Fig. 1. AL-SIM pipeline. (a) The first training stage, (b) the first test stage, (c) the second training stage, and (d) the second test stage.

Fig. 2. AL-SIM network architecture. The network is based on a U-Net structure with two parallel branches. The middle purple section represents the encoder pathway that processes the raw SIM data through progressive downsampling (via 2 × 2 max pooling operations shown in red arrows), reducing spatial dimensions while increasing feature channels. The yellow sections above and below function as decoders that predict the light fields and emitter distributions, respectively, through systematic upsampling (via bilinear interpolation shown in green arrows), which restores spatial resolution. The architecture features skip connections (gray arrows) that connect the encoder to both decoder pathways, allowing high-resolution features to bypass the bottleneck and preserve spatial information.

Fig. 3. Experimental setup. (a) Optical path of SIM. (b) Formation of different illumination light fields and phase modulation.

Fig. 4. Simulation validation of AL-SIM under light field distortions. (a) Distorted illumination patterns (left), simulated filament structures (middle), and raw SIM images (right). (b) Illumination patterns predicted by the AL-SIM first-stage model. (c) Comparative reconstruction results: (i) AL-SIM, (ii) DL-SIM, (iii) ground truth, and (iv) wide-field microscopy. Scale bars: 1 µm (main panels); 200 nm (magnified insets).

Fig. 5. Comparative experimental results. (a) Comparison between AL-SIM and wide-field microscopy. (b) Estimated light field patterns (Est LF) derived from experimental raw data [Raw (Exp)] by AL-SIM, which are then combined with simulated emitter data to generate synthetic training samples [Raw (Simulation)] for the second-stage network. (c) Imaging of region 1 using AL-SIM, DL-SIM, conventional SIM, and wide-field microscopy. (d) Spatial resolution comparison along the white dashed line, highlighting that AL-SIM outperforms DL-SIM and conventional SIM in terms of artifacts. (e) Logarithmic Fourier spectra of imaging using AL-SIM, DL-SIM, conventional SIM, and wide-field microscopy. (f) Imaging of region 1 using different techniques and the resolution improvement along the white dashed line. (g) Imaging of region 2 using different techniques and the resolution improvement along the white dashed line [(a) scale bar: 10 µm; (b)–(d) scale bar: 2 µm].