pediatric growth curves for cerebral ventricular volume. I have no disclosures. Briefly starting with some background, hydrocephalus is among the most common diseases treated by pediatric neurosurgeons. Since this condition is characterized by dilation of the cerebral ventricles, establishing an early diagnosis requires radiographic imaging to evaluate the extent of the dilation. The current standard for radiology reports is to describe the ventricles qualitatively. While this may work well for cases of obvious ventriculomegaly, more borderline cases can be described with ambiguous terms, such as mild ventriculomegaly or as showing moderate enlargement of the ventricles. This poses issues for communication between providers, as well as in generating an objective diagnosis for these borderline cases. To increase objectivity, others have created standards for quantitative evaluation of ventricular caliber. These standards, such as the Evans index or the frontal occipital horn ratio, use linear measurements of brain imaging to estimate ventricular volumes. Unfortunately, these two-dimensional measurements have some shortcomings, as they do not account for variation in shapes and distributions of ventricular systems, and they may be unreliable in comparing measurements between images due to differences in imaging protocols, such as differences in gantry angulation. In order to overcome these pitfalls, groups including our own have worked to develop softwares for rigorous three-dimensional calculations of ventricular volume from radiographic images. However, if volumetric analysis is to be used in the future, a reference of normal ventricular volumes is needed. While growth charts such as the one shown here are regularly used as clinical references to track and give context to measurements of pediatric patients, no such standard exists yet for the volume of pediatric ventricles. With this in mind, we sought to develop a similar centile estimation growth curve, but for normal three-dimensional ventricular volumes for pediatric patients. Our first step was to obtain a set of normal brain MRIs to be analyzed. To do this, we searched an in-house patient database for patients that were scanned for either headaches, convulsions, or head injuries. We then applied the exclusionary criteria listed here to remove patients with any charted history that might impact their ventricular size or morphology. Next, we performed chart reviews of 50 patients from each age group, which yielded a set of 687 brain MRIs to be analyzed. To analyze the MRI sets, we used a custom semi-automated segmentation software. The program uses k-means clustering to segment the features of each slice by intensity into air, bone, gray matter, white matter, and CSF. The user then selects the ventricles on each slice. The program then uses the input from each slice as well as MRI metadata to build a three-dimensional model and calculate the volume of the ventricles. Now that we had a set of volumes of normal ventricles for patients aged 0 to 18, we move forward to build our growth curves. To do this, we utilize one of the methods that has been used by the CDC in their construction of growth curves, the LMS or Lambda Mu Sigma method. This resulted in the generation of three growth curves. The one shown here was made using all the volumes we collected, while the two on subsequent slides were made using only volumes from male patients or only volumes from female patients. As you can see, the figure demonstrates a period of rapid growth during the first year of life, followed by a period of slower growth thereafter. There is also a slight dip in the higher percentiles between ages 1 to 3. This may be artifactual or may represent a period in which brain growth outpaces head growth. Shown here is the growth chart I just referenced, made from only volumes collected from male patients, and here is the growth chart made from only volumes collected from female patients. All three curves seem to have the same general shape, with female volumes being slightly lower than male volumes at a given age. We then evaluated the appropriateness of fit of these curves to the data, and to do so we used Q-Q plots for normalized residual analysis. Each of the correlation coefficients for these models was above 0.99, indicating an excellent fit. Before I conclude, I will walk through a couple of examples of how these normal growth curves may be useful in a clinical setting. Shown here is a brain MRI of a 3-year-old girl who was evaluated at our institution for diffuse intrinsic pontine glioma. She had no past brain MRIs for comparison. In the radiology note for the scan, the ventricles were described as displaying mild ventriculomegaly. The patient later went on to decline neurologically, and required a shunt for CSF diversion. If the scan is now evaluated volumetrically, we see that the ventricles are measured to be 49 milliliters, placing the volume well above the 95th percentile for a 3-year-old girl. Doing so gives objective context to a somewhat ambiguous scan, and makes the diagnosis of hydrocephalus more likely. To give another example, we anticipate that the curves may be used to identify patients with abnormally accelerated ventricular growth, just as other pediatric growth curves are used to track measurements such as head circumference height or weight. A pathologic process may be indicated in patients with multiple brain MRI scans that demonstrate the crossing of centile lines. Shown here is an example of such a patient plotted against one of the centile curves. In summary, we identify the problem that if volumetric analysis is to be used moving forward, a reference for normal growth of ventricular volume in children must be created. We sought to build this normal reference by making pediatric growth curves for three-dimensional ventricular volume. To do so, we used a semi-automated segmentation program to calculate the ventricular volume of normal pediatric brain MRIs. This resulted in the generation of a set of normal pediatric growth curves, which may serve as a clinical reference moving forward. Thank you for your time.

normal ventricular volumes

pediatric patients

radiology reports

ventricular size

software program