Computer-Assisted Tissue Image Analysis in Minimally Invasive Surgery: History
Please note this is an old version of this entry, which may differ significantly from the current revision.
Contributor:

Computer-assisted tissue image analysis (CATIA) enables an optical biopsy of human tissue during minimally invasive surgery and endoscopy. Thus far, it has been implemented in gastrointestinal, endometrial, and dermatologic examinations that use computational analysis and image texture feature systems.

  • tissue image analysis
  • tissue texture image analysis
  • optical biopsies
  • computer-aided diagnosis

1. Introduction

Clinicians and computer scientists have been exploring the possibilities of using computer-assisted tissue image analysis (CATIA) in distinguishing normal from abnormal tissues. The fast development and application of endoscopy for diagnosis and treatment in daily medical practice, and the use of high-definition video recording facilitate research on CATIA [1]. Processing evaluation is based on a manual or automated image interpretation that filters artefacts from a database of images. We reviewed existing studies to evaluate the potential of performing optical biopsies of human tissues and the reliability of the results, evaluated the gained experience from CATIA on various tissues, and explored the possibilities of implementing tissue image texture analysis in daily medical practice within the context of computer-assisted diagnostics (CAD) during minimally invasive surgery. In the studies reviewed, tissue image processing techniques focus on three aspects. Tissue image features, colour-spectrum characterization and filtering as well as algorithm and statistical evaluation during or after a patient’s endoscopic procedure (Table 1).
Table 1. Summary and Indexing of tissue image processing terminology and techniques.
Image Processing Evaluation:
Manual or automated image interpretation that filters artefacts from a database of images, e.g., endoscopic surgery video summary.
Colour-spectrum, Characterization and Filtering:
Image colour texture content of the region of interest (ROI).
Colour texture features are extracted over different colour spaces or hue saturation values (HSV).
(1) Red green blue (RGB), (2) Luminance (Y), (3) Chrominance (red-yellow)/chrominance (blue-yellow) (YCrCb).
For each colour space component, a standard grayscale feature is used and can be widely applied for texture characterization according to different texture features.
Tissue image features
There are 26 texture features from each colour component.
(i) Statistical Features (SF): (1) Mean, (2) Variance, (3) Median, (4) Energy, (5) Skewness, (6) Kurtosis, (7) Mode, (8) Entropy.
(ii) Spatial Gray Level Dependence Matrices (SGLDM) (1) Angular second moment, (2) Contrast, (3) Correlation, (4) Homogeneity, (5) Variance, (6) Entropy, (7) Sum Entropy, (8) Sum Average, (9) Sum Variance, (10) Difference Entropy, (11) Difference Variance, (12) Information Measurement of Correlation 1, (13) Information Measurement of Correlation 2.
(iii) Gray Level Difference Statistics (GLDS): (1) Mean, (2) Entropy, (3) Contrast, (4) Homogeneity, (5) Energy.
Algorithm and Statistical evaluation
Training and testing to distinguish normal from abnormal Regions of Interest (ROI).
CATIA system performance was evaluated using SVM algorithm and probabilistic neural networks (PNN).
C-SVM network was used to investigate the Gaussian radial basis function (RBF) kernel and the linear kernel.
Principal component analysis (PCA) reforms a dataset into a bilinear model of linear independent variables and uses a mathematical equation to explain the variation within the dataset.
Vectors within the matrix are reshaped into images that show the spatial distribution forming abundance images, which represent the abundance of each vector for each pixel.
Abundance images are then plotted in a colour-scaled image and can be combined with prominent differences between the samples highlighted.
Partial least squares discriminant analysis (PLS-DA) is a supervised data reduction technique.
It uses a versatile algorithm that can predict and describe modelling as well as select discriminative variables.
Examining suspicious tissue without an invasive procedure by in vivo optical biopsy with tissue image analysis has several advantages: preventing vascular and tissue injury, haemorrhage, haematoma, spread of malignant cells, and risk for infection and scarring. In addition, CATIA allows for comparing suspicious tissue to its neighbouring healthy regions. Thus far, final diagnosis and treatment follows histopathologic examination. Nevertheless, tissue image analysis may guide the physician during biopsy sampling by providing a high-risk or low-risk tissue malignancy score. CATIA could serve as a ‘second opinion’, augmenting the physician’s decision on the biopsy sampling location and allowing a preliminary tissue image result. In the suspicion of malignancy, the histopathological examination can be prioritised. Moreover, tissue image analysis could decrease the risk for error, especially in difficult and suspicious cases with extensive visual tissue variability [2].
Tissue images during minimally invasive surgery and endoscopic images during colonoscopy and gastroscopy for CATIA could be manually isolated and evaluated when an abnormality or lesion is suspected. An automated system could also be used to define frames of normal and abnormal characteristics in endoscopic segments with different visual appearance [3]. An expert might easily choose the frames that need further processing, though the inexperienced practitioner might find this more difficult. Specific tissue segments could be isolated from a video, and groups of frames with suspicious pathologic features could be visualized and selected for CATIA. Clustering and classification techniques facilitate the selection of automated images and allows surveillance of defined targets [1]. Skin, gastrointestinal tract [GIT], larynx, and endometrial [4][5] tissues were analysed for malignancy by optical biopsies.

2.  Computer-Assisted Tissue Image Analysis in Minimally Invasive Surgery

The camera systems, monitors, operative techniques, and skills developed with minimally invasive surgery provide tissue images and magnification with exceptional clarity. The abdomen and individual organs can be examined in situ with ease, without disturbing the anatomic features or the pathologic condition before treatment. In addition, video images can be used intra- and postoperatively to re-evaluate the pathologic condition and operative technique and for teaching purposes. They provide the surgeon with excellent quality real-time video, assessing cavities and areas of the human body impossible to observe with the naked eye. The easy access to tissue images facilitates, encourage, and accelerate the application of bioinformatics using different algorithms, which are correlated with the histopathological findings [1][6].
Dual-working channel endoscopes can enable an image-guided punch biopsy by using OCT. Matched OCT images obtained in vivo corresponding to histological biopsies can improve the accuracy and reliability of the technique [7]. An improvement in image resolution and the development of more specific imaging technology, such as polarization sensitive OCT, may also improve the accuracy of detecting buried pathologic features [7]. However, dual-working channel endoscopes increase the tip diameter of the scope, which is a big disadvantage when small cavities are observed, as in hysteroscopy. OCT is frequently used in ophthalmology and can provide information about cell architecture and morphology up to 15 nano microns below epithelial cells [8].
Tissue visual signs, image texture analysis, and selected features by electronic neural network systems can serve as biomarkers distinguishing abnormal from normal tissue. Precancerous as well as cancerous conditions are characterized as images with a complex set of attributes. Colour, texture, and relative geometry are predominately useful, while region shape is significantly less so. Regions are frequently amorphous, or, for a few region classes, exhibit a shape which may be only approximately modelled, and even in these cases, the model may be image dependent. The overall region of interest in the images may in general correlate with the histopathologic cancerous characteristics, such as abnormal tissue architecture, neo-angiogenesis, oedema, and cellular dysfunction. Images from a histopathologic section produced by microscopy may be interpreted by visual signs and tissue image features by computer-assisted diagnosis [9]. Such translation from microscopy tissue section characteristics to tissue image textures demand an allocation of data and computer system training [10]. CAD may have the potential to diagnose early disease, including cancer [1]. The loading of data with digital features of normal and abnormal tissue, with both visual and histopathologic characteristics, is essential in building the primary level of bioinformatics. The functionality and efficiency of CAD depends on network capacity, speed of data processing, and technological support [1].
The texture discrimination of capsule endoscopy (CE) video frames can be improved by modelling classical texture descriptors in the colour scale plane instead of the colour plane, as usually assumed by classical approaches [4]. Higher order statistics applied to the joint distribution of classical texture descriptors appear effective for texture characterization. Future work will include introducing different classification schemes [4]; augmenting the database, which is important in generalizing the results, especially when higher order statistics’ modelling is involved; exploring the temporal dynamics of texture information, since taking information from neighbour frames may improve classification performance [4].
Optical coherence tomography (OCT) is widely considered a real-time intraoperative tumour margin assessment because of its high-resolution (HR) images, rapid scanning, and optical properties [11]. However, although OCT provides HR images, the combination of OCTSS and spectral domain (SD) is still insufficient to effectively classify different types of internal organs [11]. The main reason is that OCT images are simply composed of the reflectivity of light (elastic scattering property), which can only reflect the texture information instead of molecular information [12]. OCT is a minimally invasive method to evaluate buried glands or other subsurface features and may be used to evaluate the efficacy of other endoscopic therapies, such as cryoablation and photodynamic therapies, not only in the GIT but also in the skin and abdominal cavity [13].
Raman spectrum is aimed at improving the accuracy of tissue margins’ delineation by detecting the margin of tumour surrounded by normal tissues, e.g., muscle. Based on the integrated system, OCT and RS can acquire the measurement with similar experimental conditions [14]. This allows for real-time review and assessment of the margins. Tumour margin detection can be evaluated with different algorithms and tissue types. 3D optical coaxial tomography and Raman spectroscopy were the two additional modalities used in combination with the tissue texture analysis to augment CATIA diagnostic ability. Coaxial tomography seems to provide extra information regarding the tissue cell layers below the superficial layer and can be used as an added tool to the optic system. Raman spectroscopy provides highly specific 3D spectra with intensity and time axes mainly used during microscopy for histopathologic sections.
Although many ENT articles have been published on CAD, research on tissue texture analysis was missing. No studies using CAD for endometriosis were found. The intensity, density, and variety of tissue hue found in cases of pelvic and abdominal endometriosis would facilitate CATIA research in clinical practice. CATIA could probably contribute to the identification and quantification of endometriosis, especially the depth and extent of the disease on one tissue, the epithelium, and could probably assist in surgical treatment and the depth of destruction by laser and other modalities. Prospective and randomized studies are needed before CATIA is implemented in clinical practice.

This entry is adapted from the peer-reviewed paper 10.3390/jcm10245770

References

  1. Neofytou, M.S.; Tanos, V.; Constantinou, I.; Pattichis, M.S.; Kyriacou, E.C.; Pattichis, C.S. Computer–Aided Diagnosis in Hysteroscopic Imaging. IEEE J. Biomed. Health Inform. 2015, 19, 1129–1136.
  2. Efthymiou, M.; Allen, P.B.; Taylor, A.C.F.; Desmond, P.V.; Jayasakera, C.; de Cruz, P.; Kamm, M.A. Chromoendoscopy versus Narrow Band Imaging for Colonic Surveillance in Inflammatory Bowel Disease. Inflamm. Bowel Dis. 2013, 19, 2132–2138.
  3. Atasoy, S.; Mateus, D.; Meining, A.; Yang, G.; Navab, N. Endoscopic Video Manifolds for Targeted Optical Biopsy. IEEE Trans. Med. Imaging 2012, 31, 637–653.
  4. Chao, Z.; Tsung-Han, T.; Hsiang-Chieh, L.; Tejas, K.; Figueiredo, M.; Yuankai, T.K.; Osman, A.O.; Desmond, A.C.; Schmitt, J.M.; Qin, H.; et al. Characterization of Buried Glands before and after Radiofrequency Ablation by Using 3-Dimensional Optical Coherence Tomography (with Videos). Gastrointest. Endosc. 2012, 76, 32–40.
  5. Abdul-Wadood, D.N.; Loay, E.G.; Nabeel, A.R. Diagnosis of skin cancer using image texture analysis. Int. J. Sci. Eng. Res. 2014, 5, 155.
  6. Neofytou, M.S.; Pattichis, C.S.; Pattichis, M.S.; Tanos, V.; Kyriacou, E.C.; Koutsouris, D. A Standardised Protocol for Texture Feature Analysis of Endoscopic Images in Gynaecological Cancer. BioMed. Eng. OnLine 2007, 6, 44. Available online: http://www.biomedical-engineering-online.com/content/6/1/44 (accessed on 30 November 2021).
  7. Fujimoto, J.G.; Pitris, C.; Boppart, S.A.; Brezinski, M.E. Optical coherence tomography: An emerging technology for biomedical imaging and optical biopsy. Neoplasia 2000, 2, 9–25.
  8. Munzenmayer, C.; Winter, C.; Rupp, S.; Kage, A.; Wittenberg, T. Texture-based computer-assisted diagnosis for fiberscopic images. In Proceedings of the 2009 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Minneapolis, MN, USA, 3–6 September 2009.
  9. Metin, G.N.; Laura, B.E.; Ali, C.; Madabhushi, R.M.N.; Yener, B. Histopathological image analysis: A review. IEEE Rev. Biomed. Eng. 2009, 2, 147–171.
  10. Zhang, Y.; Wirkert, S.J.; Iszatt, J.; Hannes Kenngott, M.W.; Mayer, B.; Stock, C.; Clancy NTElson, D.S.; Maier-Hein, L. Tissue classification for laparoscopic image understanding based on multispectral texture analysis. J. Med. Imaging 2017, 4, 015001.
  11. Boustany, N.; Boppart, S.; Backman, V. Microscopic imaging and spectroscopy with scattered light. Annu. Rev. Biomed. Eng. 2010, 12, 285–314.
  12. Szczypinski, P.; Klepaczko, A.; Pazurek, M.; Daniel, P. Texture and colour based image segmentation and pathology detection in capsule endoscopy videos. Comput. Methods Programs Biomed. 2014, 113, 396–411.
  13. Chenzhou, W.; Gleysteen, J.; Nutte, T.T.; Yi, L.; Eben, R. In-vivo optical imaging in head and neck oncology: Basic principles, clinical applications and future directions. Int. J. Oral Sci. 2018, 10, 10.
  14. Liu, C.-H.; Qi, J.; Lu, J.; Wang, S.; Wu, C.; Shih, W.-C.; Larin, K.V. Improvement of tissue analysis and classification using optical coherence tomography combined with Raman spectroscopy. J. Innov. Opt. Health Sci. 2015, 8, 2015.
More
This entry is offline, you can click here to edit this entry!
Video Production Service