Yeah, most forms of selection bias fall under collider bias, meaning you are selecting on a variable that is the effect of the exposure and outcome. For example, being young effects phone status, and the outcome effects phone status (say they are very sick thus unemployed, so no phone). So whether or not being young is associated with the outcome a backdoor path connects them biasing the estimate since you are controlling for an effect of both the IV and DV variables via your selection process.