Two-Dimensional Image-Based Screening Tool for Infants with Positional Cranial Deformities: A Machine Learning Approach

Incidence of Nonsynostotic Plagiocephaly and Developmental Disorders

Importance

This is the first population-based study quantifying the incidence of nonsynostotic positional plagiocephaly and/or brachycephaly (PPB) in infancy and its association with developmental disorders.

Objective

To report the incidence of PPB before age 1 year, the incidence of craniosynostosis, and the percentage of children with PPB diagnosed with a developmental disorder by age 7 years.

Design, Setting, and Participants

This was a retrospective, population-based cohort study of children in the Rochester Epidemiology Project (REP) born in Olmsted County, Minnesota, from January 1, 2008, through December 31, 2012, with follow-up through age 7 years. Data were analyzed from March 2021 to April 2024.

Exposure

Physical examination detecting cranial deformity.

Main Outcomes and Measures

The primary outcome was the incidence of PPB. Secondary outcomes were the incidence of craniosynostosis and the percentage of children with PPB diagnosed with a developmental disorder by age 7 years.

Results

Of 9909 infants (5084 [51.3%] male; 9205 [92.9%] born at term and 704 [7.1%] born preterm) included in the study, 575 had PPB, for a PPB incidence of 5.8% (95% CI, 5.3%-6.3%). The incidence of PPB was 5.3% (95% CI, 4.8%-5.8%) in term infants vs 11.8% (95% CI, 9.4%-14.6%) in preterm infants. The incidence of craniosynostosis was 0.16% (95% CI, 0.09%-0.26%). A developmental disorder was known or suspected in 4.2% (95% CI, 2.7%-6.2%) of infants at the time of PPB diagnosis; among 402 infants with PPB and follow-up through age 7 years, 30 (7.5%; 95% CI, 5.0%-10.7%) had a confirmed developmental disorder by 7 years of age. The prevalence of autism spectrum disorder (ASD) in children with a history of PPB who were followed up to age 7 years was 2.2% (9 of 402 children).

Conclusions and Relevance

This study found that only a small percentage of the infants had positional head deformity significant enough to be documented and/or referred for subspecialty evaluation, and only a small subset of these children went on to have a developmental disorder in childhood. This information is helpful for counseling families about their child's developmental risk at time of PPB diagnosis.

Predictive Value of Sonographic Parameters on the Effects of Cranial Molding Helmet Therapy in Infants with Positional Plagiocephaly

Positional plagiocephaly is a deformational cranial flattening frequently treated in pediatric neurosurgical practice. Positional maneuvers and orthotic helmet therapy are preferred therapeutic options for moderate-to-severe forms. Treatment response seems to be age-dependent. Nevertheless, predictive data are vague, and cost-efficiency might be a limiting factor for treatment. The purpose of this study was to investigate the early predictive value of sonographic parameters on the efficacy of orthotic helmet therapy through the assessment of changes in skull shape and correlation of the parameters with caliper cephalometry values and with age. A consecutive cohort of 49 patients < 10 months of age, undergoing orthotic helmet therapy for positional plagiocephaly, was recruited prospectively. The authors routinely assessed the patency of the lambdoid sutures by ultrasound and the following additional skull parameters were measured: suture width, adjacent full bone thickness, adjacent cortical bone thickness and occipital angle. Caliper cephalometric values, as well as demographic and clinical data were collected. Retrospective data analysis showed an inverse relation between both cortical and full skull bone thickness and early treatment efficacy, defined by a reduction in the occipital angle. The improvement of sonographic parameters correlated with the development of cranial caliper cephalometry values. In conclusion, the sonographic assessment of skull bone thickness is a safe and cost-effective tool to predict the early efficacy of orthotic helmet therapy in positional plagiocephaly and might, therefore, help the clinician to foresee the potential evolution of the deformity.

Artificial Intelligence Methods for the Argenta Classification of Deformational Plagiocephaly to Predict Severity and Treatment Recommendation

INTRODUCTION

Deformational plagiocephaly (DP) can be classified into 5 severity types using the Argenta scale (AS). Patients with type III or higher require referral to craniofacial surgery for management. Primary care pediatricians (PCPs) are often the first to encounter patients with DP, but current screening methods are subjective, increasing the risk of bias, especially for clinicians with little exposure to this population. The authors propose the use of artificial intelligence (AI) to classify patients with DP using the AS and to make recommendations for referral to craniofacial surgery.

METHODS

Vertex photographs were obtained for patients diagnosed with unilateral DP from 2019 to 2020. Using the photographs, an AI program was created to characterize the head contour of these infants into 3 groups based on the AS. The program was trained using photographs from patients whose DP severity was confirmed clinically by craniofacial surgeons. To assess the accuracy of the software, the AS predicted by the program was compared with the clinical diagnosis.

RESULTS

Nineteen patients were assessed by the AI software. All 3 patients with type I DP were correctly classified by the program (100%). In addition, 4 patients with type II were correctly identified (67%), and 7 were correctly classified as type III or greater (70%).

CONCLUSIONS

Using vertex photographs and AI, the authors were able to objectively classify patients with DP based on the AS. If converted into a smartphone application, the program could be helpful to PCPs in remote or low-resource settings, allowing them to objectively determine which patients require referral to craniofacial surgery.

The Use of Artificial Intelligence for the Classification of Craniofacial Deformities

Positional cranial deformities are a common finding in toddlers, yet differentiation from craniosynostosis can be challenging. The aim of this study was to train convolutional neural networks (CNNs) to classify craniofacial deformities based on 2D images generated using photogrammetry as a radiation-free imaging technique. A total of 487 patients with photogrammetry scans were included in this retrospective cohort study: children with craniosynostosis (n = 227), positional deformities (n = 206), and healthy children (n = 54). Three two-dimensional images were extracted from each photogrammetry scan. The datasets were divided into training, validation, and test sets. During the training, fine-tuned ResNet-152s were utilized. The performance was quantified using tenfold cross-validation. For the detection of craniosynostosis, sensitivity was at 0.94 with a specificity of 0.85. Regarding the differentiation of the five existing classes (trigonocephaly, scaphocephaly, positional plagiocephaly left, positional plagiocephaly right, and healthy), sensitivity ranged from 0.45 (positional plagiocephaly left) to 0.95 (scaphocephaly) and specificity ranged from 0.87 (positional plagiocephaly right) to 0.97 (scaphocephaly). We present a CNN-based approach to classify craniofacial deformities on two-dimensional images with promising results. A larger dataset would be required to identify rarer forms of craniosynostosis as well. The chosen 2D approach enables future applications for digital cameras or smartphones.

Artificial Intelligence Applications in Diagnosing and Managing Non-syndromic Craniosynostosis: A Comprehensive Review

Craniosynostosis is characterised by the premature fusion of one or more cranial sutures, resulting in an abnormal head shape. The management of craniosynostosis requires early diagnosis, surgical intervention, and long-term monitoring. With the advancements in artificial intelligence (AI) technologies, there is great potential for AI to assist in various aspects of managing craniosynostosis. The main aim of this article is to review available literature describing the current uses of AI in craniosynostosis. The main applications highlighted include diagnosis, surgical planning, and outcome prediction. Many studies have demonstrated the accuracy of AI in differentiating subtypes of craniosynostosis using machine learning (ML) algorithms to classify craniosynostosis based on simple photographs. This demonstrates its potential to be used as a screening tool and may allow patients to monitor disease progression reducing the need for CT scanning. ML algorithms can also analyse CT scans to aid in the accurate and efficient diagnosis of craniosynostosis, particularly when training junior surgeons. However, the lack of sufficient data currently limits this clinical application. Virtual surgical planning for cranial vault remodelling using prefabricated cutting guides has been shown to allow more precise reconstruction by minimising the subjectivity of the clinicians' assessment. This was particularly beneficial in reducing operating length and preventing the need for blood transfusions. Despite the potential benefits, there are numerous challenges associated with implementing AI in craniosynostosis. The integration of AI in craniosynostosis holds significant promise for improving the management of craniosynostosis. Further collaboration between clinicians, researchers, and AI experts is necessary to harness its full potential.

Denominator of Cranial Vault Asymmetry Index: Choosing Between Longer and Shorter Diagonal Lengths

Since it was proposed in this journal in 2001, the cranial vault asymmetry index (CVAI) has been an important parameter for assessing cranial shape. However, different publications currently use different variables in the denominator of the CVAI formula. We thus investigated the use of long and short diagonal lengths as variables in the denominator of the CVAI formula. We searched the databases of PubMed, Google Scholar, and Scopus for articles published between 2016 and 2022 that cited the original work article of CVAI. Articles were included if they were written in English and if the denominator of the CVAI formula was specified. For multiple articles by the same author, only the most recent article was included. In total, 30 articles were included; 10 articles used the longer diagonal length as the denominator and 20 articles used the shorter diagonal length. No uniform trend was observed by a country or journal of publication. Application of the CVAI formula using different denominators yielded interchangeable results, and the resulting values had only negligible differences clinically. However, it would be necessary to create a standard formula for using the CVAI as a parameter for reporting cranial shape assessments consistently.

Smartphone Integration of Artificial Intelligence for Automated Plagiocephaly Diagnosis

Positional plagiocephaly is a pediatric condition with important cosmetic implications affecting ∼40% of infants under 12 months of age. Early diagnosis and treatment initiation is imperative in achieving satisfactory outcomes; improved diagnostic modalities are needed to support this goal. This study aimed to determine whether a smartphone-based artificial intelligence tool could diagnose positional plagiocephaly.

Methods

A prospective validation study was conducted at a large tertiary care center with two recruitment sites: (1) newborn nursery, (2) pediatric craniofacial surgery clinic. Eligible children were aged 0-12 months with no history of hydrocephalus, intracranial tumors, intracranial hemorrhage, intracranial hardware, or prior craniofacial surgery. Successful artificial intelligence diagnosis required identification of the presence and severity of positional plagiocephaly.

Results

A total of 89 infants were prospectively enrolled from the craniofacial surgery clinic (n = 25, 17 male infants [68%], eight female infants [32%], mean age 8.44 months) and newborn nursery (n = 64, 29 male infants [45%], 25 female infants [39%], mean age 0 months). The model obtained a diagnostic accuracy of 85.39% compared with a standard clinical examination with a disease prevalence of 48%. Sensitivity was 87.50% [95% CI, 75.94-98.42] with a specificity of 83.67% [95% CI, 72.35-94.99]. Precision was 81.40%, while likelihood ratios (positive and negative) were 5.36 and 0.15, respectively. The F1-score was 84.34%.

Conclusions

The smartphone-based artificial intelligence algorithm accurately diagnosed positional plagiocephaly in a clinical environment. This technology may provide value by helping guide specialist consultation and enabling longitudinal quantitative monitoring of cranial shape.

Machine Learning-Driven Clinical Image Analysis to Identify Craniosynostosis: A Pilot Study of Telemedicine and Clinic Patients

BACKGROUND

The authors have developed pretrained machine learning (ML) models to evaluate neonatal head shape deformities using top-down and facial orthogonal photographs of the patient's head. In previous preliminary analysis, this approach was tested with images from an open-source data bank.

OBJECTIVE

To determine the accuracy of pretrained ML models in identifying craniosynostosis among patients seen in our outpatient neurosurgery clinic.

METHODS

We retrospectively reviewed top-down and facial orthogonal images of each patient's head and provider clinical diagnosis from the same encounters. Head shape classifications generated from 3 pretrained ML models (random forest, classification and regression tree, and linear discriminant analysis) were applied to each patient's photograph data set after craniometric extraction using a predefined image processing algorithm. Diagnoses were codified into a binary scheme of craniosynostosis vs noncraniosynostosis. Sensitivity, specificity, and Matthew correlation coefficient were calculated for software vs provider classifications.

RESULTS

A total of 174 patients seen for abnormal head shape between May 2020 and February 2021 were included in the analysis. One hundred seven patients (61%) were seen in-person and 67 (39%) through telemedicine. Twenty-three patients (13%) were diagnosed with craniosynostosis. The best-performing model identified craniosynostosis with an accuracy of 94.8% (95% CI 90.4-97.6), sensitivity of 87.0% (95% CI 66.4-97.2), specificity of 96.0% (95% CI 91.6-98.5), and Matthew correlation coefficient of 0.788 (95% CI 0.725-0.839).

CONCLUSION

Machine learning-driven image analysis represents a promising strategy for the identification of craniosynostosis in a real-world practice setting. This approach has potential to reduce the need for imaging and facilitate referral by primary care providers.

Novel Screening and Monitoring Techniques for Deformational Plagiocephaly: A Systematic Review

This article summarizes the current state of diagnostic modalities for infant craniofacial deformities and highlights capable diagnostic tools available currently to pediatricians.