Recent advancement in neuronal imaging leads to accurate measurement of cellular activity after treatment with drugs. However, neuronal activity obtained by dynamic imaging reveals complex pattern with time, the prediction of activity level corresponding to drug level remains challenging. In this context, we apply self organizing map (SOM) to estimate the drug dose according to neuronal activity level in four different time windows. Here we cluster the neuronal activity pattern and classify the activity pattern with drug dose. We also implement supervised SOM to predict the drug dose corresponding to activity pattern. The advantage of using SOM is that it is a great visualization and prediction tool to analyze high-dimensional data onto a low-dimensional (1D or 2D) SOM grid. Such computaional tool can be used to classify unknown neuronal responses according to the extent of drug level present in the system. © 2020 IEEE.