Computer-aided diagnosis of rheumatoid arthritis with optical tomography, Part 1: feature extraction.

This is the first part of a two-part paper on the application of computer-aided diagnosis to diffuse optical tomography (DOT). An approach for extracting heuristic features from DOT images and a method for using these features to diagnose rheumatoid arthritis (RA) are presented. Feature extraction is the focus of Part 1, while the utility of five classification algorithms is evaluated in Part 2. The framework is validated on a set of 219 DOT images of proximal interphalangeal (PIP) joints. Overall, 594 features are extracted from the absorption and scattering images of each joint. Three major findings are deduced. First, DOT images of subjects with RA are statistically different (p<0.05) from images of subjects without RA for over 90% of the features investigated. Second, DOT images of subjects with RA that do not have detectable effusion, erosion, or synovitis (as determined by MRI and ultrasound) are statistically indistinguishable from DOT images of subjects with RA that do exhibit effusion, erosion, or synovitis. Thus, this subset of subjects may be diagnosed with RA from DOT images while they would go undetected by reviews of MRI or ultrasound images. Third, scattering coefficient images yield better one-dimensional classifiers. A total of three features yield a Youden index greater than 0.8. These findings suggest that DOT may be capable of distinguishing between PIP joints that are healthy and those affected by RA with or without effusion, erosion, or synovitis.

Computer-aided diagnosis of rheumatoid arthritis with optical tomography, Part 2: image classification.

This is the second part of a two-part paper on the application of computer-aided diagnosis to diffuse optical tomography (DOT) for diagnosing rheumatoid arthritis (RA). A comprehensive analysis of techniques for the classification of DOT images of proximal interphalangeal joints of subjects with and without RA is presented. A method for extracting heuristic features from DOT images was presented in Part 1. The ability of five classification algorithms to accurately label each DOT image as belonging to a subject with or without RA is analyzed here. The algorithms of interest are the k-nearest-neighbors, linear and quadratic discriminant analysis, self-organizing maps, and support vector machines (SVM). With a polynomial SVM classifier, we achieve 100.0% sensitivity and 97.8% specificity. Lower bounds for these results (at 95.0% confidence level) are 96.4% and 93.8%, respectively. Image features most predictive of RA are from the spatial variation of optical properties and the absolute range in feature values. The optimal classifiers are low-dimensional combinations (<7 features). These results underscore the high potential for DOT to become a clinically useful diagnostic tool and warrant larger prospective clinical trials to conclusively demonstrate the ultimate clinical utility of this approach.

Frequency-domain optical tomographic imaging of arthritic finger joints.

We are presenting data from the largest clinical trial on optical tomographic imaging of finger joints to date. Overall we evaluated 99 fingers of patients affected by rheumatoid arthritis (RA) and 120 fingers from healthy volunteers. Using frequency-domain imaging techniques we show that sensitivities and specificities of 0.85 and higher can be achieved in detecting RA. This is accomplished by deriving multiple optical parameters from the optical tomographic images and combining them for the statistical analysis. Parameters derived from the scattering coefficient perform slightly better than absorption derived parameters. Furthermore we found that data obtained at 600 MHz leads to better classification results than data obtained at 0 or 300 MHz.

Computer-aided interpretation approach for optical tomographic images.

A computer-aided interpretation approach is proposed to detect rheumatic arthritis (RA) in human finger joints using optical tomographic images. The image interpretation method employs a classification algorithm that makes use of a so-called self-organizing mapping scheme to classify fingers as either affected or unaffected by RA. Unlike in previous studies, this allows for combining multiple image features, such as minimum and maximum values of the absorption coefficient for identifying affected and not affected joints. Classification performances obtained by the proposed method were evaluated in terms of sensitivity, specificity, Youden index, and mutual information. Different methods (i.e., clinical diagnostics, ultrasound imaging, magnet resonance imaging, and inspection of optical tomographic images), were used to produce ground truth benchmarks to determine the performance of image interpretations. Using data from 100 finger joints, findings suggest that some parameter combinations lead to higher sensitivities, while others to higher specificities when compared to single parameter classifications employed in previous studies. Maximum performances are reached when combining the minimum/maximum ratio of the absorption coefficient and image variance. In this case, sensitivities and specificities over 0.9 can be achieved. These values are much higher than values obtained when only single parameter classifications were used, where sensitivities and specificities remained well below 0.8.

Influence of oxygen saturation on the optical scattering properties of human red blood cells in the spectral range 250 to 2,000 nm.

The intrinsic optical parameters absorption coefficient mu(a), scattering coefficient micros, anisotropy factor g, and effective scattering coefficient micros were determined for human red blood cell (RBC) suspensions of hematocrit 33.2% dependent on the oxygen saturation (SAT O(2)) in the wavelength range 250 to 2,000 nm, including the range above 1,100 nm, about which there are no data available in the literature. Integrating sphere measurements of light transmittance and reflectance in combination with inverse Monte Carlo simulation were carried out for SAT O(2) levels of 100 and 0%. In the wavelength range up to 1,200 nm, the absorption behavior is determined by the hemoglobin absorption. The spectral range above the cells' absorption shows no dependence on SAT O(2) and approximates the absorption of water with values 20 to 30% below the respective values for water. Parameters micros and g are significantly influenced by the SAT O(2)-induced absorption changes. Above 600 nm, micros decreases continuously from values of 85 mm(-1) to values of 30 mm(-1) at 2,000 nm. The anisotropy factor shows a slight decrease with wavelengths above 600 nm. In the spectral regions of 1,450 and 1,900 nm where water has local absorption maxima, g shows a significant decrease down to 0.85, whereas micros increases.

Multiparameter classifications of optical tomographic images.

This research study explores the combined use of more than one parameter derived from optical tomographic images to increase diagnostic accuracy which is measured in terms of sensitivity and specificity. Parameters considered include, for example, smallest or largest absorption or scattering coefficients or the ratios thereof in an image region of interest. These parameters have been used individually in a previous study to determine if a finger joint is affected or not affected by rheumatoid arthritis. To combine these parameters in the analysis we employ here a vector quantization based classification method called Self-Organizing Mapping (SOM). This method allows producing multivariate ROC-curves from which sensitivity and specificities can be determined. We found that some parameter combinations can lead to higher sensitivities whereas others to higher specificities when compared to singleparameter classifications employed in previous studies. The best diagnostic accuracy, in terms of highest Youden index, was achieved by combining three absorption parameters [maximum(micro a), minimum(micro a), and the ratio of minimum(micro a) and maximum(micro a)], which result in a sensitivity of 0.78, a specificity of 0.76, a Youden index of 0.54, and an area under the curve (AUC) of 0.72. These values are higher than for previously reported single parameter classifications with a best sensitivity and specificity of 0.71, a Youden index of 0.41, and an AUC of 0.66.

Optimal source-modulation frequencies for transport-theory-based optical tomography of small-tissue volumes.

In frequency-domain optical tomography (FDOT) the quality of the reconstruction result is affected by the choice of the source-modulation frequency. In general the accuracy of the reconstructed image should improve as the source-modulation frequency increases. However, this is only true for noise-free data. Experimental data is typically corrupted by noise and the accuracy is compromised. Assuming the validity of the widely used shot noise model, one can show that the signal-to-noise ratio (SNR) of the amplitude signal decreases with increasing frequency, whereas the SNR of the phase shift reaches peak values in the range between 400 MHz and 800 MHz. As a consequence, it can be assumed that there exists an optimal frequency for which the reconstruction accuracy would be highest. To determine optimal frequencies for FDOT, we investigate here the frequency dependence of optical tomographic reconstruction results using the frequency-domain equation of radiative transfer. We present numerical and experimental studies with a focus on small tissue volumes, as encountered in small animal and human finger imaging. Best reconstruction results were achieved in the 600-800 MHz frequency range.

Multipixel system for gigahertz frequency-domain optical imaging of finger joints.

Frequency-domain optical imaging systems have shown great promise for characterizing blood oxygenation, hemodynamics, and other physiological parameters in human and animal tissues. However, most of the frequency domain systems presented so far operate with source modulation frequencies below 150 MHz. At these low frequencies, their ability to provide accurate data for small tissue geometries such as encountered in imaging of finger joints or rodents is limited. Here, we present a new system that can provide data up to 1 GHz using an intensity modulated charged coupled device camera. After data processing, the images show the two-dimensional distribution of amplitude and phase of the light modulation on the finger surface. The system performance was investigated and test measurements on optical tissue phantoms were taken to investigate whether higher frequencies yield better signal-to-noise ratios (SNRs). It could be shown that local changes in optical tissue properties, as they appear in the initial stages of rheumatoid arthritis in a finger joint, are detectable by simple image evaluation, with the range of modulation frequency around 500 MHz proving to yield the highest SNR.

Development of a finger joint phantom for evaluation of frequency domain measurement systems.

For development and test of new optical imaging devices, phantoms are widely used to emulate the tissue to be imaged. Phantom design gets more difficult the more complex the tissue is structured. We report on developing and testing a solid, stable finger joint phantom to simulate transillumination of finger joints in frequency-domain imaging systems. The phantom consists of the bone, capsule, skin, the capsule volume, and the joint gap. Silicone was used to build the solid parts and a glycerol-water solution for the fluid in the capsule volume and joint gap. The system to test the phantom is an optical frequency-domain scanning set-up. Different stages of joint inflammation as they occur in rheumatoid arthritis (BA) were emulated by assembling the phantom with capsule and fluid having different optical properties. Reliability of the phantom measurement was investigated by repeated assembling. The results show clear discrimination between different stages of joints within the signal deviation due to reassembling of the phantom.

Comparison of two and three-dimensional optical tomographic image reconstructions of human finger joints.

We have developed an images reconstruction algorithm to recover spatial distribution of optical properties in human finger joints for early diagnosis and monitoring of rheumatoid arthritis (RA). An optimization method iteratively employs a light propagation and scattering coefficients distribution for near-infrared (NIR) light inside the joint tissue. We developed the differences in cross-sectional images obtained by using the reconstruction algorithms with 2-dimensional and 3-dimensional light propagation models. In particular we examined how these different approaches affect the discrimination between healthy and RA joints.

Sagittal laser optical tomography for imaging of rheumatoid finger joints.

We present a novel optical tomographic imaging system that was designed to determine two-dimensional spatial distribution of optical properties in a sagittal plane through finger joints. The system incorporates a single laser diode and a single silicon photodetector into a scanning device that records spatially resolved light intensities as they are transmitted through a finger. These data are input to a model-based iterative image reconstruction (MOBIIR) scheme, which uses the equation of radiative transfer (ERT) as a forward model for light propagation through tissue. We have used this system to obtain tomographic images of six proximal interphalangeal finger joints from two patients with rheumatoid arthritis. The optical images were compared to clinical symptoms and ultrasound images.