dGMI was measured using a structured illumination created with a 405 nm laser source (Stradus 405-250, Vortran). The structured illumination was combined with other light sources using a dichroic mirror (DM1, ZT405rdc, Chroma) and focused on the flow channel using a 20× objective (OBJ1, UPLSAPO 20×, Olympus). The transmitted structured illumination was collected with a 10× objective (OBJ2, UPLFLN 10×, Olympus) and separated from other transmitted light using a dichroic mirror (DM6, ZT405rdc, Chroma). This was further passed through a 20× objective (OBJ3, UPLSAPO 20×), and in front of the objective at a distance of ~350 mm, an iris was put at an aperture size of ~3 mm. The light that entered the iris was detected with a PMT (PMT4) with a bandpass filter (405 nm, ET405/10×, Chroma) in front of it. ssGMI was measured using the same structured illumination as dGMI. However, instead of collecting the transmitted light, the scattered light from the flow channel orthogonal to the illumination light was collected using a plano-convex lens (f=50 mm, Thorlabs). Then it was separated from other scattered light using a dichroic mirror (DM7, ZT405rdc, Chroma) and detected with a PMT (PMT5) with a bandpass filter (ET405/10×, Chroma) in front of it. Fluorescence labels for training and validation were measured with either the 405 nm structured illumination, a 488 nm laser (Cobolt), or a 635 nm laser (Stradus 637-140, Vortran). The 488 nm laser and 637 nm laser were combined with a dichroic mirror (DM3, ZT488rdc), reflected with another dichroic mirror (DM2, ZT405/488/635rpc) for separating the returning fluorescence light, and combined with the 405 nm structured illumination with DM1. The three light sources were focused on the flow channel using OBJ1. The fluorescence was collected using the same objective and passed through DM1 and DM2. This was further separated into different colors with dichroic mirrors DM4 and DM5 and detected using PMTs (PMT1, PMT2, and PMT3) with different bandpass filters (ET440/40×, ET525/50m, ET675/50m, respectively, all from Chroma) in front of them. In the case of WBC differential classification, a similar setup for four-color detection using three dichroic mirrors and four PMTs was used. Forward scatter (FSC) and side scatter (SSC) for triggering acquisition and initial cell population gating were measured using the 488 nm laser or the 637 nm laser. The figure is depicted for the case of the 488 nm laser. For the FSC, from the flow channel, the transmitted 488 nm light was collected with OBJ2 and passed through DM6. Then the main transmitted beam was blocked with an obscuration bar, and the scattered light that came around the obscuration bar was collected with a lens and detected with a photodetector (PDA100A, Thorlabs) with a bandpass filter (ZET488/10×, Chroma) in front of it. For the SSC, the scattered light from the flow orthogonal to the illumination light was collected using a plano-convex lens (f=50 mm, Thorlabs). Then it was passed through DM7 and collected with a PMT (PMT6) with a bandpass filter (ZET488/10×, Chroma) in front of it. dGMI, diffractive ghost motion imaging; iSGC, in silico-labeled ghost cytometry; PMT, photomultiplier tube; ssGMI, XXX.