I have never isolated mtDNA from human tissue but the freezing might cause you problems. If your protocols are based on seperating organelles and then their DNA, then forget it. In principle, a simple total DNA isolation followed by CsCl gradient should give you enough DNA. But, as you said you struggle to isolate mitochondria, I fear you ask for something impossible as the freezing has compromised the organellar integrity. Also, as there is so much mtDNA in a cell, there is no need to isolate organelles to enrich for mtDNA. You only need to seperate it from nuclear DNA but that's why the CsCl gradient will help! I hope this helps.

We isolate mtDNA from snap-frozen mouse brain without much trouble, so I assume it should work also for human brain.

We basically thaw the brain in homogenization buffer (225 mM Mannitol, 75 mM sucrose, 10 mM Hepes pH 7,4, 10 mM EDTA) and then homogenize it in the same buffer (ca. 7 ml) using a dounce potter (we have several brands, all work fine with the tighter pestle). Afterwards twice 800g for 5 min to pellet nuclei etc., then 10 min 15.000g to pellet crude mitos. These mitos we overlay over a two-step sucrose gradient (0,8M sucrose and 1,2M sucrose in 10 mM HEPES pH 7,4, 10 mM EDTA) and ultrafuge for 1 hr at 50.000g. Collect the mitos between the two layers and extract from them the DNA as you want. We don't use RNase treatments since mtDNA tends to get damaged by it.

A mouse brain gives ca. 10-15 ug mitochondrial nucleics acids (how much of this is actually DNA I cannot tell... ), so with a 2x2x2cm piece of human brain you should be on the safe side. The purity depends a bit on your personal technique, but our preps are often >90% pure

Dont know if this is useful. There is a protocol for mtDNA extraction from insects: Mitochondrial dna variation in rhopalosiphum padi (homoptera: aphididae) populations from four spanish localities. ann. entomol. soc. am. 85:241-247