Need some advice to measure divided parts

imagej

#1

Hello,
I am using ImageJ to measure the percentage of dyed area on the root of a tooth.
I want to measure the percentage in two ways.

  1. to find total area of dyed area on the root in 1 picture, this means one side of the tooth (there are in total 4 sides, each side is photographed individually).

  2. to find dyed area in each equally-1/3 divided part (from total length of the root) please see the attached image for more details.

As I searched for tutorials, the first point may be done by normal measurement, while i am wondering for the second. Please give some advice to divide and select the part of divided and measure the percentage of dyed areas correctly.
Thank you in advance.
Pat


#2

Good day Pat,

what youu like to obtain appears possible but we need to see some typical original raw images to give you reasonable advice.

Be aware of the fact that a tooth is a 3D object and an image taken from one side doesn’t allow exact area measurements due to perspective distortion.

Regards

Herbie


#3

Hi Herbie,
Thank you for your quick response.
Attached is an example of photos I took.


I tried to eliminate the fault in presenting data of 3D subject by using a box for the tooth to put on.
Each photo was taken in different 90 degrees.
This would allow me to report the percentage of dyed area in each side of the tooth.
After that I have to crop the root in the Photoshop and then save those photos to analyze in ImageJ.

Do you have some advice for my proces? ,and also please advise how to do measurement in divided part of the root.

Thank you very much.
Pat


#4

Pat,

it turns out as I’ve expected: Things are much more complicated …

  1. No photoshop.
    Do everything in ImageJ. Provide the original uncompressed images. JPG-compression introduces artifacts. We need the original raw images in TIF- or PNG-format. (Converting JPG- to those formats doesn’t help.)
  2. The area measurements will suffer from perspective distortions. There is no way out.
  3. The dye is disposed highly irregular and I have no idea where it starts and ends. At the end of the root I see little to no dye. How do you expect to measure the area.
  4. Illumination is a problem. Use a highly diffuse illumination to avoid specular reflections and make sure the illumination is homgeneous all over the tooth.

Presently I see little hope for reasonable automatic area measurements.

Regards

Herbie


#5

Thank you for your advice
I am going to report the dyed area in yes (all shades of blue), and no (only the root exposed).

As you mentioned, no photoshop is needed. So how can I remove background before the measurement?
I thought using photoshop is the best way to edit the photo because there is a tool to auto-detect the tooth.
Sorry if i misunderstood in this. Also, any advice to divide the root in 3 parts?

Pat


#6

Pat,

please read my comments thoroughly.

As I wrote, I see no way to automatically evaluate your images. The best color separation I get is with the cyan channel after RGB => CMYK conversion.

Regarding the partition, you have to define the “cervical” position first. I have no idea how this is to be done.

Regards

Herbie


#7

Hi Herbie,
I already read through your advice above, but still trying to analyze the pics because previous researches have analyzed this by ImageJ before so I think there must be a way…

As I have tried by making the pic into 8-bit, then processed into binary and analyzed the picture
The summary was as shown, but I want to ask if i understand the results correctly…

This is the original picture and the highlighted results are the total area of the root (of this side)

Then I turned the pic into 8-bit, made binary, and analyzed particle (without any selection), the results were as shown in the picture…

-Now I understand that the black area in binary picture is the white area in the original picture, is this correct?
-So, the number I must choose to calculate the %area will be the number in “Total Area” column, is this correct?

Please clarify these questions, or give any advice. Thank you very much
Regards,

Pat


#8

I doubt that this is the same image as the sample you’ve posted before.

How did you obtain the selection?

Regards

Herbie


#9

Hi Herbie,
No it is not.
I am confused about the meaning of the column in the results,
so I am using the picture that it is clear to visualize the difference of
the staining.
This is the original picture if you want.


#10

For how I obtained the selection, first I cut the tooth and insert black background in the photoshop.
Then I export it to process with ImageJ.
I used color threshold to select whole root to measure overall area.
Hope this is what you need to know…

Pat


#11

Pat,

I can’t follow you in most aspects of what you’ve posted before …

  1. Why do you use a background that is rather similar to the staining?
  2. You are looking for an automatic analysis. From my point of view color-threshold is far from well-suited for this purpose.
  3. The new sample image shows very little staining and I wonder how you were able to select the shown part in your previous post.
  4. If you are interested in area measurements you won’t get accurate results from curved objects like teeth.

Please tell me the exact setting of the color-threshold that led to the selection in your result image. If it was hand-set, then it is not suited for the desired automatic approach.

Clueless

Herbie

re #4.
Let’s assume a cylindrical object and you have 4 orthogonal side-views (no top or bottom views), then the area you compute is about 27% larger than the real cylinder mantle area. The same holds for conical objects and those of similar shape, such as tooth roots.
A camera however, except it has very good telecentric optics, doesn’t give you orthogonal views but central-perspective ones. Therefore, area distortions may be much worse and difficult to compensate by post hoc operations applied to the images.