Shape analysis and classification in the identification of colon texture images

Khaled Marghani, Satnam Dlay, Andrew Sims, Bayan Sharif

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review


Automated medical image analysis using quantitative measurements is extremely helpful for cancer prognosis and diagnosis. It's because reliable decision-making in histological laboratories can help in the planning of appropriate surgery and therapy, which can make a difference between life and death. Interpreting natural image information to significant features for classification purpose is not a trivial task. In this paper, the strength of considering shape based geometry is investigated using two feature types known as, fractal geometric feature, fractal dimension (FD), and shape factor features as the form factor, nuclear contour index, elongation, and regularity index. The first was mainly based on traditional box counting method whilst the other are based on morphology. For better segmentation and consequently optimize feature extraction process, several morphological operations were used. In order to test our algorithm, 102 of colon texture images, consisting of 44 normal tissue images and 58 malignant tissue glands, were examined. The obtained results shows there is a strong significant of P=3.29×10 -12 in identifying abnormalities using nuclear contour index and P=5. 57013×10 -10 with fractal dimension feature. Discrimination analyses using all significant features estimated shows that a ratio of 91.4%, 84.1% in the sensitivity and specificity, respectively, with an overall accuracy of 88.2% is achieved. In brief, this study concludes that abnormalities in low-level power tissue morphology can be clearly identified using quantitative image analysis based on the shape. This investigation shows the high potential of usefulness of automated system in histopathology. Furthermore, it has the advantage of being objectively, unbiased sampling, and more important as valuable diagnostic decision support tool. Finally, combining FD with features of shape measures is expressing promising results in the classification of histological dataset.

Original languageBritish English
Title of host publicationRecent Advances in Intelligent Systems and Signal Processing
Number of pages6
StatePublished - 2003


  • Classification
  • Colon
  • Histological images
  • Identification
  • Machine vision
  • Morphology
  • Segmentation
  • Shape analysis


Dive into the research topics of 'Shape analysis and classification in the identification of colon texture images'. Together they form a unique fingerprint.

Cite this