•  
  •  
 

Abstract

Image processing plays vital role in medical sciences. Medical image processing reduces diagnosis time and cost. Early detection of cancer may reduce life risk. Earlier works on cancer detection mostly focused on specific organs and imaging modalities. This work aims a generalized framework for detecting tumors in various organs by analyzing different imaging modalities. Here, Cellular Automata helps in image segmentation through region growing, and deep learning algorithms made performance analysis in terms of accuracy and loss. Analyzing CT-Scan, MRI images of brain, breast, lungs, the proposed framework may assist to detect the region of mass being developed. Proposed CA-based segmentation technique segregates the region of interest or tumor area from its background. Segmentation through region growing is performed using Moore neighborhood concept. First, noise has been reduced using filters. Then enhanced image is converted into image matrix, and CA rule is applied to it for segmentation. Generally, area of the tumor appears with high-intensity values. Here, segmentation is done by identifying the high-intensity pixel values of the image and then gradually performing region growth to include entire tumor area. Deep learning algorithms are applied to transformed image set of cancer. Finally, performance analysis is made. The parameters of performance analysis are compared between transformed and original image sets, and the results obtained with the transformed image set produced higher accuracy than the results produced from original image set of tumors. The proposed framework may help medical practitioners in detecting tumors in different organs efficiently.

Keywords

Cellular Automata, CNN, Imaging Modality, Image Segmentation, Region Growing, ResNet, VGG

Subject Area

Computer Science

Article Type

Article

First Page

2119

Last Page

2132

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Share

COinS