Image-guided Planning System for Skull Correction in Children with Craniosynostosis
Need: Approximately 1 in 2,500 children is diagnosed with craniosynostosis, a condition where the cranial bones fuse together prematurely. This causes cranial malformations and can lead to elevated intra-cranial pressure, brain growth impairment, developmental deficiency and other lifelong health challenges. Current surgical approach for the treatment for craniosynostosis is based on the subjective assessment of cranial malformations and options for treatment rely on the experience of the surgeon and population average templates that do not adapt to the specific cranial shape of each patient.
Solution: To minimize the subjectivity in the diagnosis of craniosynostosis and to optimize the surgical treatment of each patient, we have created novel methods that quantify cranial malformations objectively from computer tomography images and provide the surgeons with the sequence of surgical tasks necessary for an optimal treatment. Moreover, we have created iCSPlan, a software platform that builds on our novel methods and that aims at translating our research into clinical practice.
Impact: iCSPlan will fill a critical need, as there is currently no publicly available reproducible standard for the objective and optimal treatment of craniosynostosis. We are working with industrial partners to refine and validate iCSPlan, and to commercialize the advanced algorithms and software prototype.
Funding: NIH R41HD081712 and NIH R42HD081712
Early and Non-Invasive Detection of Genetic Syndromes
Need: Each year, nearly one million children are born with a genetic syndrome. Children with these disorders have a high incidence of serious medical complications and intellectual disability that require immediate treatment, so it is critical to detect those genetic syndromes as early as possible. Unfortunately, the vast majority of these syndromes are not identified in utero or perinatally due to the high cost and inaccuracy of testing, limited access to specialized clinics and lack of outward symptoms at birth. Delay or error in diagnosis causes these children significant suffering, irreversible injury and even death, resulting in up to 53 percent of hospital admissions and 20-30 percent of childhood deaths.
Solution: We have created a low-cost, non-invasive facial analysis technology that identifies automatically specific clinical features associated with a wide array of genetic syndromes using only face photographs of children. We have also created mGene, a smart phone application that allows clinicians to screen newborns and young children without the need for blood tests and specialized genetic expertise, which are usually unavailable in community hospitals and the developing world. Unlike other screening tests for genetic disorders, mGene tells immediately if a child is at risk for a wide range of genetic disorders, prompting immediate referral to specialized clinics for comprehensive care.
Impact: The accuracy of our system to detect a wide range of genetic syndromes is already 95%. Our system's high accuracy combined with lower clinical costs, have a lifesaving potential especially for children whose condition is only subtly apparent in physical terms. The net effect will be a reduction in the morbidity and mortality for the one million children born worldwide annually with genetic syndromes. Ultimately, mGene will make sophisticated genetic expertise widely and affordably available.
Funding: Pediatric Innovation Fund; Ministry of Health and Prevention of the UAE; NIH UL1 TR001876/KL2 TR001877
Non-Invasive Non-Radiating Diagnosis of Pediatric Hydronephrosis
Need: The most common kidney abnormality for children is hydronephrosis. Hydronephrosis is a swelling of the kidney caused by a blockage of urine drainage. If left untreated, hydronephrosis associated with obstruction can cause permanent kidney damage and in severe cases it requires surgery. To detect hydronephrosis, physicians use ultrasound images, which are safe for children and cost effective, but they can be difficult to interpret and often lead to variable diagnoses. For these reasons, when hydronephrosis is found with ultrasound imaging, a child is often required to undertake a variable number of diuretic renograms until a final decision of surgical need is made, which is invasive and involves radiation.
Solution: We have created two different solutions to improve the management of children with hydronephrosis. On one hand, we have developed a technology that uses artificial intelligence to provide a higher diagnostic accuracy from diuretic renograms, thus reducing the amount of studies and radiation needed to assess the need for surgical correction. On the other hand, we have created a low-cost software tool called KidCAD designed to evaluate the severity of hydronephrosis using only non-invasive, non-radiating ultrasound imaging. This solution incorporates advanced quantitative imaging and machine learning methods to provide the first system able to assess kidney function from regular ultrasound images. The proposed system allows establishing a link between non-invasive non-ionizing imaging techniques and renal function, limiting the need for further invasive and ionizing diuretic renogram studies.
Impact: The net effect of our technology will be a decrease in radiation exposure and potential morbidity for children with hydronephrosis. An additional benefit will be the reduction of clinical costs associated with complex and invasive nuclear imaging tests.
Funding: Philips Ultrasound, Inc.; Joseph E. Roberts, Jr., Endowment Award
Ultrasound-Ventricular Shape Analysis in Prediction of Hemorrhagic Hydrocephalus in Premature Neonates
Need: Premature neonates with intraventricular hemorrhage followed by hydrocephalus (post-hemorrhagic hydrocephalus, PHH) are at the highest risk of severe palsy and other adverse neurodevelopmental outcomes. Cranial ultrasound (CUS) is a safe and easy imaging test to perform in neonates during the first weeks after birth to identify intraventricular hemorrhage. However, the lack of a standardized method for CUS evaluation has led to significant variability in decision making regarding treatment and causing delay in initiating treatment.
Solution: Our team of neurologists, neonatologists, radiologists and quantitative imaging specialists is developing a new quantitative imaging framework to automatically quantify the morphology of the cerebral ventricles in early CUS. This solution incorporates advanced quantitative imaging and machine learning methods to analyze the morphological features in order to predict the potential need for temporizing intervention in premature neonates. The proposed tool allowed us to establish a quantitative method for PHH evaluation on CUS in extremely premature neonates with intraventricular hemorrhage.
Impact: Our tool would be so useful for neurosurgical consultation teams (neurology, neonatology and neurosurgery) to make a decision early without any bias in a safe way. Also, it could help identify neonates who are at risk sooner and decide early preventive treatment.
Funding: Sheikh Zayed Institute for Pediatric Surgical Innovation
Radiation-Free Quantification of Intracranial Volume and Head Malformations from 3D Photography
Need: The evaluation of brain development in children with cranial pathology plays an essential role in surgical treatment and patient monitoring, in which the intracranial volume (ICV) and cranial malformations are two important measurements. In practice, ICV and cranial shape are evaluated using computed tomography (CT) or magnetic resonance imaging (MRI). However, CT involves radiation, which is harmful for young children, and MRI requires sedation or anesthesia which affects neuron development for young children. 3D photography offers non-invasive, radiation-free and anesthetic-free evaluation of craniofacial morphology. But the robust quantifications of both ICV and cranial shape measurements are not possible with current 3D photography systems.
Solution: We have developed a computational framework to quantify ICV and head malformations from 3D photography, and use them to characterize cranial shape abnormalities objectively and quantitatively in patients. A set of landmarks were automatically identified in 3D photographs of patients. That allows the cranial shape to be extracted and used to compute the head volume, cranial bone volume, intracranial/brain volume and cranial bone malformations at every location on the cranium.
Impact: The novel automatic framework has the potential to reduce the use of radiation- and sedation-based imaging for surgical planning and patient monitoring. We are working with collaborators to collect more data and validate our framework. We are also working on extensions to support facial dysmorphology analysis from 3D photography.
Funding: NIH R42HD081712
Volumetric Stratification of Risk of Vision Loss in Optic Pathway Gliomas
Need: Optic pathway gliomas (OPGs) are tumors involving the visual system. These tumors can affect hormone production, appetite and sleep, and can cause irreversible vision loss leading to permanent disability. A major impediment in understanding the effect of the progression and benefit of treatment on vision has been the inability to accurately quantify OPG growth. Given this uncertainty, some children will sustain lifelong disability from their vision loss, even despite receiving treatment for their tumor, likely because treatment is started only after the loss of vision occurs.
Solution: Our team of neuro-ophthalmologists, neuro-oncologists, radiologists and quantitative imaging specialists is developing a novel magnetic resonance imaging (MRI) analysis technology that will accurately identify subtle changes in tumor progression. Our technology will enable identifying impending vision loss, and thereby providing an opportunity for early treatment and preserving vision for children with OPG. The technology incorporates advanced imaging and machine learning methods to provide the first system able to accurately quantify anatomical and structural imaging features of the OPG and of the visual pathway to identify clinically-proven features of vision loss.
Impact: We demonstrated for the first time that greater OPG volume predicts axonal degeneration, a biomarker of vision loss, and that MRI volumetric measures can stratify the risk of visual loss. In response, the journal Neurology wrote in an editorial that “The future of the care of children with ... OPG seems brighter.”
Funding: Gilbert Family Foundation
Lower Respiratory Tract Infections related to Prematurity
Need: Prematurity is the largest single cause of death in children under five in the world and lower respiratory tract infections (LRTI) are the top cause of hospitalization and mortality in premature infants. Clinical tools to predict and prevent severe LRTI in premature pediatric patients are critically needed to allow early interventions to decrease the high morbidity and mortality in this patient group. Although imaging biomarkers of lung disease from computed tomography have been successfully used in adults, they entail heightened risks for children due to cumulative radiation and the need for sedation. Our goal is to address these gaps and improve clinical practice by developing an objective imaging biomarker framework to assess the risk of severe respiratory disease in premature babies using non-invasive low-radiation X-ray imaging.
Solution: The image processing pipeline developed by our lab integrates three novel technical components: a) automatic lung segmentation, b) obtrusive object removal in CXR, and c) severity quantification of lung pathology. These innovations enabled the development of a quantitative imaging software technology to quantify Lung Air trapping and Irregular opacities Radiological analyzer (LungAIR).
Impact: Our approach will enable better clinical management of diseases of prematurity leading to novel diagnostic strategies to improve treatment and outcomes for the highly vulnerable population of premature infants.
Funding: NIH UL1TR000075, NIH KL2TR000076 and NIH R41 HL145669