We are am now tackling with the artifacts in MEG. We have to use ICA to distangle blinks and heartbeat, as no EOG and ECG were recorded intentionaly. We don't now how to recognize the ICA components related to ECG. See the following picture, can we regard IC4 and IC10 as ECG signal, IC2I and C3 as EOG signal, and retain all other components.
Your selection of IC2 and IC3 look indeed like they are are related with eye movements.
The peaks in IC10 could look like heartbeats, but I'm not sure about IC4.
This is a complicated and very manual task, there is no clear guidelines that we could give.
One thing you could try is to change the number of components in your ICA decomposition: see if you get cleaner IC signals with more or less components.