If you want that kind of accuracy over that temperature range, than an RTD, as KISS mentioned, would likely be the best.
But you will also need careful design to meet that accuracy.
What temperature range? Even a Class A Pt RTD is just barely that accurate around 0C. By +/- 50C, it has exceeded that error. Do you need accuracy? Or repeatability?
How are you planning calibrating it to 0.2 degrees?, you will also have to calibrate it across the entire range as it's not linear, and use a lookup table to convert the readings. Repeatability is much more likely than accuracy.
I'm somewhat dubious about getting 0.2 degrees accuracy?, would 0.5 degrees not do? - if so you could use the DS18B20 which requires no calibration.
Then do it, accuracy is down to how well you build it, and even more so to how well you calibrate it - at least you have a nice low range you need to cover, so the calibration lookup table won't be too big.