metastase = usually one type of tumor cells, or?

[mastaer]

Hi,

I am a computer engineer. So I hope my question is not too dump.

I got full slides which are h&e stained. On the tissues there are lymph nodes. The pathologist marked the metastases. In this marked regions there can be found multiple types of cells.

1. Can I say that (usually) only one type of cells are the tumor cells in the metastase and represent the tumor?
   
2. If you liked to mark the tumor regions as exact as possible, would you do the prediction for every cell (tumor or not)? (I know the workload would be too high)
3. How important is the coloring of the stroma to do the prediction?
4. Is there a relationship between the capsula and the metastase?

Thanks a lot
Jonas


Appendix 1: The tile from a slide that i got from the patholoist.

Appendix 2: prediction for the cells (Doesn't need to be right)

---

Members don't see this ad.

 

[Granular]

You have two options to get answers to your questions:
First:
1. Apply to medical school.
2. Borrow money to go (unless you are already wealthy).
3. When you finish, go to an AP residency program (this will shave a year off your training).
4. Learn the answers along the way.
5. Finish automating what pathologists do.
6. Sell your system and recoup your time and money.
Trying to jump straight to step 5 (which seems to be your current plan) is not likely to work; cf. recent problems at Silicon Valley darling trying to move directly into lab medicine space.

Or, second:
1. Hire one of the many unemployed or underemployed pathologists who already know the answers. Apparently the pathologist providing and marking slides for you is unwilling to explain things to you. Good for her!
2. Pay your new pathologist handsomely to teach you how to automate what they do, since they may never work again if you are successful.

Your time might be better spent automating other aspects of healthcare, though. Doctors still often want someone to talk to about their specimens/cases and are unlikely to simply accept an artificial intelligence generated diagnosis. Unless they would bear no liability for acting on a wrong answer.

Good luck!

 

[octopusprime]

Hi,

I am a computer engineer. So I hope my question is not too dump.

I got full slides which are h&e stained. On the tissues there are lymph nodes. The pathologist marked the metastases. In this marked regions there can be found multiple types of cells.

1. Can I say that (usually) only one type of cells are the tumor cells in the metastase and represent the tumor?
   
2. If you liked to mark the tumor regions as exact as possible, would you do the prediction for every cell (tumor or not)? (I know the workload would be too high)
3. How important is the coloring of the stroma to do the prediction?
4. Is there a relationship between the capsula and the metastase?

Thanks a lot
Jonas

Bud, not trying to be a jerk, but no clue what you're asking...you're best off talking to someone in person.

 

[kchan99]

Hi,

I am a computer engineer. So I hope my question is not too dump.

I got full slides which are h&e stained. On the tissues there are lymph nodes. The pathologist marked the metastases. In this marked regions there can be found multiple types of cells.

1. Can I say that (usually) only one type of cells are the tumor cells in the metastase and represent the tumor?
   
2. If you liked to mark the tumor regions as exact as possible, would you do the prediction for every cell (tumor or not)? (I know the workload would be too high)
3. How important is the coloring of the stroma to do the prediction?
4. Is there a relationship between the capsula and the metastase?

Thanks a lot
Jonas

I don't understand the nature of your project. Is this for machine learning? What are you trying to classify? Is this something like Imagescope, or are you using Matlab?

Question 1: Please ask your pathologist.
Question 2: This sounds like a biomedical engineering question. I would presume that you would want to create different nuclei masks.
Question 3: It depends on what your doing...?
Question 4: Please ask your pathologist.