Optimization algorithm

[Brim]

Advertisement - Members don't see this ad

I wonder if this algorithm can be used to help speed up IMRT optimization...

~~---~~---~~---

Solving problems by computer just got a lot faster

Solving problems by computer just got a lot faster
A new algorithm shuffles through all possible solutions to find the best answer

A new computer program works smarter, not harder, to solve problems faster than its predecessors.

The algorithm is designed to find the best solution to a given problem among all possible options. Whereas other computer programs winnow down the possibilities one at a time, the new program — presented July 12 at the International Conference on Machine Learning in Stockholm — rules out many choices at once.

 

[scarbrtj]

I wonder if this algorithm can be used to help speed up IMRT optimization...

I wonder. In reality most times optimization for me takes 1-5 minutes total, static beam IMRT. The RapidArc optimizations still take a lot longer.

What do we really mean by optimization? Something that is optimum cannot be bettered by definition. Therefore the optimum plan is the best that could ever be obtained for treating a particular patient with a particular external shape, location of disease and arrangement of internal organs. I propose the view that this optimum plan is unachievable and that in practice this does not matter. To arrive at the optimum plan one would have to investigate the use of: (i) all types of irradiation (protons, carbon ions, photons…); (ii) of all energies (continuous not just those we have available); (iii) all possible numbers of beams from 1 to infinity; (iv) all possible ranges of fluence levels; (v) all possible beam geometry shapes; (vi) all possible fractionation schemes …and so on. It is totally apparent that as ‘‘optimizers’’ we cannot and do not do this.

- Steve Webb, 2003

 

[Brim]

In the day of dial-up modems, we used to have to wait 5 minutes to download a GIF image file. Now that's completely unacceptable and we want it to be nearly instantaneous.

Waiting 1-5 minutes for an IMRT optimization is not bad, but what if it could be 20 times as fast? What if it could be lightning fast?

In reality most times optimization for me takes 1-5 minutes total

 

[scarbrtj]

What if it could be lightning fast?

It will be, soon. Moore's Law and all. I'm also including calc times in the ~5 minute timeframe. Fifteen years ago we would wait ~30-40 minutes (at best), now about 1-5 mins. (GPUs will take us the next part of the way.)

Twenty years ago I had a 128kbps home internet connection, then considered pretty fast. Now I have 1Gbps, about 8000 times faster. IMRT optimization has NOT gotten 8000 times faster in the intervening period.

Clearly we need to let ISPs in on the radiation game.

 

[deleted18755]

Who else remembers going over volumes with dosimetry/physics during lunch time, checking in with them before leaving and seeing that the IMRT plan was "cooking", going home for the day then coming in the next morning and the computer had been going all out all night long but still wasn't done, then getting a page around 10 am to come check it out?

It's hard to believe computers were so slow as recently as ~15 years ago (or was it actually done much sooner but the physicists staff was just using the slow computer/complexity of the plan as an excuse to goof off!?!)

 

[emt409]

I wonder. In reality most times optimization for me takes 1-5 minutes total, static beam IMRT. The RapidArc optimizations still take a lot longer.

What do we really mean by optimization? Something that is optimum cannot be bettered by definition. Therefore the optimum plan is the best that could ever be obtained for treating a particular patient with a particular external shape, location of disease and arrangement of internal organs. I propose the view that this optimum plan is unachievable and that in practice this does not matter. To arrive at the optimum plan one would have to investigate the use of: (i) all types of irradiation (protons, carbon ions, photons…); (ii) of all energies (continuous not just those we have available); (iii) all possible numbers of beams from 1 to infinity; (iv) all possible ranges of fluence levels; (v) all possible beam geometry shapes; (vi) all possible fractionation schemes …and so on. It is totally apparent that as ‘‘optimizers’’ we cannot and do not do this.

- Steve Webb, 2003

RapidArc optimizations and calcs are about to be lightning fast in the next couple years with the roll-out of GPU based treatment planning boxes. I've demo'd them. They are sick.

 

[Chartreuse Wombat]

RapidArc optimizations and calcs are about to be lightning fast in the next couple years with the roll-out of GPU based treatment planning boxes. I've demo'd them. They are sick.

Add it therefore follows that reimbursement will fall and the need for dosimetrists will fall as well. Automation..

 

[thaddeus]

my dosimetrists depend on that prolonged optimization time for their internet surfing needs. I worry about falling job satisfaction if they have to actually do productive work all day.

 

[Haybrant]

my dosimetrists depend on that prolonged optimization time for their internet surfing needs. I worry about falling job satisfaction if they have to actually do productive work all day.

100% this.

 

[medgator]

.

Clearly we need to let ISPs in on the radiation game.

Wonder how well the Comcast AT&T oligopoly will work with the Varian/Elekta one

 

[medgator]

my dosimetrists depend on that prolonged optimization time for their internet surfing needs. I worry about falling job satisfaction if they have to actually do productive work all day.

I guess... they have moments like that mixed in with cranking out multiple plans/running multiple optimizations in a day. I guess it depends on your patient load to dosimetrist ratio