A minimum of 30 usable sessions is needed to have statistically significant data for webcam eye-tracking experiments, with a goal of getting 45 usable. In order to get this many usable sessions, you would need 100-150 complete participants. Keep in mind that about one-third of respondents will be disqualified during the process.
Why having 30 usables is statistically significant
We ran a sample validation test on Mashable.com with a Ford advertisement. Below are the results split up by the number of respondents. You can see that the heat maps with low base sizes contains large differences, but at about 20 respondents and up the heat map remains mainly the same.
With that being said, at 30 your results will be representative of your sample.