In lung cancer classification in CT scan images, is false positive or false negative is more important to care about? in most paper they try to reduce false positives, but in some articles they showed false negative is more important to be decreased, in my opinion both are important but false negative is a little bit more, I am right or not? please help understand it with a reference if possible?
In general, this depends on your scenario. Most important is to have the value of both as low as possible.
HTH.
Dr. Samer Sarsam
I agree with you. I believe a false negative is more dangerous. Although we are still far from completely relying on machines to detect cancer, we still need their results to be credible enough in the eyes of medical practitioners.
In medicine, they prefer to reduce false negatives. There are many articles that supports this argument, see link below
https://scholar.google.ca/scholar?hl=en&as_sdt=0%2C5&q=false+positive+in+cancer+screening&btnG=
This is because the damage from not knowing that a patient has cancer is more severe than believing falsely that he/she have cancer.
I believe you should try to minimize false negatives, i.e. avoid diagnosing a person as healthy when he/she is actually sick. If you diagnose a healthy person as sick, additional tests can help towards the right diagnosis. On the other hand, if you diagnose a sick person as healthy, they might not get additional tests done.
First, let us make some definitions:
A false positive = person is considered as sick but actually is healthy
A false negative = person is considered as healthy but is actually sick
What does it mean?
False positive cases lead to overspending due to unnecessary care and damaging health of otherwise healthy person due to unnecessary side effects of the therapy.
When you get a too high number of false positive cases, insurance companies will make a lot of noise and quite the financial support of the therapy.
A false negative case means that your patients get sicker or die. Insurance company saves money for no provided therapy but in long run, they might pay more for subsequent therapies.
This is exactly the moment when ethics enter medicine. When to provide care and when not that all combined with the net expenses.
It is very difficult to make the right decisions.
Personally, I would like to recommend you to stratify, whenever possible, the validity/certain of the result which is provided.
The second what you might think about is to find some cross-check evaluation meghod(s) of the same results. It can increase in some cases the validity of the output. Preferential I use an absolutely independent evaluating method.
Try to think out of the box and apply even exotic and less known methods with known good results in other medical and other application areas.
Good luck! :-)
many thanks to all of you for your helpful recommendation, it's clear that there is no such fact that tells your model should have lowest possible false positive or false negative.
False positive (FP) : pixels falsely segmented as foreground
False negative (FN) : pixels falsely detected as background
Both should be minimized. In medical image segmentation, false positive (FP) is big issue.
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification