GenGait: A Transformer-Based Model for Human Gait Anomaly Detection and Normative Twin Generation

Dimensions overall score 7.0

• GenGait uses a Transformer masked autoencoder trained exclusively on normative gait sequences from 150 adults for anomaly detection.method 95%
  Evidencepartial

  This work proposes a label-free framework for joint-level anomaly detection and kinematic correction based on a Transformer masked autoencoder trained exclusively on normative gait sequences from 150 adults

  Implicationpartial

  Directly stated in the abstract with specific details about the model architecture and training data.

  Verificationpartial

  partial

• The model performs anomaly detection by estimating joint inconsistency scores through occlusion of individual joints and measuring deviations from learned normative patterns.method 90%
  Evidencepartial

  first it estimates joint inconsistency scores by occluding individual joints and measuring deviations from the learned normative prior

  Implicationpartial

  Explicitly described in the abstract as part of the inference procedure.

  Verificationpartial

  partial

• GenGait can generate corrected kinematic trajectories at flagged joint positions by reconstructing the full skeleton from remaining spatiotemporal context.method 90%
  Evidencepartial

  it withholds the flagged joints from the encoder input and reconstructs the full skeleton from the remaining spatiotemporal context, yielding corrected kinematic trajectories at the flagged positions

  Implicationpartial

  Directly stated in the abstract as the second step of the inference procedure.

  Verificationpartial

  partial

• Validation on 10 held-out normative participants mimicking seven simulated gait abnormalities showed accurate localization of biomechanically inconsistent joints.result 85%
  Evidencepartial

  Validation on 10 held-out normative participants, who mimicked seven simulated gait abnormalities, showed accurate localization of biomechanically inconsistent joints

  Implicationpartial

  Directly stated in the abstract with specific validation details.

  Verificationpartial

  partial

• The approach achieved significant reduction in angular deviation across all analyzed joints with large effect sizes.result 85%
  Evidencepartial

  a significant reduction in angular deviation across all analyzed joints with large effect sizes

  Implicationpartial

  Directly stated in the abstract with quantitative performance claims.

  Verificationpartial

  partial

• GenGait enables interpretable, subject-specific localization of gait impairments without requiring disease labels.technical 90%
  Evidencepartial

  The proposed approach enables interpretable, subject-specific localization of gait impairments without requiring disease labels

  Implicationpartial

  Explicitly stated as a key advantage of the approach in the abstract.

  Verificationpartial

  partial

• The framework preserves normative kinematics while correcting anomalies.result 80%
  Evidencepartial

  and preservation of normative kinematics

  Implicationpartial

  Directly stated in the abstract as part of the validation results.

  Verificationpartial

  partial

• The model uses markerless multi-camera motion-capture system data.technical 95%
  Evidencepartial

  acquired with a markerless multi-camera motion-capture system

  Implicationpartial

  Explicitly stated in the abstract as the data acquisition method.

  Verificationpartial

  partial

Competitive landscape

A label-free Transformer model that detects and corrects gait abnormalities by learning normative human movement patterns.