Great PR, so-so accuracy

Two full weeks since my last post – holiday activities, don’t you know, starting on December 21 with a solstice party at the home of an eccentric friend.   I see that my blog viewership has declined precipitously in the past week, so I’m happy that you all have better things to do at holiday time than to read about PSP.  Don’t we all wish that the disease itself would take a few days off, too?

My re-emergent thought is about the famous “hummingbird sign.”  On an MRI scan in the sagittal plane – that’s as if you sliced someone down the middle and looked at the cut surface – the brainstem sort of looks like a side view of a hummingbird. 

In the MRIs above, the nose is on the left.  In the lower images, the arrows stop just short of the indicated structures so as not to obscure them.  Note the progressively thinner, sleeker midbrain (the hummingbird’s head and beak) with retention of the plump pons (the belly, which is plumper than than that of a real hummingbird). 

Now here’s the issue.  The appearance of the hummingbird sign isn’t as closely related to PSP as has been implied by many.  There are just too many false positives and false negatives. 

The false positives mostly occur in people with normal-pressure hydrocephalus, a condition where the fluid-filled spaces in the brain (the “ventricles”) enlarge because of an obstruction in the re-absorption of the fluid into the bloodstream.  This stretches the fibers adjacent to the ventricles, impairing control of gait, cognition and the bladder.  It also presses down on the midbrain, producing the hummingbird sign.  Then there are those individuals with corticobasal degeneration where the features resemble PSP (“CBS-PSP”).  They can also have a hummingbird sign.

The false negatives occur in the first couple of years after the initial symptoms.  They also occur if the MRI is mis-aligned on the brain or the head is a little rotated, producing an allegedly midline cut that’s actually a couple of millimeters to one side.  That means that the thinnest part of the midbrain, which is in the midline, isn’t shown in the image. 

You should also know that the hummingbird sign isn’t just about a thin midbrain.  A normal pons is also part of the sign.  That’s because in multiple system atrophy and a few rarer disorders, both the midbrain and pons become thinner.  But in PSP, it’s mostly the midbrain that does so.

I think that in the next year or two, a test of the tau protein in spinal fluid, blood or a tiny punch biopsy of skin will provide a much more accurate diagnosis of PSP than the hummingbird sign.  Soon thereafter we will probably have a PET technique that does the same. Then, clinical treatment trials can be accomplished faster because they won’t have to compensate for the statistical noise produced by participants with a false positive diagnosis.  In fact, all sorts of research on PSP will become much more powerful if people without PSP can be excluded. 

All my best for the New Year.

Mail-order diagnosis

You may have noticed that I’ve been bullish on the ability of ordinary MRI scans to help diagnose PSP.  Now there’s an on-line, automated resource to allow anyone anywhere to upload MRI images and receive an answer – for free.

We’ve known for over a decade that very careful, standardized measurement of the size of various parts of the brain can track the progression of PSP over the 1-year course of treatment trial better than the PSP Rating Scale or any other “bedside” measure.  But more recently, MRI has been found to be highly useful in the differential diagnosis of PSP – that is, telling PSP from normal aging, Parkinson’s, Alzheimer’s, and other conditions. 

For an excellent, technical, open-access review of simple MRI measurements in the diagnosis of PSP, click here. The leading authors are Dr. Aldo Quattrone and his son Dr. Andrea Quattrone at Universita Magna Graecia in Catanzaro, Italy, who pioneered most of the discoveries described.

Such MRI-based measurements use only routinely obtained images like those from your local radiologist.  But actually doing the measurements requires some experience.  The Catanzaro group has created a Web portal onto which anyone can upload de-identified MRI images from a CD.  An answer returns in a few days.  The site is

The black-and-white images below show the inputs into the automated algorithm.  Sorry if these close-up brain images look like abstract expressionism.  The drawings here may help orient you.

MRI images A and B are sagittal (A is in the midline and B is a few mm to one side), images C and D are in the coronal plane and image E is in the horizontal (or axial) plane.

from: Quattrone, et al. Brain Science, 2022.

A: midbrain area (upper outline; Amb) and pons area (lower outline; Apons) (In PSP, atrophy of the midbrain is marked but atrophy of the pons is mild.)

B: middle cerebellar peduncle diameter (This atrophies only a little in PSP.)

C: superior cerebellar peduncle diameter in a slice parallel to the midline (“parasagittal” slice; This atrophies moderately in PSP.)

D: third ventricle diameter (averaging the diameters of the front, middle and back thirds) (This enlarges markedly in PSP.)

E: maximum distance between anterior horns of lateral ventricles (This atrophies moderately in PSP.)

The number derived from these measurements is called the magnetic resonance parkinsonism index (MRPI).  Its value is (Apons/Amb) x (B/C).  Values above 13.88 indicate PSP-RS with 89% sensitivity*, 95% specificity* and 94% accuracy*. This works best in separating PSP-Richardson syndrome from Parkinson’s disease. 

The MRPI 2.0 is (MRPI) x (D/E).  This works better than the original MRPI in separating PSP-Parkinson and other non-Richardson PSP variants from Parkinson’s disease.  Values above 2.70 indicate PSP with 86% sensitivity, 92% specificity and 90% accuracy.

*Sensitivity is the fraction of people with the disease who have a positive test. 

Specificity is the fraction of people without the disease who have a negative test. 

Accuracy is the fraction of people with an accurate test, whether positive or negative. 

In this case, “the disease” means PSP and “without the disease” means PD, some other disease or no disease.

The really valuable part is that this technique works well even in early, mild cases, where a diagnosis could not be made by other means.  In a few studies, such patients were followed for years until they showed more definitive signs, which were then used to validate the initial, image-based diagnoses.

This technique has not been shown effective in differentiating PSP-P from multiple system atrophy of the parkinsonian type (MSA-P), which is a common dilemma for movement disorder specialists seeing a patient with mild symptoms.  But the MRPI and MRPI 2.0 could be combined with other supplementary tests such as supine and standing blood pressure (usually abnormal in MSA-P, normal in PSP) and still-experimental tests such as blood levels of tau, phosphorylated tau and neurofilament light chain (all elevated in PSP, not in MSA) to refine its abilities.

Another important caveat:  Sometimes PSP can be mimicked by rare cases of common diseases like Alzheimer’s or dementia with Lewy bodies, or by some rare diseases like corticobasal degeneration, frontotemporal dementia with parkinsonism, or pallidopontonigral degeneration.  There haven’t yet been enough patients with those things subjected to the MRPI or MRPI 2.0 to prove those formulas able to separate those conditions from PSP.  After all, the MRI only looks for atrophy of certain brain structures, regardless of whether that atrophy is related to tau aggregation or something else.

Bottom line:  As my medical students don’t appreciate hearing, no diagnostic test short of autopsy is ever going to be definitive on its own.  Any test will have to be combined with old-fashioned history and exam and with other imaging, fluids or physiological tests.  Knowing which of those to choose for a given patient and how to interpret the results will keep humble, human neuro-diagnosticians in business for a while longer.


In my next post: another on-line tool for the diagnosis of PSP.

Simple but effective

A chance re-encounter

A 2019 article I came across this week dragged me back into blog posting after a month-long break (sorry, fans — I have no excuse).  I remember seeing the paper at the time but blew it off as mere confirmation of previous publications.  But it actually may provide a way to diagnose PSP years before symptoms appear. 

The problem is a familiar one

As you know from my constant harping on the subject, what we really need are two things: a way to diagnose PSP in its earliest stages, preferably before it causes any disabling symptoms (or any symptoms at all); and a way to prevent the disease process from progressing further than that.  In official lingo: a marker and neuroprotection. 

All sorts of marker proposals are showing promise: leading the pack right now are tests of blood or spinal fluid for neurofilament light chain or tau, PET scans for tau, and various MRI techniques.  Two of the more distant contenders are smartphone-based eye movement measurements and skin biopsies for tau aggregates. The problem is to differentiate very early PSP from normal aging and from competing diagnostic possibilities such as Parkinson’s, MSA and dementia with Lewy bodies. 

Get out your rulers

MRI measurements of the volume of the cerebrum is a very sensitive way to track the progression of PSP and is used in drug trials routinely to compare the rate of brain loss in the treatment group to that in the placebo group.  But it doesn’t work for diagnosing the disease in the first place.  For that, you need to image a part of the brain that, unlike the cerebrum, is involved early in the course of the disease.  It also has to be easy to image using standard MRI machines.  The dorsal midbrain does both. 

As an internal comparator, the study also measured the size of the pons, which is the segment of brainstem just below the midbrain.  It atrophies little in PSP.  For both measures, they used the area in square centimeters of the structure on a mid-sagittal MRI cut (one that slices the head perfectly into left and right halves).  See the image below.

MRI in the mid-sagittal plane, with nose at left of each. The left image shows the dorsal midbrain and the right, the pons. a radiologist drew the outlines by hand with a mouse. The MRI machine’s software calculates the area with each outline in square centimeters. (From Cui et al BMC Neurology 2020)

Now, while the dorsal midbrain is where vertical eye movement, the hallmark of PSP, is situated, it’s not where PSP starts.  That happens in subthalamic nucleus, the globus pallidus and the substantia nigra.  But the dorsal midbrain gets involved soon enough, is much easier to image than those things, and is consistently involved in the classic form of PSP, Richardson syndrome. 

History is not bunk

So, with that as background, Dr. Jong Hyeon Ahn and colleagues from six university hospitals in South Korea found 27 patients with PSP with brain MRIs not only after their PSP symptoms began, but also before they began.  The scans had been performed for non-PSP symptoms such as transient dizziness, fainting, suspected stroke, or headache.  In fact, the article says that elderly South Koreans often request — and receive — brain MRIs as part of their routine checkups.  (Who knew?)  The MRIs were routine, with none of the standardization across radiology sites that are commonplace in multi-center drug studies.  In other words, these were “real world” MRIs.

The pre-symptomatic MRIs were performed an average of 28 months (range: 12-48 months) before PSP symptoms began and the researchers pored over their records to make sure there were no symptoms at the time suggestive of PSP.  They rejected MRIs done within 12 months of symptom onset to further reduce the chance that the symptoms prompting the scan were part of PSP.

They compared these pre-symptomatic MRIs to the same patients’ post-onset MRIs and to those of 27 patients with Parkinson’s and another 27 with no known brain disorder.  The 27 with PSP all had the classic PSP-Richardson syndrome, where the vertical eye movement problem is more prominent than in the less common PSP subtypes. 

I few paragraphs ago, I mentioned that the pons was also measured.  In some diseases, both the midbrain and pons atrophy together, but only in PSP is the midbrain affected far worse.  So they divided the areas of the pons by that of the midbrain, expecting that ratio to be higher in PSP than in competing diagnostic possibilities. 

The results

The graph below compares the four subject groups by their pons area, midbrain area and pons/midbrain ratio. There’s some overlap between groups, but the averages (the means) differ both for the midbrain alone and for the pons/midbrain ratio.  The horizontal bars with asterisks indicate a statistically significant difference between the means of two groups at the ends of the bar.  The pons alone showed no differences, as expected, but the midbrain alone did show a difference and the pons/midbrain ratio did even better than that.

Areas of dorsal midbrain and pons as measured on mid-sagittal MRI. The horizontal brackets with asterisks indicate statistically significant differences between groups. P=pons, M=midbrain, RS=Richardson’s syndrome, PD=Parkinson’s disease (from Ahn et al Park Rel Dis 2022)

Those differences weren’t just at the level of the group means, which would be scientifically interesting but close to useless for patient care.  For the pons/midbrain ratio, the accuracy (the fraction of subjects correctly classified by the test) for pre-symptomatic PSP vs PD was 89% and for pre-symptomatic PSP vs controls, it was 93%.  A more critical statistic from the standpoint of avoiding false positives is the specificity, which for the pons/midbrain measurement comparing PSP and PD, was an amazing 100%.  It was the same for the PSP vs controls. 

Receiver operating curves showing the trade-off between the sensitivity and specificity of the midbrain area (blue) with pons/midbrain ratio (green) in distinguishing patients with PSP from those with Parkinson’s disease (left two graphs) or controls (right two graphs). The two upper graphs compare pre-symptomatic PSP with PD or controls. The two bottom scans compare post-onset PSP with PD or controls. (From Ahn et al. Park Rel Disord 2022)

Now — for the green eyeshades

A strength of the study is that all the pre-symptomatic MRIs were more than 1 year before symptoms began.  Any shorter than that would raise questions of whether very subtle PSP features might have been present.  Another strength is that the MRIs were performed on ordinary machines available in any radiology office.

One caveat is that all 27 PSP patients had the PSP-Richardson form, and the findings may not apply to PSP-Parkinsonism or the other atypical forms.  Another is that the patients were alive and not autopsy-confirmed in their diagnoses and a third is that the neurological evaluations had been performed by general neurologists rather than by movement disorder specialists.

The take-home

So, we await confirmation by other researchers with larger subject numbers and comparisons of PSP with MSA and DLB.  We also need to standardize the measurement of the pons and midbrain areas to strengthen the real-world diagnostic value of this painless, harmless and apparently highly accurate test.  Coupling this test with other simple ones may create an even more accurate diagnostic battery.

This could be a keeper.  Then all we’ll need is a way to keep everyone pre-symptomatic.

Get out those rulers

Everyone with a suspected diagnosis of PSP should have a brain MRI.  It can find more-readily-treated things such as strokes, tumors or normal-pressure hydrocephalus.  But the MRI is not all that useful in differentiating PSP in its early, diagnostically-uncertain, stages from other neurodegenerative conditions such as Parkinson’s, MSA, Alzheimer’s, CBD, dementia with Lewy bodies, and the several forms of FTD.  Even the famous hummingbird sign of PSP doesn’t appear until the middle stages of the disease, by which time a neurologist can make the diagnosis by history and physical exam anyway.  Besides, any disorder that causes atrophy of the midbrain will produce a hummingbird sign.

But now, researchers at the University of California, San Francisco and the Universitat Autònoma de Barcelona have used an automated system to measure the degree of atrophy of several areas of brain as seen on MRI.  The system, called “FreeSurfer,” is in standard use in research requiring MRI measurements. The lead author was Ignacio Illán-Gaia and the senior author was Adam Boxer.  All of their 326 subjects had been evaluated at UCSF’s Memory and Aging Center between 1994 and 2019.  The diagnosis in each case was later established at autopsy – a major scientific strength of this study.  Autopsy showed PSP in 68, CBD in 44, various forms of FTD in 144, Alzheimer’s in 45, and PD, MSA or DLB in only 11.

The four brain areas chosen for analysis were all previously known to atrophy in PSP: cerebral cortex, midbrain, pons and superior cerebellar peduncle.  (The midbrain and pons are in the brainstem and the SCP is one of three tracts connecting the cerebellum to the rest of the brain.)  They used not only the size of each, but also a previously reported index called the “magnetic resonance parkinsonism index” (MRPI), a formula involving the size of the midbrain, pons, SCP and middle cerebellar peduncle. (See note below for details.) The MRPI does very well in distinguishing PSP from PD, but has not been adequately evaluated against all possible alternative diagnoses.  Actually, an updated version called “MRPI 2.0” can distinguish PSP from MSA because it takes into account atrophy of the thalamus, but it’s too new to have an automated version, so this project satisfied itself with the MRPI.

The result was that the MRPI showed an excellent ability to distinguish PSP from the other diseases as a group.  The area under the receiver operating curve (AUROC; see my previous post for an explanation) was excellent: 0.90 of a possible 1.00.  But the AUROC for distinguishing PSP from CBD was only moderate at 0.83.  A more sophisticated statistical analysis, a “multiple logistic regression model” (MLRM), worked even better, distinguishing PSP from the others with a superb AUROC of 0.98.  The CBD- vs-others comparison also benefited from the MLRM, rising to 0.86.

To put the AUROC into more-relatable terms: The AUROC of 0.98 in this case corresponds to an “accuracy” of 95%.  That means that the MLRM got the diagnosis correct (i.e., PSP or not PSP) in 95% of patients.  But that simple calculation can be misleading, which is why the AUROC is used by researchers. 

As mentioned above, the total number of patients with PD, DLB and MSA was only 11.  That’s because the study was performed at a memory center, not a movement center.  While the MRPI has proven its utility in distinguishing PSP from PD, the same can’t be said for the PSP vs DLB or the PSP vs MSA comparisons.  So we need more work with a statistically robust number of patients with DLB and MSA.

For an admittedly biased assessment of the importance of this study, here’s Dr. Illán-Gaia in emailed comments in response to my request for a couple of quotable blurbs:

Our study demonstrates in a large autopsy-proven cohort that combining a set of cortical and subcortical measures of cerebral atrophy could represent a powerful diagnostic tool. These measures can be obtained with a simple MRI and could be combined with other biomarkers to improve the diagnosis of patients with PSP or CBD.

More work needs to be done to ensure the translation of our method to clinical practice and we are now working to validate our results in other large multicenter studies.


The MRPI is calculated as follows: (area of pons on mid-sagittal section / area of midbrain on midsagittal section) X (diameter of middle cerebellar peduncle on parasagittal section / diameter of superior cerebellar peduncle on coronal section). 

The MRPI 2.0 multiplies the MRPI by the (maximum width of the third ventricle / maximum width of the frontal horns of the lateral ventricles).