Two‑photon calcium imaging is the standard tool for recording large neural populations in vivo, yet accurate spike inference across the growing variety of calcium indicators remains unresolved. Existing supervised methods achieve reasonable in‑domain accuracy but generalize poorly to unseen indicators because different indicators produce distinct fluorescence kinetics and signal statistics, while current architectures are simple generic temporal regressors lacking dynamics‑matched inductive bias. We introduce SpikeSSL, a universal spike‑inference framework whose temporal backbone consists of a bank of bidirectional IIR state‑space layers broadly motivated by calcium dynamics. A multi‑modal conditioning encoder maps indicator identity, sampling rate, and trace‑level signal statistics into a global conditioning vector that modulates the backbone via Adaptive Layer Normalization. A heteroscedastic variance head provides calibrated per‑frame uncertainty.
On a benchmark built from 33 public ground‑truth datasets with five fixed evaluation splits, SpikeSSL attains state‑of‑the‑art performance in both in‑domain and zero‑shot leave‑one‑indicator‑out settings. We also develop a biophysical simulation pipeline capable of systematically varying kinetic parameters, spike statistics, response nonlinearities, baseline drift, and noise to generate paired fluorescence‑spike traces. Augmenting training with ~11,000 simulated traces effectively closes the cross‑indicator domain gap and improves zero‑shot generalization. Code is publicly available at https://github.com/detimage123/SpikeSSL.
Review: By embedding calcium dynamics directly into the network and employing multi‑modal conditioning, SpikeSSL achieves robust cross‑indicator spike inference, offering a scalable unified platform for neural imaging analysis.