Thank you so much for inviting me and also in general for organizing the seminar and

putting together the session.

Today, I'm actually not going to talk about the ensemble common filter, but I'm going

to present an approach that we developed in the context of a specific application for

sequential decision support.

But the method is not restricted to that problem and could be applied to problems of various

fields.

And I would like to acknowledge my collaborators on this work who are Corinna Meyer, Niklas

Hartung, William Hussinger, all from the University of Potsdam.

And then there's Charlotte Hockhoff, who is from the pharmacology department of the

University of Berlin.

So in the beginning, I'm going to present you with the original problem from the field

of pharmacology that inspired us actually to develop the method I'm going to propose

to you today.

And it will also be the key numerical example during my talk that I will always come back

to.

So it has been known now for a while that the response to medical treatment is subject

to high variability across patients, which basically also gave rise to the field of personalized

medicine, which has become very popular.

And this variability is particularly harmful when the effects of the treatment can be severe

or in the worst case, even life-threatening.

So our focus here will be on cancer-induced chemotherapy, which is one of the scenarios

where treatment needs to be carefully adjusted to the patients, otherwise the patient might

die.

So here on the right-hand side, you can see a graph that actually visualizes the effect

of chemotherapy on the neutrophil counts in the human body.

So you have the neutrophil counts on the y-axis here.

And so the neutrophil concentration is significantly reduced by chemotherapy drugs, which basically

corresponds to the immune system being severely weakened.

And physicians have categorized these concentrations in five categories, ranging from grade 0 to

grade 4.

So you can see the red lines here.

And grade 0 is associated with an intact immune system.

So while grade 4 means that the patient basically has to go to the ICU because the immune system

is not capable to defend the body from otherwise harmless infections that are now becoming

really dangerous to the patient.

So for the treatment to be successful, physicians try to reach grade 2 to 3, since this typically

means not only that the patient's immune system is intact, but also that the cancer cells

are intact.

And that's the ultimate goal in the treatment.

And of course, they're still trying to avoid to actually get down to grade 4.

And now we have three curves here.

And they represent three individuals reacting very differently to the same amount of drug

that is administered right here at the same time.

And it's also the same amount of the drug.

And as you can see, for example, this curve does not even reach the grade 1.

While the other two are going like this one is actually doing OK because it's still staying

in grade 3.

But the other one is already touching grade 4.

So maybe it should not have received that much of the drug.