Frequency-domain Near-infrared Diffuse Spectroscopy For The Diagnosis Of Cervical Cancer

Cervical cancer greatly threatens women's health for its high mortality rate and younger trend. While cervical cancer is curable if early diagnosis can be reached. The near infrared diffusion technology has been a hot topic on the inspection of early cancer because of its good penetration depth and non-invasive character. This dissertation presents a frequency-domain inspection system, in which the light source is modulated by a high frequency, and the diffusion reflex output from the tissue, i.e., attenuation of intensity (or modulation depth: M) and phase delay-φ, are detected. From the two values we can get the inverse result:μ_a,μ_s and g, which suggest the tissue's physiology status at that time.In this system light source is a high frequency (100MHz) modulated laser diode,λ=830nm, and the light output from the sample solid phantom (μ_a=0.1cm^-1,μ_s=110cm^-1,g=0.9) is detected by a photo multiplier tube (PMT) modulated by another frequency (100.001MHz). There will be a beat frequency (1kHz) signal output from the PMT. After a circuit changing current to voltage and signal process we can get what we want: M and ? .At first we get the bandwidth of the PMT by experiment. At the second we compare two signal process methods, and find the digital signal process is better than the analog method. The phase accuracy is about 0.02? and amplitude accuracy is about 0.2% when the digital collected signal's voltage is above 1V. The digital method meets well since similar system's inspection accuracy is about 0.1? and 0.4%[1]. The dissertation also shows light intensity measurement results at several different source-detector separation (9¹8mm) separately by direct current lights and 1kHz modulation light inspiring the phantom. Comparing the experimental data to that from Continuous Wave Monte Carlo (CW-MC) simulation, we find the error is no more than 1% at 9¹2mm, while increases after 12mm; the accuracy is within 0.03 except few points. At the same time we get the reconfiguration optical parameters,μ_a=0.09983cm^-1,μ_s=106.992cm^-1, with the inverse Monte Carlo simulation, the corresponding error is about 0.17%, 2.7%. At last, the dissertation gives analysis for the errors and shows the possible amelioration methods in the future.