Neuromodulators such as monoamines or neuropeptides play crucial roles in modulating diverse brain functions. They can be released phasically, i.e. in response to stimuli, or tonically, in a spontaneous and more discrete manner. Understanding how these spontaneous releases occur is challenging yet essential, as their disruption is associated with behavioral and psychiatric disorders. A paradigm shift in this domain has been the generation of fluorescent probes sensitive enough to detect, using 2-photons microscopy, discrete tonic release of specific neuromodulators including dopamine, noradrenaline or oxytocin. Yet, analysis of high sensitivity 2-photons microscopy to optical signals presents a significant challenge, as these appear randomly, anywhere and at any time, with low signal-to-noise. This kind of analysis in terms of spatiotemporal dynamics is very complex due to the variability in signal size, shape, frequency, location, potential displacement, or propagation. This complicates the pattern matching of fluorescent signals indicative of neurotransmitter activity. In response to these challenges, we propose an automated fluorescent signals extraction and analysis framework based on machine learning and image processing principles, which enables accurate detection of the neurotransmitter fluorescence signal through background denoising, object segmentation and multiple object tracking. Our algorithm allowed to seamlessly analyze 2-photons imaging of monoamine spontaneous releases from neurons and astrocytes.The versatility of this algorithm exceeds 2-photons imaging data and demonstrates robust detection efficiency on confocal and in vivo datasets as well as on calcium imaging recordings.