This study developed an automated image analysis method to quantify neuronal response to intracortical microelectrodes by counting neurons in histological images. The method used image processing in Fiji and data analysis in Matlab. It achieved high correlation with manual counts, taking 5 minutes versus over 5 hours. While the automated counts were consistently lower, the difference was not statistically significant. This method provides a consistent, reproducible and faster way to quantify histology and better understand the cellular response to microelectrodes.

1. Automated Image Analysis Method to Quantify Neuronal Response to Intracortical Microelectrodes
Ray Ward 1, Janak Gaire 2, and Kevin J. Otto1,2,3,4,5,6
J. Crayton Pruitt Family Department of Biomedical Engineering
Correlation Coefficients
for Automatic and Manual Counts
Image
Neuron
Count
Neuronal
Density
106 0.9946 0.9890
107 0.9890 0.9386
112 0.9940 0.8987
113 0.9906 0.9493
117 0.9853 0.8868
120 0.9877 0.9345
Average 0.9985 0.9798
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
106 107 112 113 117 120
PercentDifference
Image
Average Percent Difference
Introduction Conclusion
• Time: Manual count took over 5 hours- the
automatic process took under 5 minutes.
• Variance: More variability in manual counts due to
fluctuations in user definition of cell, attention, and
the effects of adjacent cells on perception.
• T-Test: Difference between mean of manual and
automatic counts not statistically significant.
• Difference: Manual count consistently higher than
automatic count. May result from over/under-
segmentation or unclear potential cells. (See below)
• Correlation: High correlation indicates clear linear
relationship.
• This method saves time and effort, providing
consistent and easily reproducible results for
histological quantification.
Future Work
• Method could potentially be improved by
combining sequences of image filters or using
alternative threshold algorithms
• Examine correlation to fluorescent intensity
• Use method for histological quantification of
IMEs, to better understand the cellular
response
Intracortical microelectrodes (IMEs) have a wide variety of
applications ranging from monitoring neuron activity to treating
neurological disorders. But the lack of reliable functionality limits
their use in long-term experiments and clinical implementation.
Functionality loss is associated with the formation of glial scar
around the implant and a loss of nearby neurons (Figure 1).
Quantification of the cell types involved is challenging and
time-consuming, particularly in larger datasets. Without accurate
histological quantification, it’s difficult to accurately describe the
relationship between this cellular response and IME
functionality.
Using Fiji and Matlab, established cell counting
techniques1,2,3 can be adapted to automatically quantify the
number and density of neurons as a function of distance from
the implant.
References
[1]Schindelin, J.; Arganda-Carreras, I. & Frise, E. et al. (2012), "Fiji: an open-source platform for
biological-image analysis", Nature methods 9(7): 676-682.
[2]I Grishagin. Automatic cell counting with ImageJ. Analytical Biochemistry, (473).2015.
[3]https://www.unige.ch/medecine/bioimagingfiles/3714/1208/5964/CellCounting.pdf
1J Crayton Pruitt Family Department of Biomedical Engineering, 2Department of Neuroscience, 3Department of Materials Science and Engineering, 4Department of Neurology, 5Department
of Electrical and Computer Engineering, 6Nanoscience Institute for Medical and Engineering Technology, University of Florida, Gainesville, FL, USA
Methods
• 6 histological sections of implanted cortical tissue stained with anti-NeuN antibody to detect neuronal nuclei
• Images were captured (850 um X 850 um) using confocal microscopy
• Post processing of microscopy images was done in Fiji1
• Data analysis was done in Matlab
• Neuron count and density as a function of distance from implant site was quantified for manually and
automatically identified cells using the following workflow:
Distance Map
Bin Cells by Distance
20µm bins
Select Site
Normalize to Area
Neurons / µm2
Particle Analysis
Redirect to
Distance Map
Gaussian Blur
3-pixel radius
Auto Threshold
Fiji Default Algorithm
FillHoles &
Watershed
Results
Percent Difference
Total Counts and Percent Differences
Image
Automatic
Count
Manual
Count
Total
Percent
Difference
106 722 773 6.60%
107 839 910 7.80%
112 730 765 4.58%
113 565 610 7.38%
117 633 697 9.18%
120 864 965 10.47%
Average 725.5 786.67 7.67%
Correlation
0.0
0.5
1.0
1.5
2.0
2.5
20 60 100 140 180 220 260 300 340 380
CellDensity(cells/µm2)
Distance away from Implant Site (µm)
Auto Count
Manual Count
Neuron Density vs Distance away from Implant Site
Image 120
Figure 1. Histological section of implanted tissue (1 month)
stained with markers for neuronal nuclei and reactive
astrocytes. Red box : Implant location
• In total, 4720 neurons were
identified manually, and 4355
neurons were identified using the
automated method.
• The average standard deviation for
manual counts (7.859) was higher
than that of automatic counts
(7.121).
• When Matlab was used to perform
two-sample t-tests for the manual
and automatic neuron counts of
individual images, the average P-
value was 0.642 (α = 0.05)
▲ Examples of (A) under-segmentation and (B) over-segmentation
x10-3
Acknowledgements
This project was sponsored by Defense Advanced Research Projects Agency (DARPA) Microsystems
Technology Office (MTO), under the auspices of Dr. Jack W. Judy (jack.judy@darpa.mil) and Dr. Doug
Weber (Douglas.Weber@darpa.mil) as part of the Reliable Neural Technology Program, through the
Space and Naval Warfare Systems Command (SPAWAR) Systems Center (SSC) Pacific grant No. N66001-
11-1-4013 and the University of Florida Preeminent Initiative Start-up Funds.