An improved AlexNet deep learning method for limb tumor cancer prediction and detection
Arunachalam Perumal1, Janakiraman Nithiyanantham2, Jamuna Nagaraj3
1Department of Biomedical Engineering, Vel Tech Rangarajan Dr Sagunthala R&D Institute of Science and Technology, Chennai, 600062, India.
Biomedical Physics & Engineering Express
|October 22, 2024
Summary
This study introduces an improved deep learning model for diagnosing synovial sarcoma (SS) from pathological images. The novel approach enhances diagnostic accuracy, aiding in earlier detection of this rare soft tissue cancer.
Area of Science:
- Oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Synovial sarcoma (SS) is a rare soft tissue cancer requiring early detection for improved patient outcomes.
- Digital pathology offers potential for enhanced diagnostic accuracy but faces challenges with image quality and complex feature identification.
- Current diagnostic methods for SS can be limited, necessitating advanced computational approaches.
Purpose of the Study:
- To develop and evaluate an improved deep learning model for the accurate diagnosis of synovial sarcoma (SS) using digital pathological images.
- To enhance the performance of convolutional neural networks (CNNs) for SS detection through architectural modifications and advanced preprocessing techniques.
- To establish a more precise and efficient method for identifying SS in pathological samples.
Main Methods:
- A convolutional neural network (CNN) based on an improved AlexNet architecture was developed for SS classification.
- Image preprocessing included dataset augmentation, adaptive median filtering (AMF), and histogram equalization for noise reduction and quality enhancement.
- Feature extraction utilized the Gray-Level Co-occurrence Matrix (GLCM) and Improved Linear Discriminant Analysis (ILDA), with segmentation performed using repetitive phase-level set segmentation (RPLSS).
Main Results:
- The improved AlexNet model demonstrated superior performance in diagnosing SS compared to existing methods.
- Key performance metrics showed significant improvements: accuracy (3%), sensitivity (1.70%), specificity (6.08%), and Area Under the Curve (AUC) (8.86%).
- The enhanced architecture with additional convolutional layers and resized inputs contributed to the model's superior diagnostic capabilities.
Conclusions:
- The proposed improved AlexNet CNN model offers a highly accurate and effective tool for the automated diagnosis of synovial sarcoma from digital pathological images.
- This deep learning approach has the potential to significantly improve early detection rates and subsequent patient survival for synovial sarcoma.
- The study highlights the efficacy of combining advanced image processing techniques with deep learning for challenging cancer diagnoses.
