Oxygenation is an important marker in many clinical settings, e.g. diagnosing ischaemic brain injuries in preterm infants or determining treatment effectiveness in cancer patients. Despite significant efforts to determine the oxygenation state of the human tissue with conventional imaging modalities, e.g. positron emission tomography (PET) and magnetic resonance imaging (MRI), a method which is fit for continuous, routine use in clinics does not exist.

Near infrared optical tomography (NIROT) is a compelling alternative which can be employed to measure tissue oxygenation. This method is based on the illumination and subsequent detection of the tissue response to light in the near infrared (NIR) spectrum. NIROT is non-invasive and safe for continuous monitoring of patients. However, conventional NIROT systems have demonstrated a limited spatial resolution, in the 1-2 cm range, and prevented widespread clinical uptake. A major reason for this limited resolution is the low numbers of sources and detectors in conventional systems.

In an effort to improve the spatial resolution of NIROT, researchers have recently applied time-resolved cameras based on single photon avalanche diodes (SPADs) to NIROT phantom measurements, achieving a resolution of 5 mm. Despite these promising results, conventional SPAD cameras are unsuitable for clinical measurements due to a slow image acquisition time. In this thesis, time-resolved cameras were developed which have the potential to perform image acquisitions in NIROT measurements for a multi-source multi-wavelength system in a number of minutes. The main objective was to develop a large format, 252 × 144 pixel, time-resolved SPAD camera capable of wide field measurements in a clinical setting.

Two new pixel circuits were developed in a backside-illuminated (BSI) 3D IC technology to increase the signal-to-noise ratio (SNR) and dynamic range (DR) in time-resolved measurements. The first circuit demonstrates, for the first time, a technique to increase the excess bias range, and thus the photon detection efficiency (PDE) and timing performance of conventional SPAD pixels. Coupled to an active recharge circuit, the pixel achieves a minimum dead time of 8 ns, and is thus suitable for high DR measurements. The second presents the first pixel circuit able to interface with a SPAD via the anode or the cathode terminal, thus enabling the possibility of a general purpose time-correlated single-photon counting (TCSPC) die connecting to multiple application specific photodetector dies.

A new time-resolved SPAD sensor architecture is presented which employs a time-to-digital converter (TDC) sharing architecture to achieve both high PDE and high throughput parallel measurements. A 32 × 32 sensor based on this architecture is produced in a 180nm CMOS technology. At the maximum throughput, 10^6 photons can be obtained for every pixel in the array in parallel in 4.6 seconds.

Finally, a wide-field, 252 × 144 pixel time-resolved sensor is presented. To maintain a fast acquisition speed the architecture includes a per-pixel integrated histogramming readout. This enables the compression of the readout data by up to a factor of 14.9. To the best of the author¿s knowledge, this is the first implementation of integrated histogramming on a per-pixel basis for a full sensor array. This new sensor has game changing potential for NIROT, opening the door to high resolution wide-field clinical measurements.