Stoltz, Petronella

OBJECTIVE Sagittal craniosynostosis is the most common form of craniosynostosis and typically results in scaphocephaly, which is characterized by biparietal narrowing, compensatory frontal bossing, and an occipital prominence. The cephalic index (CI) is a simple metric for quantifying the degree of cranial narrowing and is often used to diagnose sagittal craniosynostosis. However, patients with variant forms of sagittal craniosynostosis may present with a "normal" CI, depending on the part of the suture that is closed. As machine learning (ML) algorithms are developed to assist in the diagnosis of cranial deformities, metrics that reflect the other phenotypic features of sagittal craniosynostosis are needed. In this study the authors sought to describe the posterior arc angle (PAA), a measurement of biparietal narrowing that is obtained with 2D photographs, and elucidate the role of PAA as an adjuvant to the CI in characterizing scaphocephaly and the potential relevance of PAA in new ML model development. METHODS The authors retrospectively reviewed 1013 craniofacial patients treated during the period from 2006 to 2021. Orthogonal top-down photographs were used to calculate the CI and PAA. Distribution densities, receiver operating characteristic (ROC) curves, and chi-square analyses were used to describe the relative predictive utility of each method for sagittal craniosynostosis. RESULTS In total, 1001 patients underwent paired CI and PAA measurements and a clinical head shape diagnosis (sagittal craniosynostosis, n = 122; other cranial deformity, n = 565; normocephalic, n = 314). The area under the ROC curve (AUC) for the CI was 98.5% (95% confidence interval 97.8%–99.2%, p < 0.001), with an optimum specificity of 92.6% and sensitivity of 93.4%. The PAA had an AUC of 97.4% (95% confidence interval 96.0%–98.8%, p < 0.001) with an optimum specificity of 94.9% and sensitivity of 90.2%. In 6 of 122 (4.9%) cases of sagittal craniosynostosis, the PAA was abnormal while the CI was normal. This means that adding a PAA cutoff branch to a partition model increases the detection of sagittal craniosynostosis. CONCLUSIONS Both CI and PAA are excellent discriminators for sagittal craniosynostosis. Using an accuracy-optimized partition model, the addition of the PAA to the CI increased model sensitivity compared to using the CI alone. Using a model that incorporates both CI and PAA could assist in the early identification and treatment of sagittal craniosynostosis via automated and semiautomated algorithms that utilize tree-based ML models.

OBJECTIVE Quantitative measurements of trigonocephaly can be used to characterize and track this phenotype, which is associated with metopic craniosynostosis. Traditionally, trigonocephaly metrics were extracted from CT scans; however, this method exposes patients to ionizing radiation. Three-dimensional optical scans are another option but are not routinely available in most outpatient settings. Recently, the authors developed semiautomated artificial intelligence algorithms that extract craniometric data from orthogonal 2D photographs. Although 2D photographs are safe, inexpensive, and straightforward to obtain, the accuracy of photograph-based craniometrics in comparison to CT and 3D optical scan correlates has not been established. In this study the authors compared the classification power of 2D photograph–based metrics of trigonocephaly with four CT-based metrics and one 3D optical scan–based metric in a heterogeneous series of patients who presented to an outpatient craniofacial clinic. METHODS In this study the authors performed retrospective craniometric analyses of patient 2D photographs, 3D optical scans, and CT scans. Imaging-derived craniometrics include the 2D photograph–based anterior arc angle (AAA2D-photo), anterior-posterior ratio (APR2D-photo), and anterior-middle ratio (AMR2D-photo); the CT-based anterior arc angle (AAACT), metopic index (MICT), endocranial-bifrontal angle (eBFACT), and interfrontal angle (IFACT); and the 3D optical scan–based anterior arc angle (AAA3D-optical). Receiver operating characteristics (ROCs) were used to identify craniometrics strongly descriptive of trigonocephaly. Interrater comparisons were made between paired trigonocephaly measurements obtained from photographs and either CT scans or 3D optical scans. RESULTS There were 13 photograph-based and CT-based pairs and 22 paired measurements from 2D photographs and 3D optical scans. AAA displayed the strongest classification capacity across all three imaging modalities. Significant agreement was observed between AAACT and AAA2D-photo (intraclass correlation coefficient [ICC] = 0.68 [95% CI 0.24–0.89], p = 0.0035), and AAA3D-optical and AAA2D-photo (ICC = 0.70 [95% CI 0.41–0.87], p < 0.0001). There was no significant correlation between APR2D-photo or AMR2D-photo and conventional CT-based metrics describing longitudinal width ratios (MICT). CONCLUSIONS Photograph-based craniometrics are powerful tools that can be used to quantify the severity of trigonocephaly and exhibit high concordance with standard measurements derived from CT scans and 3D optical scans. The authors developed and freely share a research-use application to calculate trigonocephaly metrics from 2D photographs. Given the availability of digital photography, lack of ionizing radiation, and low cost of photograph-based craniometric derivation, this technique may be useful to supplement routine ambulatory care and objectively track outcomes following treatment.

OBJECTIVE The authors sought to evaluate the accuracy of a novel telehealth-compatible diagnostic software system for identifying craniosynostosis within a newborn (< 1 year old) population. Agreement with gold standard craniometric diagnostics was also assessed. METHODS Cranial shape classification software accuracy was compared to that of blinded craniofacial specialists using a data set of open-source (n = 40) and retrospectively collected newborn orthogonal top-down cranial images, with or without additional facial views (n = 339), culled between April 1, 2008, and February 29, 2020. Based on image quality, midface visibility, and visibility of the cranial equator, 351 image sets were deemed acceptable. Accuracy, sensitivity, and specificity were calculated for the software versus specialist classification. Software agreement with optical craniometrics was assessed with intraclass correlation coefficients. RESULTS The cranial shape classification software had an accuracy of 93.3% (95% CI 86.8–98.8; p < 0.001), with a sensitivity of 92.0% and specificity of 94.3%. Intraclass correlation coefficients for measurements of the cephalic index and cranial vault asymmetry index compared to optical measurements were 0.95 (95% CI 0.84–0.98; p < 0.001) and 0.67 (95% CI 0.24–0.88; p = 0.003), respectively. CONCLUSIONS These results support the use of image processing–based neonatal cranial deformity classification software for remote screening of nonsyndromic craniosynostosis in a newborn population and as a substitute for optical scanner– or CT-based craniometrics. This work has implications that suggest the potential for the development of software for a mobile platform that would allow for screening by telemedicine or in a primary care setting.