New insights about automatic bladder tumors detection

envie a um amigo share this

New insights about automatic bladder tumors detection

Terça, 30.01.2018

Bladder tumor is one of the leading causes of death, especially among men and its diagnosis is highly dependent on the interpreter’s experience of different exams. There is no doubt that the development of automatic detection systems will improve the detection accuracies of this pathology. Therefore, Nuno Freitas a Master student of Biomedical Engineering in University of Minho, oriented by Professor Carlos S. Lima (School of Engineering) and Professor Estevão Lima, MD (School of Medicine and Head of the Urology Department in Hospital of Braga), was able create an initial module to be part of an automatic and real-time bladder tumors detector, using white light cystoscopy frames.

In this approach each frame was processed by a method divided into 3 but complementary blocks, namely the preprocessing step, the segmentation of each frame in two regions and the extraction of statistical measures. One of the observations drawn from our work is that the pre-processing followed by the segmentation increased the performances in every proposed approach. Although it is a preliminary study, the obtained results are a good start to produce complete automatic diagnosis, since a sensitivity of 91% was obtained when using a database with 353 images.

 

Authors and Affiliations:

Nuno R. Freitas1, Pedro M. Vieira1, Estevão Lima2,3, and Carlos S. Lima1

1 CMEMS-UMinho Research Unit, University of Minho, Guimarães 4800-058, Portugal

2 Life and Health Sciences Research Institute, University of Minho, Campus Gualtar, Braga 4710-057, Portugal

3 Department of Urology, Hospital of Braga, Braga 4710-243, Portugal

 

Abstract:

Correct cystoscopy images classification depends on interpreter's experience. Bladder cancer is a common lesion that can only be confirmed by biopsying the tissue, wherefore the automatic identification of tumors has a significant role in the early stage diagnosis and its accuracy. To our best knowledge the use of white light cystoscopy images to bladder tumor diagnosis wasn't reported so far. In this paper a texture analysis-based approach is proposed for bladder tumor diagnosis presuming that tumors change tissue texture. As it is well accepted by the current scientific community texture information is more present in the medium to high frequency range which can be selected by using the Discrete Wavelet Transform (DWT). Tumor enhancement can be improved by using automatic segmentation, since a mixing with normal tissue is avoided in ideal conditions. The segmentation module proposed in this paper takes advantage of the wavelet decomposition tree to discard poor texture information in such a way that both steps of the proposed algorithm segmentation and classification share the same focus on texture. A Multilayer Perceptron (MLP) and Support Vector Machine (SVM), with a stratified 10-fold cross-validation procedure was used for classification purposes by using the hue-saturation-value (HSV), red-green-blue (RGB), and CIELab color spaces. Performances of 91% in sensitivity and 92.9% in specificity were obtained regarding HSV color by using both pre-processing and classification steps based on DWT. The proposed method can achieve good performances on identifying bladder tumor frames. These promising results open the path towards a deeper study regarding the applicability of this algorithm in computer aided diagnosis.

 

Journal: Physics in Medicine & Biology

 

Link: http://iopscience.iop.org/article/10.1088/1361-6560/aaa3af