In the interest of determining the relative merits of different instruments and reconstruction techniques, we wish to encourage the exchange of data between research groups working in the biomedical optics community. Investigators are invited to evaluate their own reconstruction algorithms using data acquired with the UCL 32-channel time-resolved imager (MONSTIR).

Data format

The temporal point spread functions (TPSFs) acquired for each source-detector combination are available as individual ASCII files. Each TPSF file consists of between one and four thousand pairs of numbers t, i(t), where t is the time (in picoseconds) following a laser pulse first entering the object, and i(t) is proportional to the number of photons detected within a 5 picosecond interval centered on that time. Note that absolute values of i(t) are not calibrated for variation in detector efficiency, and we recommend that investigators do NOT attempt to derive un-normalized amplitude information from our data. Reasons for avoiding the use of absolute intensity measurements are outlined below in the discussion on data types. The specific source and detector used to acquire each TPSF is identified in the corresponding filename. The absolute positions of all the sources and detectors on the surface of the object is provided in a so-called QM file. Please click here for a detailed description of the format of this file.

The UCL TOAST reconstruction algorithm makes use of specific data types derived from the TPSFs. Our data type files are also available for download. Each consists of an ASCII list of numbers, one number for each acquired TPSF. The ordering of the data within this file is specified within the corresponding QM file, which also lists the source and detector positions. If you are still uncertain about how to interpret this format, please contact us by email.

Data types

Photon interactions at the surface adjacent to a source or detector will have an overwhelming influence on measurements of absolute intensity. Consider, for example, the effect a hair or a skin blemish immediately below a source fiber would have on intensity, compared to the effect of more interesting structure deep below the skin! One potential way to avoid this problem is to acquire data at two (or more) wavelengths at which the deep structure has highly contrasting properties. Alternatively, one can avoid the use of absolute intensity measurements by employing normalized time-domain data types. This is what we have attempted to do at UCL. So far, our TOAST reconstruction algorithm has been adapted to accomodate all of the following:

• Temporal Moments. (Also known as Mellin Transforms).
• Central Moments. (Moments about the mean).
• Normalized Laplace Transforms.
• Mellin-Laplace Transforms. (A combination of Mellin and Laplace).

These are calculated directly from the experimental TPSFs for each source-detector combination. For more information about these data types, see publications by Schweiger and Arridge [1,3], Hebden et al [2], and Hillman et al [4].

Experiment list

Below is a list of the experiments which have resulted in data available for download. (The first data set became on-line in February 1999, and we anticipate that further data will be added on a regular basis during subsequent months). Each link provides a brief description of the experiment, and a further link to download the corresponding data and QM files.

1. Cylindrical phantom: 16 sources, 16 detectors. Time-resolved data types.
2. Cone phantom: 32 sources, 32 detectors. Time-resolved data types.

References

1. Schweiger, M, and Arridge, SR (1997): In Information Processing in Medical Imaging, Editors: Duncan, J, and Gindi, G, Lecture Notes in Computer Science, 1230, (Springer-Verlag: Berlin) 71-84.
2. Hebden, JC, Arridge, SR, and Schweiger, M (1998): OSA Trends in Optics and Photonics on Advances in Optical Imaging and Photon Migration, Editors: Alfano, RR, and Fujimoto, JG, (OSA, Washington DC) 21, 162-167.
3. Schweiger, M, and Arridge, SR (1999): Application of temporal filters to time resolved data in optical tomography, Physics in Medicine and Biology 44, 1699-1717.
4. Hillman, EMC, Hebden, JC, Schmidt, FEW, Arridge, SR, Schweiger, M, Dehghani, H, and Delpy, DT (2000): Calibration techniques and datatype extraction for time-resolved optical tomography, Review of Scientific Instruments 71(9), 3415-3427.

Return to BORL Research Topics.
Return to BORL Homepage.

Last update: June 12, 2001