recommend five key strategies, including: (1) educating patients on the limitations of existing data protections; (2) conducting targeted research, including forensic analyses, into secondary data exchanges and to identify privacy breaches or reidentification risks; (3) enacting regulations that mandate greater transparency in health data transactions; (4) implementing computational mechanisms, such as distributed ledger technologies, to enhance data traceability and auditability; and (5) adopting dynamic consent models that allow patients to continuously manage and update their consent preferences. </sec>

  PublisherDOI

• Measuring Dependencies between Biological Signals with Self-supervision, and its Limitations

  ArXiv.org · 2025-07-29

  preprintOpen accessSenior author

  Measuring the statistical dependence between observed signals is a primary tool for scientific discovery. However, biological systems often exhibit complex non-linear interactions that currently cannot be captured without a priori knowledge regarding the nature of dependence. We introduce a self-supervised approach, concurrence, which is inspired by the observation that if two signals are dependent, then one should be able to distinguish between temporally aligned vs. misaligned segments extracted from them. Experiments with fMRI, physiological and behavioral signals show that, to our knowledge, concurrence is the first approach that can expose relationships across such a wide spectrum of signals and extract scientifically relevant differences without ad-hoc parameter tuning or reliance on a priori information, providing a potent tool for scientific discoveries across fields. However, dependencies caused by extraneous factors remain an open problem, thus researchers should validate that exposed relationships truly pertain to the question(s) of interest.

  PublisherOA PDFDOI

• Bitbox: Behavioral Imaging Toolbox for Computational Analysis of Behavior from Videos

  ArXiv.org · 2025-12-19

  articleOpen accessSenior author

  Computational measurement of human behavior from video has recently become feasible due to major advances in AI. These advances now enable granular and precise quantification of facial expression, head movement, body action, and other behavioral modalities and are increasingly used in psychology, psychiatry, neuroscience, and mental health research. However, mainstream adoption remains slow. Most existing methods and software are developed for engineering audiences, require specialized software stacks, and fail to provide behavioral measurements at a level directly useful for hypothesis-driven research. As a result, there is a large barrier to entry for researchers who wish to use modern, AI-based tools in their work. We introduce Bitbox, an open-source toolkit designed to remove this barrier and make advanced computational analysis directly usable by behavioral scientists and clinical researchers. Bitbox is guided by principles of reproducibility, modularity, and interpretability. It provides a standardized interface for extracting high-level behavioral measurements from video, leveraging multiple face, head, and body processors. The core modules have been tested and validated on clinical samples and are designed so that new measures can be added with minimal effort. Bitbox is intended to serve both sides of the translational gap. It gives behavioral researchers access to robust, high-level behavioral metrics without requiring engineering expertise, and it provides computer scientists a practical mechanism for disseminating methods to domains where their impact is most needed. We expect that Bitbox will accelerate integration of computational behavioral measurement into behavioral, clinical, and mental health research. Bitbox has been designed from the beginning as a community-driven effort that will evolve through contributions from both method developers and domain scientists.

  PublisherOA PDF

• Concurrence: A dependence criterion for time series, applied to biological data

  arXiv (Cornell University) · 2025-12-17

  preprintOpen accessSenior author

  Measuring the statistical dependence between observed signals is a primary tool for scientific discovery. However, biological systems often exhibit complex non-linear interactions that currently cannot be captured without a priori knowledge or large datasets. We introduce a criterion for dependence, whereby two time series are deemed dependent if one can construct a classifier that distinguishes between temporally aligned vs. misaligned segments extracted from them. We show that this criterion, concurrence, is theoretically linked with dependence, and can become a standard approach for scientific analyses across disciplines, as it can expose relationships across a wide spectrum of signals (fMRI, physiological and behavioral data) without ad-hoc parameter tuning or large amounts of data.

  PublisherOA PDFDOI

• A Transdiagnostic AI-Based Measure of Interpersonal Coordination in Autism and Other Conditions

  2025-10-31

  articleOpen access

  Background: Interpersonal coordination is a fundamental social behavior that has been shown to be reduced in autism, though less is known about other psychiatric conditions. An automated quantitative measure of interpersonal coordination would enhance assessment, diagnosis, and monitoring of treatment-related change in autism and other psychiatric conditions. We introduce and apply a novel AI-based measure (’concurrence’) to quantify and compare nonverbal interpersonal coordination during naturalistic conversation in individuals with and without various psychiatric presentations. Methods: The primary analysis included 380 12–18-year-olds with neurotypical development (NT), autism (AUT), or other psychiatric conditions (PSY), recorded during videoconference get-to-know-you conversations with a research staff member (‘partner’). Replication analyses included 72 12–18-year-olds with NT or AUT, recorded during face-to-face conversations. A self-supervised AI method (concurrence) was applied to time series data representing facial expressions and head movements of participants and their conversation partners. This yielded interpersonal coordination scores for all participant-partner dyads, which were then compared transdiagnostically. Convergent and discriminant validity were assessed using annotated subsamples from a combined sample of 609 5–52-year-olds. Convergent validity was assessed with measures of social gaze, motor imitation ability, and conversation quality; discriminant validity was assessed with IQ scores. Results: In the videoconference sample, AUT demonstrated significantly lower interpersonal coordination than PSY (unadjusted Cohen’s d=0.46, p&amp;lt;0.001) and NT (d=1.03, p&amp;lt;0.001), with PSY also lower than NT (d=0.50, p&amp;lt;0.001). The AUT&amp;lt;NT effect was replicated in the face-to-face sample (d=0.73, p&amp;lt;0.05). The group-by-context interaction was nonsignificant (p=0.33), suggesting group differences are robust to recording context. Convergent and discriminant validity was demonstrated through positive associations between interpersonal coordination and mutual social gaze (r(108)=0.46, p&amp;lt;0.0001), gross motor imitation ability (r(35)=0.41, p&amp;lt;0.05), and conversation quality ratings (r(364)=0.34, p&amp;lt;0.0001), but not IQ (r(367)=0.03, p=0.55).Limitations: Generalizability is limited by sample characteristics including cognitive and verbal ability, age, and sex.Conclusions: The study demonstrates reduced interpersonal coordination in adolescents with autism and other psychiatric conditions (AUT&amp;lt;PSY&amp;lt;NT) using a novel, transdiagnostic computational measure. This finding replicated across different conversational contexts and the concurrence measure demonstrated convergent and discriminant validity, highlighting its potential as an automated and scalable individual-level measure of social skills.

  PublisherDOI

Frequent coauthors

• Robert T. Schultz

  University of Pennsylvania

  124 shared

• Evangelos Sarıyanidi

  58 shared

• Casey Zampella

  Children's Hospital of Philadelphia

  54 shared

• John D. Herrington

  Children's Hospital of Philadelphia

  47 shared

• Kai Wang

  34 shared

• Ragini Verma

  University of Pennsylvania

  31 shared

• Beth A. Malow

  Vanderbilt University Medical Center

  20 shared

• Drew Parker

  University of Pennsylvania

  19 shared

Education

• PhD, Computational Science & Engineering, Informatics Institute

  Istanbul Technical University

  2012