AI-Powered Overhead Squat Assessor

Goal

The primary goal of this project was to develop an intelligent assistant for health and fitness professionals. I aimed to create a tool that bridges the gap between qualitative movement screening and quantitative, evidence-based program design. By leveraging AI, the application streamlines the assessment process, reduces administrative time, and empowers clinicians to create highly personalized corrective exercise plans more efficiently.

Reason for Creation

The Overhead Squat Assessment (OHSA) is a cornerstone of movement screening, renowned for its ability to reveal kinetic chain dysfunctions. However, the process of interpreting the visual data, cross-referencing it with potential muscle imbalances, and then formulating a comprehensive corrective plan is both complex and time-consuming. Clinicians often rely on mental heuristics or manual lookups, which can be prone to subjectivity and inconsistency.

I saw an opportunity to automate the heavy lifting of this analysis. I wanted to build a tool that could take raw assessment data and instantly provide a structured, data-driven hypothesis of muscle imbalances, suggest confirmatory tests, and generate a robust starting point for a corrective program—all while keeping the clinician firmly in control.

How It Works

1. Initial Assessment: The clinician inputs observational data by selecting checkboxes for common movement compensations seen during the squat (e.g., "Knees Bow In," "Arms Fall Forward"). They can supplement this with objective goniometric (range of motion) measurements and a checklist of the client's available exercise equipment.
2. AI Analysis & Hypothesis: Upon submission, the Gemini API analyzes the combined data. It generates a report detailing likely overactive (tight) and underactive (weak)muscles, providing a clinical rationale for each finding. Crucially, it also suggests a prioritized list of follow-up tests (e.g., Thomas Test, Glute Medius Strength Test) needed to confirm its hypotheses.
3. Dynamic Follow-up: The application generates a dynamic form containing only the AI-suggested tests. This keeps the assessment focused and efficient. The clinician retains full clinical autonomy with an "Add/Remove Tests" feature to modify the test battery as they see fit.
4. Personalized Plan Generation: After the clinician enters the results of the follow-up tests, the AI synthesizes all collected data—initial observations, goniometry, test results, and available equipment—to create a comprehensive corrective exercise plan. The plan is structured into two parts: "Release & Stretch" for overactive muscles and "Activate & Strengthen" for underactive ones.
5. Customization & Export: The final plan is fully interactive. The clinician can click a "Suggest Alternatives" (✨) button on any exercise to get AI-generated replacements that target the same muscle, tailored to the client's equipment. Once finalized, the assessment data and exercise plan can be exported as a professional, client-ready PDF report or formatted for email.

Future Development

This tool is currently in its alpha stage. The long-term vision is to integrate it into a larger suite of clinical tools under the banner of kinesiologytools.com. Future enhancements may include user accounts for tracking client progress, integration of computer vision for automated video analysis, and an expanded library of clinical assessments.