Accurate detection and segmentation of brain cancer in MRI scans are
critical for effective diagnosis and treatment planning. Traditional
methods often struggle with the complexities of tumor morphology and
variations in scan quality. Existing detection systems can be slow and
may not effectively handle the variability in tumor appearances,
leading to potential delays in diagnosis and treatment. To address these
challenges, we propose an enhanced detection framework using a
Siamese Regional Proposed Network (SRPN). The SRPN integrates
template branch and bounding box regression to expedite detection
processes. The system utilizes an extended Siamese network to learn
the distance between tracklet pairs, capturing the local and global
features of tumors. These features are transferred to bidirectional gated
recurrent units (GRUs), which generate tracklets and segment them
into shorter sub-tracklets based on local distances. The segmented sub-
tracklets are then reconnected into longer trajectories using similarities
derived from temporal pooling global features. Additionally, fuzzy logic
fusion is employed to combine segmented regions for improved
accuracy. The SRPN-based framework demonstrated a significant
improvement in detection speed and accuracy. Experimental results
show an accuracy increase of 12% over traditional methods, achieving
94% accuracy with a detection time reduction of 30%. The system also
improved segmentation precision, with a mean Intersection over Union
(IoU) score of 85%, compared to 75% in conventional approaches.
S.A. Anlet Sharmili1, M. Sankari2
Manonmaniam Sundaranar University, India1, Lekshmipuram College of Arts and Science, India2
Brain Cancer Detection, MRI Scans, Siamese Regional Proposed Network, Bounding Box Regression, Fuzzy Logic Fusion
| January | February | March | April | May | June | July | August | September | October | November | December |
| 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 3 | 0 | 0 | 0 |
| Published By :
ICTACT
Published In :
ICTACT Journal on Image and Video Processing
( Volume: 15 , Issue: 1 , Pages: 3347 - 3356 )
Date of Publication :
August 2024
Page Views :
34
Full Text Views :
3
|