Related Experiment Video
Updated: Jun 5, 2025
Quantified Assessment of Infant's Gross Motor Abilities Using a Multisensor Wearable
Published on: May 17, 2024
Deep-learning-based markerless tracking of distal anatomical landmarks in clinically recorded videos for assessing
Hamid Abbasi1, Sarah R Mollet1, Sian A Williams2,3
1Auckland Bioengineering Institute (ABI), University of Auckland, Auckland, New Zealand, New Zealand.
Insights
Automated infant movement analysis using deep learning can now track distal limb rotations, crucial for early detection of neurodevelopmental disorders like cerebral palsy.
Area of Science:
- Biomedical Engineering
- Developmental Neuroscience
- Medical Imaging
Background:
- Abnormal General Movements (GMs) in infants are key indicators of neurodevelopmental disorders, such as cerebral palsy.
- Automated General Movements Assessments (GMA) platforms can enhance early identification of at-risk infants.
- Previous deep learning models accurately tracked limb longitudinal axes but missed crucial rotational data.
Purpose of the Study:
- To develop and validate a deep learning model capable of accurately capturing distal limb rotational movements in infants.
- To improve the accuracy of automated General Movements Assessments (GMA) by incorporating rotational data.
- To enhance the early detection of neurodevelopmental disorders through advanced infant pose reconstruction.
Main Methods:
- Utilized a ResNet-152 deep neural network model.
- Trained the model on 26 diverse videos combining laboratory and clinical data.
- Focused on improving accuracy for distal limb landmarks, which were previously challenging to track.
Main Results:
- Achieved >85% accuracy for distal limb landmark tracking.
- Attained an overall accuracy of 98.28% (SD 2.29) across 24 landmarks.
- Demonstrated that increased data diversity and sample size improved performance on challenging distal markers.
Conclusions:
- The developed deep learning scheme effectively captures clinically relevant infant rotational movements.
- This approach forms a foundation for robust infant pose reconstruction for automated GMA.
- Improved automated GMA holds significant potential for earlier and more accurate screening of neurodevelopmental disorders.
Abstract:
Abnormal patterns in infants' General Movements (GMs) are robust clinical indicators for the progression of neurodevelopmental disorders, including cerebral palsy. Availability of automated platforms for General Movements Assessments (GMA) could improve screening rate and allow identifying at-risk infants. While we have previously shown that deep-learning schemes can accurately track the longitudinal axes of infant limb movements (12 anatomical locations, 3 per limb), information about the distal limb segments' rotational movements is important for making an accurate clinical assessment, but has not previously been captured. Here we show that training schemes that are highly successful at tracking trunk and proximal limb landmarks perform less well for the distal limb landmarks, and this problem is exacerbated when landmarks are more precisely defined in the training-set to capture rotational movements. Increasing the sample size to 26 videos using a mixture of laboratory and clinical data pre-selected for diversity of pose and video conditions in a ResNet-152 deep-net model was sufficient to permit accuracy of >85% for the distal markers, and overall accuracy of 98.28% (SD 2.29) across the 24 landmarks. This scheme is suitable to form the basis of an infant pose reconstruction algorithm that captures clinically relevant information for an automated GMA.
