A resting brain is not silent: its regions talk to each other, and these conversations can be eavesdropped on with a scanner. Usually, scientists track signal strength — how loudly each area 'sounds.' But a conversation also has rhythm: some areas light up simultaneously, others with a delay. Rhythm is captured by comparing the phases of oscillations, like in music, where not only the notes matter but also the alignment of beats.

The new method Neural Network Oscillations combines both approaches: signal strength and phase coherence. Strength is measured with a sliding window, short time segments, to see how connections change. Phases are compared directly: do activity peaks in two areas coincide? These two views complement each other like text and intonation: separately you can grasp the meaning, but together you get it more precisely.

Testing showed that the dual approach outperforms previous models in finding both autism and depression. And the explanation via SHAP (a method that shows each feature's contribution to the decision) confirmed that both types of features matter — strength and phase. This means disease alters not only the volume but also the rhythm of conversations between brain regions.

Such a tool could help clinicians spot abnormalities earlier and tailor treatments more precisely. The brain is no longer a black box: now we can hear where exactly its internal orchestra goes off beat.