Personalized phenotype encoding and prediction of pathological head development from cross-sectional images
Related Experiment Video
Updated: Jan 9, 2026
Author Spotlight: Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
Connor Elkhill1, Ines A Cruz-Guerrero1, Jiawei Liu1
1Department of Biostatistics and Informatics, Colorado School of Public Health, Aurora, CO.
Insights
This study introduces a new deep learning model for predicting head development in children, even with existing pathologies. It enables personalized growth predictions using only current data, improving surgical planning.
Area of Science:
- Biomedical Engineering
- Medical Imaging
- Artificial Intelligence
Background:
- Accurate prediction of anatomical development is vital for pediatric surgery.
- Young patients' rapid growth and potential pathologies complicate anatomical change prediction.
Purpose of the Study:
- To present a novel deep learning architecture for personalized predictions of normative and pathologic head development.
- To enable accurate predictions using only cross-sectional data, overcoming the need for longitudinal data.
Main Methods:
- Developed a novel phenotype encoder using domain adversarial training for age- and sex-independent representations.
- Created a growth predictor generating head shape based on age, sex, and pathology.
- Trained the model on CT images and 3D photograms, evaluated on longitudinal data.
Main Results:
- Achieved high accuracy in head surface growth prediction (4.93 ± 2.29 mm for pathology, 4.61 ± 3.28 mm for normative).
- Demonstrated low volumetric error (0.16 ± 0.11 L for pathology, 0.27 ± 0.19 L for normative).
- Successfully created age- and sex-agnostic phenotype representations.
Conclusions:
- The proposed deep learning model enables personalized predictions of pathological head development without longitudinal data.
- This approach significantly advances the ability to plan pediatric surgical treatments by accurately forecasting anatomical changes.
- The method offers a breakthrough in creating patient-specific developmental predictions for both healthy and pathologically affected individuals.
Abstract:
Prediction of anatomical development plays a crucial role in the selection and planning of pediatric surgical treatments. However, the rapid growth of young patients and the potential involvement of pathology makes prediction of anatomical changes challenging. We present a novel deep learning architecture to make personalized predictions of normative and pathologic head development using only cross-sectional data. We designed a novel phenotype encoder that uses domain adversarial training to create age- and sex-independent representations of patient phenotypes and growth predictor that learns to generate the head shape of patients given the anatomical effects of age, sex and pathology. These modules cooperate to instantiate patient anatomies to any age, enabling personalized predictions conditioned to the presence of pathology. We trained our model using head segmentations from cross-sectional CT images and 3D photograms and evaluated model performance on an independent longitudinal dataset. The proposed model achieved a head surface growth prediction error of 4.93 ± 2.29 mm and a volumetric error 0.16 ± 0.11 L in patients with cranial pathology, and 4.61 ± 3.28 mm and 0.27 ± 0.19 L for normative subjects, demonstrating high accuracy. Our method is the first to create age- and sex-agnostic patient phenotype representations, and to enable personalized predictions of pathological development without requiring longitudinal data for training.
