Conference Proceedings
The International Conference on Emerging Technologies in Electronics, Computing and Communication 2022
(ICETECC`22)
Hybrid Deep Learning for Blood Cancer Detection: A CNN-GCN Approach for Enhanced Diagnostic Accuracy
Abdul Rehan1; Adeel Mukhtar1; Hunain Zaidi1; Zain ul Abidin1; Waqas Tariq Toor1*;| 1University of Engineering and Technology Lahore, Narowal Campus |
ABSTRACT
In medicine, a fast and accurate diagnosis of blood cancer is very important. To achieve that goal, in this paper, a novel hybrid deep learning architecture based on the combination of Convolutional Neural Networks and Graph Convolutional Networks for microscopic blood images classification into categories containing normal vs. cancerous is proposed. Using the feature extraction ability of CNNs and GCNs relational learning strengths, the model can extract both local features as well as more complex relationships within the image data. Performance tests: The designed hybrid model was tested on 6,220 microscopic blood images and outperformed state-of-the-art models like CNNs VGG16, DenseNet, and EfficientNet-B0, as it was almost 89\% accurate. Of course, the introduction of GCN layers dramatically improves the performance of the model, equipping it with the ability to identify those complex patterns characteristic for cancerous cells. Indeed, the article quite honestly demonstrates how hybrid models really appear to be helpful for medical image analysis: an opportunity has finally been created for early, accurate diagnostics of blood cancer.