Please check out the accompanying podcast of this blog post (discussion of this article begins at 15:07)
Current evaluation of idiopathic normal pressure hydrocephalus (NPH) depends predominantly on clinical examination, although there are some imaging features, including enlarged lateral ventricles, bowing of the corpus callosum, and enlargement of the sylvian fissures out of proportion to the other cerebrospinal fluid (CSF) spaces. However, these imaging features are largely subjective, depending heavily on the opinion of the interpreting physician.
In this paper, the authors sought to improve evaluation of idiopathic normal pressure hydrocephalus by developing an automated method for calculating CSF volumes within the calvarium, including volume in the lateral ventricles, total intracranial CSF volume, and brain parenchymal fraction. They compared this to manual segmentation and measured volumes in patients being evaluated for NPH both before and after a large volume lumbar puncture.
The results demonstrate that the method was useful for measuring CSF volumes, with good correlation between the automatic method and manual segmentation. Furthermore, the ventricular volume decreased after the large volume lumbar puncture, with the difference most pronounced 30 minutes after the procedure and gradually returning towards baseline over a 24 hour period. The method was also useful for calculating brain parenchymal fraction, which cannot be easily calculated manually.
These findings are interesting for several reasons. First, it is valuable to know that the automated method is reliable when compared to manual measurements and can be used to see how much change there is in the CSF volumes of patients being evaluated for normal pressure hydrocephalus. It is a natural extension of this study to see what the clinical outcomes were for these patients, and potentially know if there was a difference in patient that were ultimately diagnosed with NPH. Ideally, these values could potentially predict which patients might respond to shunting.
Second, and perhaps more importantly, it is nice to see further development of automated techniques for the reading room. Although the authors used a custom sequence not commonly used in clinical practice, it is likely that extensions of this technique could be used to apply them to more commonly performed techniques, such as heavily T2-weighted sequences. Automated measurements such as these will likely be a big part of radiology over the next several decades, and a big part of radiology research in that time will be learning how to intelligently apply them to clinical practice. This paper represents some early steps in that direction.
