Single-photon avalanche diode imagers in biophotonics: review and outlook

Abstract

Single-photon avalanche diode (SPAD) arrays are solid-state detectors that offer imaging capabilities at the level of individual photons, with unparalleled photon counting and time-resolved performance. This fascinating technology has progressed at a very fast pace in the past 15 years, since its inception in standard CMOS technology in 2003. A host of architectures have been investigated, ranging from simpler implementations, based solely on off-chip data processing, to progressively "smarter" sensors including on-chip, or even pixel level, time-stamping and processing capabilities. As the technology has matured, a range of biophotonics applications have been explored, including (endoscopic) FLIM, (multibeam multiphoton) FLIM-FRET, SPIM-FCS, super-resolution microscopy, time-resolved Raman spectroscopy, NIROT and PET. We will review some representative sensors and their corresponding applications, including the most relevant challenges faced by chip designers and end-users. Finally, we will provide an outlook on the future of this fascinating technology.

Keywords: Biophotonics; Imaging and sensing.

Fig. 1. SPAD arrays and comparison of the SPAD pixel architectures.

a Artist’s impression of a SPAD array (top view) and b an example of the corresponding cross-section for a substrate isolated SPAD in a conventional CMOS process, depicting some of the key components (diode anode/cathode and corresponding p-n junction, multiplication region in which the avalanche is triggered, and the substrate and isolation from it). The SPAD fill factor can be enhanced with microlenses (c), and the inset shows an SEM image from ref. . The design of individual pixels ranges from d basic structures, which are only capable of generating digital pulses corresponding to individual photon arrivals on the SPAD, to e pixels including counters, which add the individual arrivals over a given time window that is possibly gated, or f more advanced electronics such as a complete TDC, which make it possible to time-stamp individual photon arrival times. The corresponding examples of pixel micrographs are displayed in g–i, as reprinted from refs. ^,,

Fig. 2. Comparison of the SPAD array architectures.

a In linear arrays, the pixel electronics can be placed outside the pixel area, leading to an increase in the fill factor; in 2D arrays, the fill factors are generally smaller, because b electronics is needed inside the pixel itself, or at least c at the array periphery, e.g. for column-based TDCs. The related advantages and disadvantages are discussed in detail in the text, and the corresponding examples of array micrographs can be found in d–f, as reprinted from refs. ^,,. Finally, g provides an overview of the evolution of SPAD imagers over the last 15 years in terms of the total number of pixels (on the vertical axis), the technology node (indicated at the top of the image), and some salient architectural characteristics, such as random access or event driven (indicated at the bottom of the image). Only some representative examples, primarily targeted at biophotonics applications, are shown here (details are reported in Table 2). The diagonal lines indicate the developments along a given technology node (800, 350 and 130 nm), which are usually started by optimising the SPADs before designing full imagers. Recent years have seen a trend towards higher spatial resolutions and 3D IC solutions

Fig. 3. Example fluorescence intensity and/or lifetime results.

a FluoCam system used in a point-like mode for the study of monomeric ICG-c(RGDfK) injected in a mouse with a glioblastoma mouse model. A subtle lifetime shift between tumour and non-tumour tissue is observed. b Dual-colour intensity fluorescence image of a thin slice of a plant root stained with a mixture of Safranin and Fast Green, taken with the SwissSPAD widefield time-domain gated array. c Triple-colour intensity fluorescence image of HeLa cells labelled with DAPI, Alexa 488 and Alexa 555, taken with SwissSPAD2. d, e Label-free FLIM of an unstained liver tissue excised from a tumourigenic murine model, imaged with a 64 × 4 SPAD array. f, g A Convallaria FLIM measurement performed with a linear 32 × 1 SPAD array. The images are reprinted from refs. ^,,,,

Fig. 4. Widefield SPIM-FCS images of monomeric eGFP oligomers in HeLa cells as recorded with a SwissSPAD widefield imager.

a Fluorescence intensity, b diffusion coefficient and c dye concentration. d Diffusion coefficients for three HeLa cells expressing different oligomers. e Particle concentration for the three HeLa cells with different oligomers. The images are reprinted from ref.

Fig. 5. SPAD super-resolution images.

a The first super-resolution image captured with SwissSPAD, compared to b EMCCD and c widefield images. The images show the microtubuli of an U2OS cell labelled with Alexa Fluor 647, in Vectashield. d, e Comparison of the SPCImager using “smart” aggregation and microlenses with an EMCCD. The images show multiple GATTA-PAINT 40G nanoruler localisations. f Comparison of the differences in localisation uncertainty with and without “smart” aggregation and the impact of the microlenses^,. g SwissSPAD super-resolution image of microtubuli labelled with Alexa 647 in OxEA buffer compared to h sCMOS and i widefield images. The white bar indicates 1 μm. The images are reprinted from refs. ^,,

Fig. 6. SPAD optical tomography images and applications.

a, b NIROT camera system prototype and measurements versus simulation results for a phantom. c, d Fluorescence molecular tomography (FMT) image as an overlap of the optical image obtained with the RadHard2 32 × 32 photon-counting sensor with the corresponding MRI image. C51 cells (a colon cancer-derived cell line) have been implanted in the flank of a mouse. A clear spread in the protease activity, indicated by the significantly higher fluorescence intensity in some parts of the tumour, is shown. c Complete MR + FMT image, and d zoom of the cancer region. The images are reprinted from refs. ^,

Fig. 7. Recent SPAD concepts for imagers revolve around 3D integration, possibly combined with microlenses to further maximise the fill factor.

a A 3D integration concept image, b a two-tier implementation with additional microlenses and c, d cross-sections of different imagers using three tiers^,. Frontside illumination is used in c, whereas backside illumination is used in b and d. The images b–d are reprinted from refs. ^,,

Fig. 8. SPAD system complexity vs. biophotonics applications and evolution of representative SPAD sensor figures of merit.

a Schematic overview of the SPAD-based system complexity, in terms of key functionalities (counting/gating/time-stamping) versus the main biophotonics applications. b–f Overview of the representative SPAD sensor figures of merit as a function of the main target applications, based on data from Table 2: b total number of SPADs (corresponding to the effective spatial resolution in the imagers) versus time; c–e total number of SPADs, PDE and DCR per unit area grouped based on the application types (dashed lines: individual sensors, top/bottom of each box: maximum/minimum); and f the DCR per unit area versus the PDE