Normal Pressure Hydrocephalus Radiology Findings

Radiology is the backbone of diagnosing normal pressure hydrocephalus (NPH), a condition where cerebrospinal fluid builds up in the brain’s ventricles despite normal opening pressure on lumbar puncture. The hallmark imaging finding is ventriculomegaly, or enlarged ventricles, that appears out of proportion to any visible brain shrinkage. But a single glance at big ventricles on a scan is not enough to pin down NPH, because enlarged ventricles also show up in Alzheimer’s disease and ordinary age-related brain atrophy. Distinguishing NPH from its mimics requires a combination of specific measurements, characteristic patterns on MRI, and sometimes functional CSF flow studies.

The Evans Index and Ventricular Size

The most widely used radiological yardstick for NPH is the Evans index. It compares the widest span of the frontal horns of the lateral ventricles to the maximum inner diameter of the skull, both measured on the same axial slice of a CT or MRI scan. An Evans index above 0.3 is the standard threshold for defining ventricular enlargement, and it remains a required criterion in most published diagnostic guidelines for NPH.¹ In one large study comparing NPH patients to controls, imaging features with very high specificity for NPH included an Evans index of 0.37 or above.²

The Evans index has endured for decades because it is quick to calculate, needs no special software, and works on both CT and MRI. Its weakness, though, is that it tells you the ventricles are big without telling you why. Age-related brain atrophy can push the Evans index past 0.3 in older adults who have no hydrocephalus at all. That is why clinicians almost never rely on the Evans index in isolation. It serves as the entry gate: if the number is below 0.3, NPH is unlikely. If it is above 0.3, you look for additional signs.

The Callosal Angle

Where the Evans index measures the ventricles from the front, the callosal angle captures their shape from above and behind. It is measured on a coronal MRI slice at the level of the posterior commissure, looking at the angle formed by the roofs of the two lateral ventricles where they meet at the corpus callosum. In NPH, the ventricles balloon outward, pushing the corpus callosum upward and making this angle narrow and acute. In Alzheimer’s disease or normal aging, the angle stays wider because the ventricles enlarge more diffusely.

A study comparing NPH patients with Alzheimer’s patients and healthy controls found the average callosal angle was about 66 degrees in NPH, compared with 104 degrees in Alzheimer’s and 112 degrees in normal controls. Using a cutoff of 90 degrees achieved an accuracy of 93% and a sensitivity of 97% for distinguishing NPH from Alzheimer’s.³ A later systematic review and meta-analysis pooling data across multiple studies confirmed strong pooled sensitivity of about 91% and specificity of about 93% for the callosal angle, outperforming the Evans index, which had a lower area under the curve.⁴ A callosal angle at or below 71 degrees carried 95% specificity for NPH in one head-to-head comparison with vascular dementia and atypical parkinsonism.⁵

The callosal angle does require a properly oriented MRI, which means it cannot be measured reliably on routine CT scans. Still, because of its high diagnostic accuracy, it has become one of the most valued single measurements in NPH imaging.

The DESH Pattern

DESH stands for disproportionately enlarged subarachnoid-space hydrocephalus, and it describes a distinctive distribution of cerebrospinal fluid that sets NPH apart from other causes of big ventricles. The pattern has two components: the ventricles are enlarged and the Sylvian fissures (the deep grooves on the sides of the brain) are widened, but the subarachnoid spaces at the top of the brain, over the high convexity and along the midline, are tight and compressed. In other words, the fluid is concentrated in the wrong places. It collects below but cannot flow upward to be reabsorbed normally.

This mismatch between wide Sylvian fissures below and tight sulci above is a visual signature that experienced radiologists can spot on standard MRI sequences. Research suggests DESH functions as an imaging biomarker for NPH and helps predict who will respond to shunt surgery.⁶ Enlarged Sylvian fissures and focally dilated sulci were each independently significant predictors of NPH in logistic regression models, with the presence of focally enlarged sulci carrying a tenfold increase in the odds of the diagnosis.⁷

Another clue on structural imaging is disproportionate widening of the temporal horns of the lateral ventricles, beyond what can be accounted for by hippocampal atrophy alone.⁸ Temporal horn width of 7 mm or more reached 95% specificity for NPH in one study.⁹ This finding is particularly useful when you are trying to decide whether the enlarged ventricles on an elderly patient’s scan reflect hydrocephalus or Alzheimer’s-related atrophy, because in Alzheimer’s the hippocampus shrinks to match the temporal horn widening, while in NPH it does not.

Separating NPH from Alzheimer’s Disease on Imaging

The overlap between NPH and Alzheimer’s disease is the single most common diagnostic headache in NPH radiology. Both conditions are prevalent in the elderly, both produce enlarged ventricles, and both cause cognitive decline. To make matters worse, some patients have both simultaneously. Systematic reviews have concluded that no single MRI marker is sufficient on its own and that combining multiple sequences and measurements provides the best diagnostic separation.¹⁰

In practice, clinicians look at a constellation of findings. NPH tends to produce dramatic ventriculomegaly with relatively preserved sulci at the high convexity, a narrow callosal angle, widened Sylvian fissures, and sometimes visible periventricular white matter changes. Alzheimer’s produces more uniform brain shrinkage, wider perihippocampal fissures, and dilated sulci distributed evenly across the brain surface. In NPH, the volume of cerebrospinal fluid is shifted: up in the ventricles and Sylvian fissures, down at the superior convexity. In Alzheimer’s, it tends to increase more uniformly.

Periventricular white matter changes deserve special attention. Bright caps and halos around the ventricles on FLAIR or T2-weighted MRI are common in NPH and represent increased water content in those tissues. Research has shown that the absolute water content in these periventricular caps is higher than in ordinary white matter lesions, and a threshold of about 15% increased water content relative to normal white matter pointed toward periventricular caps rather than generic white matter lesions with high specificity.¹¹ These changes likely reflect transependymal migration of fluid from the distended ventricles into the surrounding brain tissue.

CSF Flow Studies

Phase-contrast MRI can measure the actual movement of cerebrospinal fluid through the cerebral aqueduct, the narrow channel connecting the third and fourth ventricles. In NPH, this flow is characteristically hyperdynamic: the fluid sloshes back and forth with each heartbeat more vigorously than normal, even though the overall drainage of fluid is impaired.

Studies using phase-contrast MRI have found that NPH patients show significantly higher peak flow velocities and stroke volumes through the aqueduct compared to both healthy controls and patients with brain atrophy.¹² The stroke volume parameter proved especially reliable in one study, achieving perfect accuracy for identifying NPH.¹³ Other groups have similarly reported markedly elevated systolic peak velocity and mean velocity in NPH, consistent with a hyperdynamic CSF circulation pattern.¹⁴

This hyperdynamic pattern contrasts sharply with obstructive hydrocephalus, where a physical blockage in the CSF pathways reduces aqueductal flow velocities and stroke volume.¹⁵ That distinction matters clinically, because the two types of hydrocephalus require different treatment approaches.

An older technique, radionuclide cisternography, involves injecting a radiotracer into the spinal fluid and tracking where it goes over 24 to 72 hours. The classic NPH pattern shows reflux of the tracer into the ventricles and failure of the tracer to reach the brain’s convexities on delayed images.¹⁶ Cisternography has largely been replaced by phase-contrast MRI in most centers, but it remains available and can be useful when MRI is contraindicated.

Perfusion and Diffusion Imaging

Beyond structural measurements and flow studies, more advanced MRI techniques can probe what NPH is doing to the brain’s tissue and blood supply. Arterial spin labeling, a method of measuring blood flow without contrast injection, has shown that NPH patients have reduced perfusion in the periventricular white matter, the deep gray matter nuclei, and the thalamus compared to healthy individuals.¹⁷ This reduced blood flow in deep brain structures may partly explain the gait difficulties and cognitive slowing that define NPH clinically.

Diffusion tensor imaging (DTI) offers a window into the integrity of the brain’s white matter tracts. In NPH, the corticospinal tract, which carries motor commands from the brain to the spinal cord, shows increased fractional anisotropy in patients who go on to respond to shunt surgery. That sounds counterintuitive, since higher anisotropy usually means healthier tissue. The explanation is that the enlarged ventricles mechanically stretch the fibers, aligning them more tightly along one axis. After shunting, these values tend to normalize as the ventricles shrink and the stretching resolves.¹⁸ More advanced diffusion models have confirmed this stretching effect and added the finding that estimated axon density may be lower in NPH patients and does not recover after surgery, potentially serving as a marker for irreversible damage.¹⁹

A newer application of DTI looks at the brain’s glymphatic system, the waste-clearance network that runs along perivascular spaces. Researchers have developed a metric called the ALPS index (for “analysis along the perivascular space”) that measures how easily water diffuses along these channels. In NPH patients, the ALPS index is significantly lower than in controls, suggesting impaired glymphatic drainage.²⁰ The index also differentiated between NPH patients who responded to treatment and those who did not, with non-responders showing lower values.²¹ A separate study confirmed that the ALPS index could cleanly separate confirmed NPH from controls and from suspected-but-unconfirmed cases.²²

Can Imaging Predict Who Will Improve After Shunting?

This is the question clinicians care about most, and the honest answer is that conventional imaging markers alone are not very good at it. A systematic review and meta-analysis of radiological predictors found that only the callosal angle and periventricular white matter changes significantly differentiated shunt responders from non-responders, and both had low discriminative power.²³ None of the other standard markers, including the Evans index, DESH score, and various volumetric measures, reliably distinguished responders from non-responders. A separate study specifically found no significant preoperative differences in DESH score, Evans index, or callosal angle between patients who improved after shunt surgery and those who did not.²⁴

This gap between diagnostic accuracy and prognostic accuracy is a real limitation. The same measurements that tell you a patient probably has NPH do not reliably tell you whether that patient will walk better or think more clearly after a shunt. Clinical testing, particularly the response to a temporary lumbar drain or a large-volume lumbar puncture, remains the primary tool for predicting surgical benefit. Imaging sets the stage, but the tap test often closes the deal.

Machine learning approaches are trying to bridge this gap. One recent study trained algorithms on combinations of T2-weighted and FLAIR MRI sequences and achieved area-under-the-curve values around 0.88 for predicting gait improvement and around 0.74 for predicting overall functional improvement after shunting.²⁵ These models extract patterns from the images that human readers may not easily quantify, though they still need validation in larger, multi-center studies before entering routine clinical use.

Automated and AI-Driven Diagnosis

Measuring the Evans index and callosal angle by hand is not difficult, but it does take time and introduces variability between readers. Multiple groups are now developing automated systems that extract these measurements from scans without human input. A deep learning pipeline for automated callosal angle measurement achieved a mean error of under 3 degrees compared to expert manual measurement, with a correlation coefficient of 0.98.²⁶ When classifying angles as narrow or wide using the 90-degree threshold, the algorithm misclassified only 7 out of 216 scans in an external validation set.

Other AI models go further, combining multiple imaging features into a single diagnostic score. One group used an XGBoost model trained on volumetric measures including high-convexity tightness, Sylvian fissure enlargement, callosal angle, and Evans index, achieving an area under the curve above 0.93 on an unseen test set.²⁷ A deep learning approach using convolutional neural networks reported even higher accuracy for NPH classification, reaching 98% accuracy with perfect specificity in its validation cohort.²⁸

A newer metric called the ventricle-to-subarachnoid volume ratio (VSR), calculated automatically from CT scans, identified NPH with about 94% sensitivity and 93% specificity. Importantly, this ratio was not correlated with age, which is a practical advantage over some other volumetric measures that drift upward as people get older.²⁹ An age-independent marker is especially useful because NPH overwhelmingly affects the elderly, and age-related atrophy is the main confounder in all of these measurements.

Imaging After Shunt Surgery

Radiology does not stop being relevant after a shunt is placed. Follow-up imaging, typically with CT, is used to confirm that the ventricles are shrinking appropriately and to watch for complications. The most common post-shunt complication is subdural fluid collections, either hygromas (clear fluid) or hematomas (blood), which develop when the brain pulls away from the inner skull as the ventricles decompress. In one series, subdural hematomas or hygromas accounted for 27 of the post-operative complications, though only 3 of those required reoperation.³⁰ Most complications were caught through a combination of routine scheduled CT scans and extra appointments triggered when patients reported new symptoms, underscoring the value of both systematic imaging follow-up and patient communication after surgery.

Asymptomatic Ventriculomegaly and the Preclinical Window

An intriguing radiological finding is what researchers call AVIM: asymptomatic ventriculomegaly with imaging features of NPH on MRI. These are people whose scans look like NPH, with an Evans index above 0.3 and tight high-convexity sulci, but who have no gait problems, urinary symptoms, or obvious cognitive decline. A population-based study of elderly Japanese adults found that about 1% of those over 61 had this pattern without any clinical symptoms.³¹ During follow-up, some of these individuals went on to develop symptoms.

A larger prospective study followed 52 people with AVIM over several years and found that roughly half developed clinical NPH during the follow-up period.³² Even those who remained nominally asymptomatic showed subtle declines on sensitive cognitive tests and motor tasks compared to matched controls without ventriculomegaly.³³ The implication is that the radiological changes of NPH may precede the clinical symptoms by years, and what appears to be an incidental finding on a brain scan could be an early warning.

Whether early intervention in asymptomatic individuals would prevent or delay progression is unknown. No trials have tested shunting in the AVIM population, and it would be a hard sell ethically given that brain surgery carries real risks and many of these people never develop symptoms. For now, the practical takeaway is that a radiologist who spots this pattern on an otherwise unremarkable brain scan should flag it, and the patient should be monitored clinically over time rather than having the finding dismissed as mere aging.