Structured Detection for Enhanced Resolution and Optical Sectioning Microscopy

Image scanning microscopy (ISM) is revolutionizing laser scanning fluorescence microscopy. Replacing the bucket detector used in conventional confocal systems with an array of detectors provides access to additional spatial information below the diffraction scale [1]. Single photon avalanche diode (SPAD) array detectors are becoming the tool of choice for ISM thanks to their single-photon sensitivity and excellent temporal resolution, which enables the measurement of fast fluorescence dynamics [2]. ISM systems produce a multi-dimensional dataset, which can be seen as a collection of confocal-like images, one for each sensitive element of the detector array. Therefore, the reconstruction of a single image requires a tailored algorithm. Traditional approaches, such as adaptive pixel reassignment (APR) and multi-image deconvolution, enhance the lateral resolution, surpassing the diffraction limit up to a factor of two [3]. However, these methods underperform with thick samples, which produce a defocused background that reduces the contrast and hides high-resolution details. The conventional solution consists of limiting the size of the detector, which acts as a pinhole. However, it cannot fully remove out-of-focus light and deteriorates the in-focus signal-to-noise ratio (SNR). In this work, we demonstrate that the raw ISM dataset inherently informs about the axial position of the emitter, paving the way for computational optical sectioning. We propose a comprehensive algorithm - named super-resolution optical sectioning ISM (s2ISM) - to invert the physical image formation model of ISM. Our algorithm discards out-of-focus light and reconstructs only the in-focus plane while enhancing the SNR and lateral resolution.