Detection of Lung and Colon Cancer Using XRx-Net With Histopathological Images.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Lung and Colon Cancer (LCC) diagnosis traditionally relies on histopathology, a complex and time-consuming microscopic examination.
• Factors like genetics, smoking, and alcohol consumption influence LCC prevalence.
• Current diagnostic methods present challenges in efficiency and complexity.

Purpose of the Study:

• To introduce the Xception ResNeXT Network (XRx-Net) framework for efficient Lung and Colon Cancer (LCC) detection.
• To improve the accuracy and speed of LCC diagnosis using deep learning techniques.
• To develop an automated system for classifying LCC subtypes and benign tissues.

Main Methods:

• Histopathological images of LCC were preprocessed using an Adaptive Bilateral Filter (ABF).
• Cancerous cells were segmented using hookNet, followed by feature extraction via Wavelet texture analysis and Convolutional Neural Networks (CNNs).
• The XRx-Net framework, a hybrid of Xception and ResNeXt models, was employed for LCC classification.

Main Results:

• The XRx-Net framework successfully identified lung benign tissue, lung adenocarcinoma, lung squamous cell carcinoma, colon benign tissue, and colon adenocarcinoma.
• Achieved a testing accuracy of 91.34%.
• Demonstrated a True Positive Rate (TPR) of 92.36% and a True Negative Rate (TNR) of 90.14%.

Conclusions:

• The XRx-Net framework offers a highly accurate and efficient AI-driven solution for LCC detection.
• This deep learning approach can significantly aid pathologists in diagnosing LCC subtypes.
• The study highlights the potential of hybrid CNN architectures in improving cancer diagnostic workflows.