Pancreatic Cancer Detection from CT Images using Deep Learning

IEEE BASE PAPER TITLE:

Pancreatic Tumor Recognition from CT Images through Advanced Deep Learning Techniques

IEEE BASE PAPER ABSTRACT:

Cancer constitutes a major affection nowadays, leading to death in many situations. Pancreatic malignant tumors represent the fourth most frequent cause of cancer related death in United States and Europe. The most trustworthy cancer diagnosis method is the biopsy, but this technique raises several risks, the most relevant one being the danger of tumoral spread through the human body. Thus, in the context of our research, we employ computerized methods for achieving highly accurate abdominal tumor recognition and segmentation within medical images. In the current approach, we performed the automatic recognition of pancreatic malignant tumors within Computed Tomography (CT) images, with the aid of Convolutional Neural Networks (CNN), by employing last generation architectures, as well as their improved versions, respectively combinations of these structures at classifier and decision level. The classification performance was assessed through specific metrics, an accuracy above 98% being achieved.

PROJECT OUTPUT VIDEO:

ALGORITHM / MODEL USED:

AlexNet CNN Model.

OUR PROPOSED PROJECT ABSTRACT:

Pancreatic cancer remains one of the most challenging malignancies to diagnose and treat due to its asymptomatic nature in early stages. The pancreas, located deep within the abdomen, is difficult to visualize clearly with traditional imaging methods, which often leads to late-stage diagnoses with limited treatment options. Early and accurate differentiation between normal pancreatic tissue and malignant tumors is crucial for improving patient outcomes.

This work proposes a deep learning-based classification approach using the AlexNet Convolutional Neural Network (CNN) to distinguish between normal pancreatic tissue and pancreatic tumors in CT images using MATLAB. The dataset consists of 998 CT images, categorized into two classes: 421 normal images and 577 images exhibiting pancreatic tumors.

The AlexNet model was trained and tested on this dataset, achieving exceptional performance with a training accuracy of 100% and a testing accuracy of 100%. The model is designed to automatically extract and classify features from CT scans, offering a more efficient and accurate method for identifying pancreatic tumors compared to manual interpretation. By training AlexNet on a labelled dataset of normal and tumor images, the proposed model demonstrates promising accuracy and reliability in distinguishing between the two classes.

The class-specific evaluation metrics further validate the model’s reliability and robustness. The model achieved a precision of 100%, eliminating false positives in both normal and tumor classifications. Sensitivity (recall) also reached 100%, ensuring all tumor cases were correctly identified. Specificity was equally perfect at 100%, confirming accurate identification of all normal cases with no false negatives. Furthermore, the F1-score, balancing precision and recall, demonstrated an outstanding value of 100%, reflecting the model’s comprehensive detection capabilities.

This study underscores the potential of deep learning in advancing medical diagnostics, particularly for pancreatic cancer detection, by providing a reliable and efficient tool for analyzing CT images. The results highlight the promise of employing AI-driven solutions in enhancing early detection rates and improving clinical outcomes for pancreatic cancer patients.

OBJECTIVES

Develop a deep learning-based classification model using AlexNet to accurately differentiate between normal pancreatic tissue and pancreatic tumors in CT images.
Enhance diagnostic accuracy and efficiency by automating the feature extraction and classification process, reducing reliance on manual interpretation and minimizing misclassification risks.
Support early detection of pancreatic tumors to enable timely intervention, ultimately improving patient outcomes and survival rates in pancreatic cancer diagnosis.

SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS:

System : Pentium i3 Processor.
Hard Disk : 500 GB.
Monitor : 15’’ LED.
Input Devices : Keyboard, Mouse.
Ram : 8 GB.

SOFTWARE REQUIREMENTS:

Operating system : Windows 10 / 11.
Coding Language : MATLAB.
Tool : MATLABR2023B.

REFERENCE:

Delia Mitrea, Raluca Brehar, Razvan Itu, Sergiu Nedevschi, Mihai Socaciu, Radu Badea, “Pancreatic Tumor Recognition from CT Images through Advanced Deep Learning Techniques”, IEEE Conference, 2024.

Matlab IEEE Projects

Pancreatic Cancer Detection from CT Images using Deep Learning