Multi-scale hyperspectral imaging of cervical neoplasia

Discrimination between normal and necrotic small intestinal tissue using hyperspectral imaging and unsupervised classification

Objective and automatic clinical discrimination of normal and necrotic sites of small intestinal tissue remains challenging. In this study, hyperspectral imaging (HSI) and unsupervised classification techniques were used to distinguish normal and necrotic sites of small intestinal tissues. Small intestinal tissue hyperspectral images of eight Japanese large-eared white rabbits were acquired using a visible near-infrared hyperspectral camera, and K-means and density peaks (DP) clustering algorithms were used to differentiate between normal and necrotic tissue. The three cases in this study showed that the average clustering purity of the DP clustering algorithm reached 92.07% when the two band combinations of 500-622 and 700-858 nm were selected. The results of this study suggest that HSI and DP clustering can assist physicians in distinguishing between normal and necrotic sites in the small intestine in vivo.

Comparison of preprocessing techniques to reduce nontissue-related variations in hyperspectral reflectance imaging

SIGNIFICANCE

Hyperspectral reflectance imaging can be used in medicine to identify tissue types, such as tumor tissue. Tissue classification algorithms are developed based on, e.g., machine learning or principle component analysis. For the development of these algorithms, data are generally preprocessed to remove variability in data not related to the tissue itself since this will improve the performance of the classification algorithm. In hyperspectral imaging, the measured spectra are also influenced by reflections from the surface (glare) and height variations within and between tissue samples.

AIM

To compare the ability of different preprocessing algorithms to decrease variations in spectra induced by glare and height differences while maintaining contrast based on differences in optical properties between tissue types.

APPROACH

We compare eight preprocessing algorithms commonly used in medical hyperspectral imaging: standard normal variate, multiplicative scatter correction, min-max normalization, mean centering, area under the curve normalization, single wavelength normalization, first derivative, and second derivative. We investigate conservation of contrast stemming from differences in: blood volume fraction, presence of different absorbers, scatter amplitude, and scatter slope-while correcting for glare and height variations. We use a similarity metric, the overlap coefficient, to quantify contrast between spectra. We also investigate the algorithms for clinical datasets from the colon and breast.

CONCLUSIONS

Preprocessing reduces the overlap due to glare and distance variations. In general, the algorithms standard normal variate, min-max, area under the curve, and single wavelength normalization are the most suitable to preprocess data used to develop a classification algorithm for tissue classification. The type of contrast between tissue types determines which of these four algorithms is most suitable.

The detection of cervical neoplasia via optical ımaging: a pilot clinical study

PURPOSE

The present study aims to develop a new high-resolution imaging system for the early diagnosis of cervical neoplasia based on increased vessel density of the cervical tissue.

METHODS

An optical device was developed to obtain high contrast and resolution images of vascular structures of the cervix in the present study. The device utilizes a telecentric lens to capture cervix images under light illumination with a wavelength of 550 nm emitted from LEDs. Images were obtained using the telecentric lens with or without acetic acid application to the cervix. Image processing algorithms were used to contrast and extract the skeleton of the vascular structures on the cervix. In the evaluation of the vascular density, the cervical images were divided into 12 o'clock positions, and the fractal dimension of the vascularity was calculated for each dial area between the o'clock positions. The region with the largest fractal dimension was accepted as the region with the highest probability of lesion. The range of vessel sizes was split into small classes of "bins" for each dial area with the highest fractal dimension. To validate the system's success in differentiating between normal and HSIL lesions, forty five patients who underwent colposcopy and biopsy were included in a pilot study.

RESULTS

The system correctly classified four HSIL cases out of five and failed to detect one HSIL case, achieving an accuracy rate of 97.8% with an 80% sensitivity and 100% specificity.

CONCLUSION

The developed high-resolution optical imaging system may potentially be used in detecting cervical neoplasia just before the biopsy and reduce the number of false-positive cases.

Contrast-Enhancing Snapshot Narrow-Band Imaging Method for Real-Time Computer-Aided Cervical Cancer Screening

Composition of cervical precancerous lesions and carcinoma in situ is rich in hemoglobin, unlike healthy tissues. In this study, we aimed to utilize this difference to enhance the contrast between healthy and diseased tissues via snapshot narrow-band imaging (SNBI). Four narrow-band images centered at wavelengths of characteristic absorption/reflection peaks of hemoglobin were captured with zero-time delay in between by a custom-designed SNBI video camera. Then these spectral images were fused in real time into a single combined image to enhance the contrast between normal and abnormal tissues. Finally, a Euclidean distance algorithm was employed to classify the tissue into clinical meaningful tissue types. Two pre-clinical experiments were conducted to validate the proposed method. Experimental results indicate that contrast between different grades of diseased tissues in the SNBI generated image was indeed enhanced, as compared to conventional white light image (WLI). The computer-aided classification accuracy was 100% and 50% as compared to the gold standard histopathological diagnosis results with the SNBI and the conventional WLI methods, respectively. Further, the boundary contour between health tissue, cervical precancerous regions, and carcinoma in situ can be automatically delineated in SNBI. The proposed SNBI method was also fast, and it generated automatic diagnostic results with clear boundary contours at over 11 fps on a Pentium 1.6-GHz laptop. Hence, the proposed SNBI is of great significance to enlarge worldwide the coverage of regular cervical screening program, and to live guide surgeries such as biopsy sample collection and accurate cervical cancer treatment.