Every sample is biased.
The question is whether these samples are meaningfully biased in a way that would like affect these results.
Yes, those getting a head MRI likely have some symptoms related to their head or brain area, but the vast majority of those will have nothing to do with pineal cysts or tumors; in addition, some of the work in the paper you cite was on healthy volunteers. Healthy volunteers are often over-represented by college-aged students, which likely don't overlap as much with the medical patients.
Similarly, those who require autopsy are not the same of the general population, but I can't think of any reason that the population getting autopsies would have a strong relationship to the presence of pineal cysts.
Pu et al, the paper you cite, refer to differences with age and gender, and it's likely that both MRI- and autopsy-receiving cases differ on those groupings, but not to an extent that would produce a 20-fold difference in prevalence.
So then why such a big difference in autopsy versus imaging results?
I presume this question is what really led you to asking about the autopsy studies. Both the paper you link to, and a related one (Al-Holou, et al 2011) provide some quick clues. Quoting from Al-Holou, et al:
It is not surprising that autopsy studies, because they include cysts that may be too small to be detected on MR imaging, have reported consistently higher prevalence rates for pineal cysts, ranging between 21% and 41%.25,65 In contrast to imaging studies, autopsy studies have included small or even microscopic cysts. For example, in the series reported by Hasegawa et al.,25 45% of all pineal cysts were 2 mm or less in maximum diameter. This detection variability may be a factor even between different estimates of prevalence on MR imaging.
The paper you refer to, Pu et al 2007, specifically used a finer-detail scan:
A possible explanation for the difference between the prevalence of the pineal
cysts demonstrated in our study and that in other MR imaging
studies is that we used high-resolution MR imaging with a 1.0
mm resolution on the 3D T1WI and an in-plane resolution of 1.0
mm with an axial resolution of 2.0 mm on the T2WI. An in-plane
resolution of 1.0mm and an axial resolution of 3.0mm were used
in the prior MR imaging study of healthy volunteers.17
If 45% of cysts from autopsies are themselves smaller than 2mm, it isn't reasonable to expect that most of those will be found on MRI scans, especially those with 3mm spacing.
The autopsy numbers are certainly more accurate from a cyst-counting perspective, but as the Pu et al paper states:
Most cysts are asymptomatic, with diameters ranging from 2 to 15 mm.5-8 When
present, however, symptoms are usually noted in patients with
cysts larger than 15 mm in diameter.9-10
If only cysts larger than 15mm are commonly clinically relevant, all we learn from counting the smaller cysts is in confirming that small cysts are common and not normally associated with symptoms.
Probably the best estimate of general prevalence would come from a pre-registered sample of patients whose brains would be donated specifically for pineal histology post-mortem. I can't see a lot of support or funding for such a study, though, given the lack of clinical relevance.
Al-Holou, W. N., Terman, S. W., Kilburg, C., Garton, H. J., Muraszko, K. M., Chandler, W. F., ... & Maher, C. O. (2011). Prevalence and natural history of pineal cysts in adults. Journal of neurosurgery, 115(6), 1106-1114.
Pu, Y., Mahankali, S., Hou, J., Li, J., Lancaster, J. L., Gao, J. H., ... & Fox, P. T. (2007). High prevalence of pineal cysts in healthy adults demonstrated by high-resolution, noncontrast brain MR imaging. American journal of neuroradiology, 28(9), 1706-1709.