The data rate requirements for raw data transmission through the slip ring of a CT scanner can be quite
challenging. While lossy compression can offer a significant reduction in data set size, it introduces errors
to the measurements. The design and evaluation of compression methods need to take into account the
nature of the sinogram. In this paper, we present two compression noise shaping methods for fixed rate lossy
compression of CT raw data that achieve a lower error level in the center region of the reconstructed image:
error feedback filters and sub-band coding with bit allocation.
As CT scanners continue to increase in speed and to collect more data per rotation, transmitting raw data across the slip
ring and storing raw data can be very challenging. While lossy compression can offer a significant reduction in data set
size, it introduces errors to the measurements. We examined the effect of noise of different frequencies in the view
(temporal) direction as well as at different locations in the detector arc. Our results showed that only low temporal
frequency errors in the center detectors can contribute to errors in the center of the reconstructed field-of-view (FOV).
On the other hand, high temporal frequency errors only contribute to errors in the periphery of the reconstructed FOV.
Whether image errors arise from compression or electronic noise, their relative sensitivity to different frequencies and
detectors is an important consideration for applications such as cardiac CT, where the center of the FOV may be
considered the most critical region. Therefore, when limiting data rate is essential, detectors could be allocated different
bit-rates for compression based on the tolerable frequency content of their errors and their spatial location.
In this paper, we present new Adaptive and Robust Techniques (ART) for microwave-based thermoacoustic
tomography (TAT) and laser-based photo-acoustic tomography (PAT), and study their performances for breast
cancer detection. TAT and PAT are emerging medical imaging techniques that combine the merits of high
contrast due to electromagnetic or laser stimulation and high resolution offered by thermal acoustic imaging.
The current image reconstruction methods used for TAT and PAT, such as the widely used Delay-and-Sum
(DAS) approach, are data-independent and suffer from low resolution, high sidelobe levels, and poor interference
rejection capabilities. The data-adaptive ART can have much better resolution and much better interference
rejection capabilities than their data-independent counterparts. By allowing certain uncertainties, ART can
be used to mitigate the amplitude and phase distortion problems encountered in TAT and PAT. Specifically,
in the first step of ART, RCB is used for waveform estimation by treating the amplitude distortion with an
uncertainty parameter. In the second step of ART, a simple yet effective peak searching method is used for phase
distortion correction. Compared with other energy or amplitude based response intensity estimation methods,
peak searching can be used to improve image quality with little additional computational costs. Moreover, since
the acoustic pulse is usually bipolar: a positive peak, corresponding to the compression pulse, and a negative
peak, corresponding to the rarefaction pulse, we can further enhance the image contrast in TAT or PAT by using
the peak-to-peak difference as the response intensity for a focal point. The excellent performance of ART is
demonstrated using both simulated and experimentally measured data.