Pathological estimation of tumor necrosis after chemotherapy is essential for patients with osteosarcoma. This study reports the first fully automated tool to assess viable and necrotic tumor in osteosarcoma, employing advances in histopathology digitization and automated learning. We selected 40 digitized whole slide images representing the heterogeneity of osteosarcoma and chemotherapy response. With the goal of labeling the diverse regions of the digitized tissue into viable tumor, necrotic tumor, and non-tumor, we trained 13 machine-learning models and selected the top performing one (a Support Vector Machine) based on reported accuracy. We also developed a deep-learning architecture and trained it on the same data set. We computed the receiver-operator characteristic for discrimination of non-tumor from tumor followed by conditional discrimination of necrotic from viable tumor and found our models performing exceptionally well. We then used the trained models to identify regions of interest on image-tiles generated from test whole slide images. The classification output is visualized as a tumor-prediction map, displaying the extent of viable and necrotic tumor in the slide image. Thus, we lay the foundation for a complete tumor assessment pipeline from original histology images to tumor-prediction map generation. The proposed pipeline can also be adopted for other types of tumor.

Metopic suture closure can manifest as a benign metopic ridge (BMR), a variant of normal, to “true” metopic craniosynostosis (MCS), which is associated with severe trigonocephaly. Currently, there is no gold standard for how much associated orbitofrontal dysmorphology should trigger surgical intervention. In our study, we used three-dimensional (3D) curvature analysis to separate the phenotypes along the spectrum, and to compare surgeons’ thresholds for operation. Three-dimensional curvature analyses on 43 subject patients revealed that the mean curvature of mid-forehead vertical ridge was higher for patients who underwent operation than those who did not undergo operation by 1.3 m⁻¹ (p < 0.0001). In addition, these patients had more retruded supraorbital areas by −16.1 m⁻¹ (p < 0.0001). K-means clustering classified patients into two different severity groups, and with the exception of 2 patients, the algorithm’s classification of deformity completely agreed with the surgeons’ decisions to offer either conservative or operative therapy (i.e. 96% agreement). The described methods are effective in classifying severity of deformity and in our experience closely approximate surgeon therapeutic decision making. These methods offer the possibility to consistently determine when surgical intervention may be beneficial and to avoid unnecessary surgeries on children with benign metopic ridge and associated minimal orbitofrontal deformity.

Background: Restoring craniofacial symmetry is an important objective in the treatment of many craniofacial conditions. Normal form has been measured using anthropometry, cephalometry, and photography, yet all of these modalities have drawbacks. In this study, the authors define normal pediatric craniofacial form and craniofacial asymmetry using stereophotogrammetric images, which capture a densely sampled set of points on the form. Methods: After institutional review board approval, normal, healthy children (n = 533) with no known craniofacial abnormalities were recruited at well-child visits to undergo full head stereophotogrammetric imaging. The children’s ages ranged from 0 to 18 years. A symmetric three-dimensional template was registered and scaled to each individual scan using 25 manually placed landmarks. The template was deformed to each subject’s three-dimensional scan using a thin-plate spline algorithm and closest point matching. Age-based normal facial models were derived. Mean facial asymmetry and statistical characteristics of the population were calculated. Results: The mean head asymmetry across all pediatric subjects was 1.5 ± 0.5 mm (range, 0.46 to 4.78 mm), and the mean facial asymmetry was 1.2 ± 0.6 mm (range, 0.4 to 5.4 mm). There were no significant differences in the mean head or facial asymmetry with age, sex, or race. Conclusions: Understanding the “normal” form and baseline distribution of asymmetry is an important anthropomorphic foundation. The authors present a method to quantify normal craniofacial form and baseline asymmetry in a large pediatric sample. The authors found that the normal pediatric craniofacial form is asymmetric, and does not change in magnitude with age, sex, or race.

Quantifying ear deformity using linear measurements and mathematical modeling is difficult due to the ear’s complex shape. Machine learning techniques, such as convolutional neural networks (CNNs), are well-suited for this role. CNNs are deep learning methods capable of finding complex patterns from medical images, automatically building solution models capable of machine diagnosis. In this study, we applied CNN to automatically identify ear deformity from 2D photographs. Institutional review board (IRB) approval was obtained for this retrospective study to train and test the CNNs. Photographs of patients with and without ear deformity were obtained as standard of care in our photography studio. Profile photographs were obtained for one or both ears. A total of 671 profile pictures were used in this study including: 457 photographs of patients with ear deformity and 214 photographs of patients with normal ears. Photographs were cropped to the ear boundary and randomly divided into training (60%), validation (20%), and testing (20%) datasets. We modified the softmax classifier in the last layer in GoogLeNet, a deep CNN, to generate an ear deformity detection model in Matlab. All images were deemed of high quality and usable for training and testing. It took about 2 hours to train the system and the training accuracy reached almost 100%. The test accuracy was about 94.1%. We demonstrate that deep learning has a great potential in identifying ear deformity. These machine learning techniques hold the promise in being used in the future to evaluate treatment outcomes.

A precise volumetric assessment of maxillary alveolar defects in patients with cleft lip and palate can reduce donor site morbidity or allow accurate preparation of bone substitutes in future applications. However, there is a lack of agreement regarding the optimal volumetric technique to adopt. This study measured the alveolar bone defects by using two cone-beam computed tomography (CBCT)-based surgical simulation methods. Presurgical CBCT scans from 32 patients with unilateral or bilateral clefts undergoing alveolar bone graft surgery were analyzed. Two hands-on CBCT-based volumetric measurement methods were compared: the 3D real-scale printed model-based surgical method and the virtual surgical method. Different densities of CBCT were compared. Intra- and inter-examiner reliability was assessed. For patients with unilateral clefts, the average alveolar defect volumes were 1.09 ± 0.24 and 1.09 ± 0.25 mL (p > 0.05) for 3D printing- and virtual-based models, respectively; for patients with bilateral clefts, they were 2.05 ± 0.22 and 2.02 ± 0.27 mL (p > 0.05), respectively. Bland–Altman analysis revealed that the methods were equivalent for unilateral and bilateral alveolar cleft defect assessment. No significant differences or linear relationships were observed between adjacent different densities of CBCT for model production to obtain the measured volumes. Intra- and inter-examiner reliability was moderate to good (intraclass correlation coefficient (ICC) > 0.6) for all measurements. This study revealed that the volume of unilateral and bilateral alveolar cleft defects can be equally quantified by 3D-printed and virtual surgical simulation methods and provides alveolar defect-specific volumes which can serve as a reference for planning and execution of alveolar bone graft surgery.