An early tool of medical optical imaging was the "oximeter," devised in the 1930s to detect the amount of oxygen in blood by measuring the ratio of the light absorbed at two wavelengths. Great improvements to this concept came in the 1970s with the advent of microprocessors and light-emitting diodes that permitted the use of many more wavelengths, thus allowing measurement of the absolute amount of oxygen and elimination of background effects. Assessment of the oxygen content of arterial blood through such methods has become a major diagnostic tool for studying acutely ill patients.
The potential of imaging with light was reinforced with the successful application of optical tools to determine the levels of oxygen in the brain of a cat. Later, this concept was used in monitoring brain and muscle oxygenation in humans, as well as in other applications.
Optical imaging devices are now used in the neonatal clinic at the Stanford University Medical Center, the University Hospital in London, and possibly elsewhere, and Hamamatsu Corp., in Japan, produces an infrared spectrometer for bedside monitoring of oxygenation in the brains of babies.
Objects have been illuminated for medical imaging by the continuous beam, time-resolved beam, and phase-modulated beam methods. The last method has the advantage of allowing measurement of the mean optical pathlength without the size and cost problems associated with ultrashort light pulses and a fast optical detector. The phase-modulated beam as a measure of optical length has proved to be very useful in quantifying absorption and scattering coefficient measurements.
Data collection methods can be divided into two broad classes: (1) those that try to isolate photons undergoing no or very little scattering from source to detector and that thus may be able to use simpler reconstruction algorithms because the path of each photon is considered known; and (2) those that collect light over a longer period, with or without taking into account photon arrival times.
Preliminary mathematical analysis indicates that the preferred approach is to sort photon counts by arrival time and then proceed to solve the complex nonlinear inversion problem.