Usually, most of information gathered on a lithium-battery cell is done through capacity measurement, impedance spectroscopy or other common technics. One important characteristic, but few considered, is the entropy-variation of a cell. Indeed, there are no methodologies to efficiently exploit entropy variations to get information such as the State-Of-Charge (SOC) and the State-Of-Health (SOH). To tackle this issue, lithium-ion cells have been cycled and entropy variations have been measured for different SOH though potentiometric method. Then the entropy variations have been modeled as functions of the SOH using machine learning techniques.The final goal of this study is to show that the entropy-variation measurement can be an efficient tool for the knowledge of the battery state. This new methodology is going to lead to a safer use of batteries.To get information on battery state such as the State-Of-Health (SOH), different common technics are used. For instance it can be done through capacity measurement, impedance spectroscopy, internal resistance measurement or else. However, one important characteristic is seldom used: the entropy-variations of a battery cell. The reason is that there is no methodology to quantitatively link SOH to entropy-variations.To deal with this issue, different commercial cells have been cycled and entropy-variations profiles have been measured with the potentiometric method at different SOH. Machine learning techniques have then been considered to quantitatively estimate SOH form entropy-variations profiles.The purpose of this study is to show that entropy-variations measurement can be an efficient way to acquire a better knowledge on battery state. It is a new step toward a safer use of batteries.