Case study 1 – Alzheimer’s Dementia
The first case study is on Alzheimer’s Dementia.
We are presented with a 78-year-old woman with progressive memory loss. A 3D volumetric T1 acquired sequence is used for this woman’s MRI scan and the Neuroreader® Report is used as the output to interpret the results.
Explanation of terms
The Neuroreader® Report shows the brain volume assessment, where you will find the brain volume structures, colors, corresponding segmentation, the volume in milliliters, the volume corrected for total intracranial volume with the total intracranial volume being representation of gray matter, white matter and cerebrospinal fluid.
You will also find the Neuroreader® Index and Z scores, which give indications of how far away an individual brain structure is from the normative database, and a percentile to interpret that metric and for the purposes of Neuroreader® the percentile for brain volumes of less than 25 percentile is considered abnormally low. And for ventricles, typically the ventricles show percentile that is above 75th percent of the abnormally high.
The 25/75 ration is the default setting and will be shown in summary with an orange color. The margins can, however, be customized to the customer’s needs.
Interpretation of data
It is important to understand that the numbers are not hard and fast numbers, but they are normal ranges to consider to help interpretation.
For this particular woman, the whole brain matter is abnormally low at the percentile point, 54 percentiles. Gray matter is also low, just around the 10th percentile, but white matter, even though it’s not color coded, is clearly still not normal, given that it’s about six Neuroreader® indices, almost seven away from the control database, and the percentile still on the lower end of normal at 33rd percentile. The hippocampus is abnormally low, as well as the amygdala volumes and the thalamus at the same time.
This means that we already now should be concerned that the hippocampal volume loss alone suggests that this is Alzheimer’s Dementia. And even though that it is not the only reason why the hippocampus might shrink in older individuals with these clinical presentations, that’s the most likely reason we could encounter these findings.
At the same time, we also see that there are additional structures highlighted by Neuroreader®, close to 50 brain structures in all, and we can see that the Pallidum and all of the low bar structures are abnormally low. If we look at the low bar structures, we see that there’s atrophy included in the temporal lobes as well as the parietal lobes, but diffusely, and at the same time, the lateral ventricles are abnormally enlarged, color coded orange because they’re above the 75th percentile. When we see this diffuse volume loss throughout the brain affecting particularly the medial temporal structures hippocampus low bar structures, then this is very suggestive of a diagnosis of Alzheimer’s Dementia.
Quality Check
The Neuroreader® Report also has segmentation images showing color coded areas representing the different segmented areas. The purpose of these anatomical overlay images is to provide an approximate inspection of the quality of the images.
Typically, when there are segmentation issues, they can be spotted pretty easily, and they can show up as distortions of the colors and the brain structures. And we can see here that the overlap of colors to brain regions is very clearly delineated, suggesting an accurate segmentation, which is the purpose of these quality assessments images.
Heat Maps
The Neuroreader® Report also shows Brain Heat Maps, which show the areas of the brain thatcorrespondsto lower brain volumes or abnormally high ventricular volumes.
Validation
This is another way of indicating a visual correlate to the statistical output presented earlier in the Neuroreader® Report. We know from prior published literature that there’s validation technically of Neuroreader® and several diseases. The initial validation paper shows that you can have accurate segmentation of the hippocampus.
Manual segmentation vs automated segmentation
In particular, whether you’re talking about controls, mouth cognitive impairment, or Alzheimer’s dementia, these are dissimilarity coefficients, which are very high, around 88% or so. And this number typically holds across all types of cognitive decline.
The reason why that’s important is that it shows that if we compare the automated volumetric segmentation with Neuroreader® to so called gold standard anatomical segmentation by an anatomist that there’s a very high degree of agreement between the two methods. Except with volumetrics, you have a very high, thorough, and accurate segmentation in a short time period and hand drawn segmentations typically take longer on the order of five minutes for Neuroreader® on average for 45 brain structures compared to 40 minutes for the volumetric manual segmentation.
