Experimental Validation

A collection of simulations is processed to measure the efficiency of the projected method in an empirical fashion. The brief testing diagnosis is carried out in the upcoming sections.

Data Set Used

A group of around 3,500 WCE images was extracted from a set of 16 bleeding as well as 16 nonbleeding videos taken from 32 patients [17]. These details related to

Sample images with respective magnitude spectrums

FIGURE 5.6 Sample images with respective magnitude spectrums.

the images are explained here. A similar frame is removed for eliminating the recurrence of images. The noninformative frame is destroyed under the application of remaining food, turbid fluid, bubbles, specular reflection, or fecal content that cannot be removed. By practical applications, it is more significant for classifier bleeding as well as nonbleeding images to the easy access of physicians. Hence, an effective as well as real-time application is capable of resolving the problems involved in frames. All filtered frames from a collection of 32 videos is used to develop a data set. A training data set is comprised with a group of 600 bleeding as well as 600 nonbleeding frames obtained from 12 various people. It can be employed in training the SVM classifier. Simultaneously, the testing data set captured 860 bleeding whereas 860 nonbleeding frames were collected from a residual group of patients. Fig. 5.6 depicts the instance of bleeding and nonbleeding cases. Also, the corresponding magnitude spectrums are obviously depicted.

Results Analysis

Figs. 57-5.9 show the results offered by distinct models on the applied WCE images. Fig. 5.7 depicts the comparative study of the applied models in terms of accuracy. It is exhibited that the raw histogram and ratio (RHR) model has resulted in a worse classification of WCE images with a minimal accuracy of 73.72%. In line with, the PNN model has outperformed the earlier model and attained a somewhat maximum accuracy of 74.48%. Along with that, the color moment-local binary patterns (CM-LBP) method has tried to show a manageable WCE image classification outcome and ended up with an accuracy of 77.74%. Concurrently, a competitive classification accuracy of 91.51% has been attained by the super pixel based (SP) model. But, the proposed NGLCM-IGWO-SVM model has resulted to superior classification outcome w ith the highest accuracy of 92.76%.

Accuracy analysis of different approaches on WCE images

FIGURE 5.7 Accuracy analysis of different approaches on WCE images.

Fig. 5.8 demonstrates the sensitivity analysis of NGLCM-IGWO-SVM model with other models. It is exhibited that the RHR model has resulted to a worse classification of WCE images with a minimal sensitivity of 75.35%. In line with, the PNN

Sensitivity analysis of different approaches on WCE images

FIGURE 5.8 Sensitivity analysis of different approaches on WCE images.

Specificity analysis of different approaches on WCE images

FIGURE 5.9 Specificity analysis of different approaches on WCE images.

model has outperformed the earlier model and attained a somewhat maximum sensitivity of 75.93%. Along with that, the CM-LBP method has tried to show a manageable WCE image classification outcome and ended up with a sensitivity of 76.28%. Concurrently, a competitive classification sensitivity of 89.07% has been attained by the SP model. But, the proposed NGLCM-IGWO-SVM model has resulted in a superior classification outcome with the highest sensitivity of 90.35%.

Fig. 5.9 performs a comparison of specificity values of the NGLCM-IGWO-SVM model with other models. It is exhibited that the RHR model has resulted in a worse classification of WCE images with a minimal specificity of 72.09%. In line with, the PNN model has outperformed the earlier model and attained a somewhat high specificity of 73.02%. Along with that, the CM-LBP method has tried to show manageable WCE image classification outcome and ended up with a specificity of 79.19%. Concurrently, a competitive classification specificity of 93.95% has been attained by the SP model. But, the proposed NGLCM-IGWO-SVM model has resulted in a superior classification outcome with the highest specificity of 94.87%. By observing the mentioned tables and figures, it is confirmed that the NGLCM-IGWO-SVM model has outperformed all the compared methods in a considerable way.

Conclusion

This chapter has introduced a novel NGLCM-IGWO-SVM model for the detection of bleeding regions from WCE images. The presented NGLCM-IGWO-SVM model involves a set of different processes, namely data collection, preprocessing, feature extraction, and classification. Once the data is collected and preprocessed, a proficient NGLCM method is utilized to extract the features from the provided GI images. Then, the classification process is carried out by the use of IGWO-SVM, where the parameters of SVM have been tuned by the IGWO algorithm. The simulation of the NGLCM-IGWO-SVM model takes place using benchmark GI images. The experimental outcome pointed out that the NGLCM-IGWO-SVM model is superior to other models with a maximum accuracy of 92.76%, specificity of 94.87%, and sensitivity of 90.35%. In the future, the NGLCM-IGWO-SVM method could be utilized in real-time applications.

References

[1] Emmanouilidou, M., The Internet of Healthcare Things: A European perspective and a review of ethical concerns. International Journal of Business, Human and Social Sciences, 13 (5), 675-679, 2019.

[2] Deva, S. V. S. V. P„ Akashe. S., and Kim. H. J., Feasible challenges and applications of IoT in healthcare: Essential architecture and challenges in various fields of Internet of Healthcare Things. In Smart Medical Data Sensing and IoT Systems Design in Healthcare IGI Global, 2020. 178-200.

[3] Ianculescu. M., and Alexandru. A., Internet of Health Things as a win-win solution for mitigating the paradigm shift inside senior patient-physician shared health management. International Journal of Computer and Information Engineering, 13 (10). 573-577, 2019.

[4] Fu. Y„ Zhang, W„ Mandal, M„ and Meng. M.-H. Computer-aided bleeding detection in WCE video. IEEE Journal of Biomedical and Health Informatics, 18 (2), 636-642. 2014.

[5] Kumar, R.. Zhao, Q., Seshamani. S., Mullin, G„ Hager, G.. and Dassopoulos, T. Assessment of Crohn's disease lesions in wireless capsule endoscopy images, IEEE Transactions on Biomedical Engineering,59 (2) 355-362, 2012.

[6] Li. B., and Meng, M.-H. Computer-aided detection of bleeding regions for capsule endoscopy images, IEEE Transactions on Biomedical Engineering, 56 (4). 1032-1039. 2009.

[7] Hwang, S.. Oh, J.. Cox. J., Tang, S. J.. and Tibbals. H. F. Blood detection in wireless capsule endoscopy using expectation maximization clustering. Proceedings SPIE, 6144, 1-11,2006.

[8] Pan, G., Yan. G.. Qiu, X., and Cui. J. Bleeding detection in wireless capsule endoscopy based on probabilistic neural network. Journal of Medical Systems 35 (6), 1477-1484. 2011.

[9] Liu. J., and Yuan, X. Obscure bleeding detection in endoscopy images using support vector machines, Optimization and Engineering 10 (2), 289-299. 2009.

[10] Hegenbart, S„ Uhl, A., Vecsei, A. and Wimmer. G, Scale invariant texture descriptors for classifying celiac disease, Medical Image Analysis, 17 (4), 458-474, 2013.

[11] Cunha, J.. Coimbra, M„ Campos, P„ and Soares, J. M. Automated topographic segmentation and transit time estimation in endoscopic capsule exams, IEEE Transactions on Medical Imaging, 27 (1), 19-27, 2008.

[12] Lakshmanaprabu. S. K., Mohanty, S. N.. Shankar, K.. Arunkumar, N„ and Ramirez. G. Optimal deep learning model for classification of lung cancer on CT images. Future Generation Computer Systems, 92. 374-382, 2019.

[13] Elhoseny, M., and Shankar, K. Optimal bilateral filter and convolutional neural network based denoising method of medical image measurements. Measurement, 143. 125-135, 2019.

[14] Shankar, K., Lakshmanaprabu, S. K., Khanna. A.. Tanwar, S.. Rodrigues. J. J.. and Roy. N. R. Alzheimer detection using Group Grey Wolf Optimization based features with convolutional classifier. Computers & Electrical Engineering, 77, 230-243, 2019.

[15] Shankar. K., Perumal, E„ and Vidhyavathi. R. M. Deep neural network with moth search optimization algorithm based detection and classification of diabetic retinopathy images. SN Applied Sciences, 2 (4), 1-10, 2020.

[16] Elhoseny, M„ Shankar. K., and Uthayakumar, J. Intelligent diagnostic prediction and classification system for chronic kidney disease. Scientific Reports, 9 (1). 1-14. 2019.

[17] Hassan, A.R. and Haque, M.A. Computer-aided gastrointestinal hemorrhage detection in wireless capsule endoscopy videos. Computer methods and programs in biomedicine, 122 (3), 341-353, 2015.

 
Source
< Prev   CONTENTS   Source   Next >