Some questions about data collections

[Colin Gauvin]

I was hoping to get a sanity check on my SerialEM and SmartScope settings. for our Arctica + K3 system

I have SerialEM set to the following:

Low Dose Parameters:

* Search - 280x for grid squares, no defocus offset

* View - 1600x, -50 µm defocus offset

* Record - 36,000x

* Preview - Same as record (both record and preview have dose fractionation/save frames)

Non-Low-Dose Paramters:

Mont-Map: 62x

I have Record aligned to View, and View aligned to Search, so Record --> Search is relatively consistent.

In SerialEM I have done the AFIS calibration. Typically we record 5x5 multishot, 1 per hole, on 300 mesh 1.2/1.3 grids.

Here is my SmartScope data collection:

I'm having a few problems.

1) The montages routinely don't montage the whole grid. I set the same parameters that work for SerialEM: 7x10. In SmartScope, however, I'm getting some clipping at two edges of the atlas (highlighted in red):

2) SmartScope routinely picks grid squares that are off the edge of the stage boundaries, pausing SerialEM and requiring the process to be interrupted. 

Is there any way to set the stage limits and/or distance from edge for square selection? 

3) SmartScope routinely fails to align to the hole. Here is an image where it failed despite otherwise clear  holes. See for example, this image where it failed to align properly:

Then, in the multishot, it's missing the holes. It's tough to tell if this is because of the misalignment, or some otherwise ubiquitous problem, at this point.

I'm wondering if it's actually grabbing a template that's carbon, and not hole. Is there a way to check what the template is?

4) How do I set the multishot number? Is that the BIS max distance parameter?

[Colin Gauvin]

Another thing I'm finding is that eucentric is occasionally way off, and it needs 25+ microns of defocus to correct.

It also seems to want to collect empty holes - not the worse, but I'm wondering if that's intended or not.

[Colin Gauvin]

Image of grid square where it's selecting empty holes:

[Jonathan Bouvette]

Hi Colin,

Thanks for the valuable feedback! I'll try to respond to everything.

1- Partial atlas: I presume that you usually collect the using the Setup Grid Montage Dialog. Up until only a few months ago, this command was unavailable in SerialEM so we implemented the atlas using the regular OpenNewMontage command which, per my understanding, behaves differently with regards to grid limits. This causes the final montage to be different. Also, I'm not sure if SerialEM uses 2 different sets of overlap for the tiles in the 2 different dialogs. SmartScope uses the overlap that were last used un the File -> Open new Montage menu. I would check that they're the same.

2- Squares outside the stage boundaries: We haven't really tested taking full atlases since SmartScope was initially aimed at screening. I personally also always tend to acquire slightly smaller atlases even for data collection since not everything can be used. This is a new problem for us. I can probably implement a radius in um around the origin (Stage X=0,Y=0) where everything outside of it wouldn't be selected. Would that fulfill the need?

3- Alignment to hole failing: That depends on the the protocol that was selected. If you selected Auto with the current settings, it should have defaulted to SPA-Ptolemy. The way this protocol works is the following. When the first square is acquired, a view image will be acquired, the Ptolemy algorithm will find all the holes in the image and average them to create a hole template that is stored in buffer T. You can take a look at that image and see if the image looks like what it's supposed to be. Also, we alway used bin 1 when acquiring the images in SmartScope. You may be ok at bin 2 but i'm not sure beyond that.

If the protocol was SPA, then you need to save a hole template, usually a crop of a single hole in the SerialEM Smartscope directory in References/holeref.mrc. The problem with this procedure is when screening multiple grid types in the same session, you'd have to switch the template.

4 - Multishot number: You are correct, SmartScope uses a radius to group the holes. In the case of R1.2/1.3 and with the value (3um) that was used, the largest group would be 5 in a cross pattern. For 25 holes, i usually enter 7.5 since is the corners are at ~7.3um. You may also set a max group number to reject and groups smaller than a certain amount. SmartScope will try to make as many groups of 25, then 24 with the remainder and so on until it reaches the min group size value. 

5 - Imprecise eucentric height: As you may have seen, the eucentric height is done at the Search (square) mag. This can be imprecised but was designed to speed up screening by avoiding switching back and forth between LM and SA. For data collection, you may select the "-precise" version of the protocol SPA-Ptolemy protocol. This will use the SerialEM default Rough Eucentricity command at the View mag.

6 - Selecting empty holes: The main issue is that SmartScope doesn't have a hole classifier....yet. We're working on it! What I have for now, is a rudimentary gray-level clustering that mainly aims at not collecting the very dark holes at the edge of the square (Cluster 0). However, it is difficult to determine whether the brightest cluster is empty holes or good holes automatically because it's dependent of the grid. I am planning at some point to rework the UI and the clustering to have a slider. In the meantime, you can manually select the empty holes and mark them as empty. Smartscope will not collect them. Usually, for data collection, as squares get collected, I start working on manually marking the bad holes on already acquired squares in the following manner:

• Open the square
• Select all the empty holes by clicking on them
• Click the Actions dropdown and assign the label to the holes
• Repeat for other labels
• At this point, all the holes that have a bad category will not be collected
• (Optional but recommended) Rerun the hole grouping procedure with the `All Hole -> Regroup BIS` button. This will regroup holes into larger groups and omit the newly labeled holes from the groups

I hope this responds to everything,

Best,

Jonathan