TheraTrack: Digital Therapeutic Tracking Platforms

• TheraTrack is a suite of digital platforms and methodologies that integrate multi-modal data to support clinical decision-making and patient feedback.
• It employs therapist-facing dashboards, wearable tracking, and XR systems to deliver real-time analytics and AI-powered summarization.
• Key applications include mental health homework review, rehabilitation tracking, and closed-loop feedback that improve therapy adherence and clinical insights.

TheraTrack refers to a set of digital, algorithmic, and data-driven platforms and methodologies supporting tracking and feedback in therapeutic contexts, spanning clinical mental-health homework, subjective experience logging, and kinematic monitoring for rehabilitation. Modern implementations operate as therapist-facing web platforms with AI summarization, wearable-based behavioral tracking systems, or closed-loop extended reality (XR) exercise environments. Central to all instantiations is the integration of heterogeneous data and the provision of structured, traceable insights to support clinical decision-making, adherence promotion, and patient self-awareness.

1. System Architectures

TheraTrack encompasses varied system architectures adapted for their clinical application domain.

In the context of therapist-facing homework support platforms, TheraTrack is a web-based dashboard which ingests, stores, and synthesizes multi-modal patient-generated health data (PGHD) within a canonical JSON schema. Data sources include free-text journal entries, structured assessments (PHQ-9, GAD-7), and device exports (e.g., actigraphy summaries, heart rate), all indexed for metadata-driven filtering and traceability. The platform’s frontend supports drag-and-drop UI customization and widget configuration, responding to an onboarding survey capturing therapist requirements for visualization, summary depth, and data feeds. The backend orchestrates a LLM pipeline (e.g., via Azure GPT-4o with temperature 0.7, zero-shot) for natural language querying and summary generation, integrating context retrieval, prompt construction, and intent classification (Wang et al., 26 Jan 2026).

Wearable-based TheraTrack implementations consist of a compact, single-button wristband housing a mechanical push mechanism with debounce circuitry, a Bluetooth Low Energy SoC, and minimalistic firmware for robust event capture. Event data are transmitted in real time to a companion smartphone application over BLE, with cloud synchronization handled by a REST API backend (Node.js/Express or Flask). End-to-end, the architecture enables time-stamped, audit-logged, and analyzable symptom registration for integration into therapy (Larsen et al., 2017).

In XR rehabilitation applications, TheraTrack is conceptualized as a telehealth-plus-XR system with markerless depth capture, real-time kinematic analysis, and synchronous or asynchronous remote feedback. The architecture fuses skeleton tracking (Microsoft Kinect v2/OpenNI/MediaPipe), real-time analytics, WebGL rendering (three.js or Unity3D), and a low-latency networking stack (WebRTC/Socket.io) to stream joint angle data, animate avatars, and deliver immediate feedback for patient and therapist (Conroy et al., 19 Feb 2025).

2. Core Tracking Methodologies

TheraTrack systems rely on diverse methodological foundations, reflecting their context.

In therapist-facing homework review tools, data integration, LLM-based summarization, and traceable evidence mapping are fundamental. The multi-stage intent classification and context retrieval pipeline supports both structural AI summaries (with rigid section headers and evidence anchors) and natural-language query handling. Verification is facilitated by explicit pointers back to underlying JSON artifacts, with every AI-generated assertion mapped to its data provenance, and user interaction enabling drill-down to raw entries (Wang et al., 26 Jan 2026).

For one-button experience tracking, the methodology is centered around precise, high-fidelity symptom logging. The detection protocol requires two rapid button presses (within 500 ms) to mark a valid registration, minimizing false positives. Data synchronization routines periodically upload buffered events to a central server. Statistical aggregation of these events (mean daily count, variance, Poisson modeling) facilitates retrospective clinical analytics. Protocol adherence is monitored through event timestamp distributions visualized by time-of-day and day-of-week (Larsen et al., 2017).

In the XR and rehabilitation setting, TheraTrack includes pipelines for markerless skeletal tracking and kinematic computation. Depth-sensor data is processed to extract joint coordinates, with smoothing (e.g., 4 Hz Butterworth filtering) applied to mitigate jitter. Joint angles $\theta(t)$, angular velocities $\omega(t)$, and trajectory error metrics (RMSE and MAD) are computed in real time. Feedback mechanisms leverage threshold-based visual overlays and audio prompts, with feedback latency maintained below 100 ms to preserve closed-loop interactivity (Conroy et al., 19 Feb 2025).

3. Data Processing, Summarization, and Visualization

Data processing pipelines are tailored for transparency, interpretability, and clinical validation.

In the therapist dashboard, raw PGHD artifacts are processed into canonical forms and indexed by timestamp, type, and therapist-focused metadata. Data presentation leverages widget-based modularity: bar charts of homework durations, color-coded quality or mood ratings, timeline trackers for standardized assessment instruments, and trend visualizations (sparklines, streak counters). Summarization employs rigid prompt-engineered LLM flows, enforcing required reporting order of section headers and "No data" compliance for missing values. All generated summaries are furnished with evidence anchors, enabling backward navigation for clinical verification (Wang et al., 26 Jan 2026).

For wearable event data, aggregation at multiple temporal granularities enables line charts (cumulative press events over days), box plots (daily distributions), and heatmaps (event density across hours and weekdays). This multi-resolutional analytic approach allows therapists to overlay event density with clinical notes, facilitating context-sensitive hypothesis generation and intervention planning (Larsen et al., 2017).

In XR tracking, data from 3D joint time series is reported via real-time numerical streams, summary graphics, and qualitative overlays (limb highlighting for spatial error). Statistical analyses—ranging from Shapiro–Wilk normality tests to paired t-tests and Wilcoxon tests on kinematic outcomes—quantify performance variations. Feedback and adherence metrics (percent time within performance band) are visualized gamified progress bars to drive patient engagement (Conroy et al., 19 Feb 2025).

4. Clinical and User Impact

Pilot studies with therapist-facing TheraTrack systems demonstrate quantifiable reductions in cognitive load, improvement in information access efficiency, and positive usability ratings. Qualitative analyses identify cognitive off-loading, session-preparation acceleration, and dynamic verification behaviors as recurrent usage patterns. Experienced therapists primarily appropriate LLM summaries for private review, while novices leverage these tools as confidence-enhancing cross-checks. Trust calibration is observed as therapists exploit evidence anchors for explanatory debugging, especially when initial skepticism is present. Appropriation is adaptive: deep summary exploration precedes sessions, while client-facing views are selectively filtered during consultation. Customization desiderata include support for modality-specific assessments and reconfigurability as therapy phases evolve. Practitioners emphasize the criticality of ethical safeguards: explicit human oversight, clear AI suggestions labels, strict HIPAA-compliance, and granular consent for sensitive data (Wang et al., 26 Jan 2026).

Active experience self-tracking via single-button wearables yields granular, temporally resolved symptom data with direct clinical integration. In a PTSD case study, high-density event records permitted disambiguation of symptom clusters, enabled targeted behavioral experiments, and resulted in stabilization of hyperarousal event rates. Patient-reported sense of agency and therapist-perceived insight increased relative to baseline. Nonetheless, generalizability remains limited due to N=1 evidence and platforms without passive physiological collection (Larsen et al., 2017).

XR-enabled home exercise tracking, as implemented analogously in proof-of-concept platforms, exhibits strong patient and therapist preference for real-time feedback, with 70–100% reporting perceived utility for adherence and communication. Quantitative movement accuracy metrics (RMSE < 5° in calibration) affirm technical feasibility, though some angular velocity reductions are noted compared to paper-based protocols (Conroy et al., 19 Feb 2025).

5. Implementation Guidelines and Technical Recommendations

TheraTrack systems benefit from explicit modularity and adherence to best practices in implementation.

Therapist dashboards require robust data import pipelines, flexible widget frameworks, dynamic configuration interfaces (for sources and summary depth), and programmable LLM prompt templates restricting hallucination and supporting traceability. Evidence linking uses index-based token-source mappings, leveraging frameworks such as LangChain’s “source_document” metadata features (Wang et al., 26 Jan 2026).

Wearable devices should integrate mechanical debounce, BLE communication protocols with fail-safe local event storage, and scheduled upload routines. Battery and quiescent current budgeting must follow precise estimation formulas (e.g., $T_{\text{batt}}$ as documented for coin cells), and user protocols prioritize in-situ event marking over retrospective recall.

XR-based rehabilitation platforms should incorporate anatomical calibration routines (e.g., T-pose sequencing), sensor-to-world alignment algorithms, algorithmic filtering for joint smoothing, and robust statistical validation against gold-standard motion capture. Pipeline latency should be maintained below 100 ms, with UX design incorporating dead-band feedback and progressive disclosure for performance metrics. Hardware choices (depth sensors, GPU class) and software stack recommendations (OpenNI, Unity3D, Python/JavaScript analytics) must be closely followed, with future integration of head-mounted displays and IMUs for fidelity (Conroy et al., 19 Feb 2025).

6. Limitations and Future Directions

TheraTrack implementations are constrained by several technical, clinical, and scalability challenges.

Limitations observed include single-participant or short-duration pilot studies, incomplete sensor integration (e.g., absence of passive physiological markers), data loss due to software bugs (week 2 BLE event gap), and unexamined effects on longitudinal outcomes (Larsen et al., 2017, Wang et al., 26 Jan 2026). In the mental-health context, further work is required for deeper in-session workflow integration, real-time intervention triggers, and broader N-of-1 and randomized studies.

AI-augmented dashboards must address biases in summarization, dynamic adaptation to evolving clinical goals, and persistent concerns over privacy and ethical use, particularly with free-text and sensitive content storage (Wang et al., 26 Jan 2026).

For XR-based home exercise platforms, technical reliability of markerless tracking, calibration stability, and infrastructure requirements remain ongoing areas for enhancement. Future work suggests expansion to multi-event or multi-symptom tracking, haptic or immersive feedback, integration of passive wearable sensors, and deployment at population scale with real-world adherence/outcome impact assessment (Conroy et al., 19 Feb 2025).

---

TheraTrack thus denotes a multi-domain family of clinical tracking and feedback tools characterized by rigorous data integration, modular analytics, and evidence-traceable AI summaries, each adapted to the high-precision requirements of mental health, behavioral self-tracking, and physical rehabilitation. Further field validation and technical diversification are needed to generalize and sustain clinical impact.