Sorry but you have pretty much already provided the solution. Which is that such a scan is not at all attainable.
The ATR and the DMI plots are completely different species. Those two plots never cross.
The ATR measures the average of the true range from one candle to the next.
The DMI measures the difference between the high of each candle, and the difference between the low of each candle. Then from that it computes a moving average of those two values, and divides that value buy the ATR. So you see the two studies measure completely different things and there is no way to combine them to create any sort of signal at all. And we find the DMI already contains the ATR as one of it's core components.
For reference, you mentioned a previous post and I want to provide a link to that post for the benefit of the rest of our viewers:
In that post we see an example of a chart overlay and a full explanation of why they are completely useless (even harmful) for trading. Harmful because they create an optical illusion that has no basis in reality. To say they are "not reliable" is missing the point entirely. They are useless because they create an illusion of something that does not exist.
An effective analogy is something called "spurious correlations". If you want to have some fun, enter that phrase in your favorite search engine and check some of the results that come up. Once example that I found is "Divorce rate in Maine correlates with Per capita consumption of margarine". Just because they correlate does not mean they are related in any way.