Signal Processing

The Signal Processing & Data Inference team interprets these data, extracting clinically-meaningful information and presenting it in the most effective manner to the clinicians.

The Proteus team develops novel optical imaging platforms using a range of detectors. Whether used alone, or in conjunction with our bespoke SmartProbes, these platforms generate vast amounts of data from a range of biological models.

To date, the Signal Processing & Data Inference team has concentrated its efforts into data pre-processing (reconstructing and enhancing the raw data) as well as post-processing approaches (extracting clinically valuable information from the pre-processed data). In particular, areas of interest include (but are not limited to):

Data compression

The amount of data generated on our custom imaging platforms require effective data compression approaches in order to enable real-time operation. This is especially needed for  imaging modes where variations of emitted light across the electromagnetic spectrum, as well as fluorescence variation over time, is recorded for each image coordinate.

Image restoration

Restoration algorithms can compensate for a range of fibre, optics and detector artefacts (see figures below).  Well recognised artifacts include:

  • Inter-core coupling: where light spreads from a core to its neighbours within a multi-core fibre bundle, resulting in blurring of the imaged scene and consequently reduction of the sensing performance of the device.
  • Optical blur: where limited magnification capabilities of the optics within the device result in further blurring of the imaged scene.
  • Irregular sampling: where the irregular distribution and optical properties of fibre cores within the multi-core fibre bundle result in an undesired honeycomb pattern projected in the imaging sensor.

Figure a: Black and white image of a USAF 1951 target with the locations of a fibre cores irregularly distributed across the field of view.
Figure b: An image of the same target as imaged through a multi-core fibre bundle, with all the associated artefacts (i.e. blurring, honeycomb patterns, etc.).

Image understanding, analysis and quantification

Multiple channels of information, including images, variations in emission rate and spectrum of the fluorescent light can be incorporated in a diverse range of data analysis frameworks. Such analysis frameworks can effectively improve the understanding and quantification of the underlying anatomical structure and function.

Data visualisation

The effective and efficient visualisation of multiple (and potentially complex) data streams is essential for the successful integration of the developed imaging platforms in the clinic.