Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Lung Cancer Classification from CT Images Using ResNet

Olajumoke O. Adekunle, Joseph D. Akinyemi, Khadijat T. Ladoja, Olufade F. W. Onifade

TL;DR

The paper tackles multiclass lung cancer classification by adapting a pre-trained ResNet-50 for lung tissue images from the LC25000 dataset. By replacing the final layer and adding two dense layers, the model learns a 3-class distinction among benign, adenocarcinoma, and squamous cell carcinoma using 15,000 lung-tissue images (10,200 train, 2,550 val, 2,250 test) resized to 224×224 and trained for 18 epochs with a 25-sample batch and 0.001 learning rate. The approach achieves a test accuracy of 98.8%, with benign samples classified perfectly and high F1-scores for malignant classes, while misclassifications occur primarily between the two malignant categories. Saliency maps are used to interpret learned features, and the results outperform several prior LC25000 studies, indicating strong generalization and potential clinical utility. The work highlights the effectiveness of transfer learning with customized head layers for histopathological lung tissue classification and suggests avenues for scaling to larger datasets and additional cancer subtypes.

Abstract

Lung cancer, a malignancy originating in lung tissues, is commonly diagnosed and classified using medical imaging techniques, particularly computed tomography (CT). Despite the integration of machine learning and deep learning methods, the predictive efficacy of automated systems for lung cancer classification from CT images remains below the desired threshold for clinical adoption. Existing research predominantly focuses on binary classification, distinguishing between malignant and benign lung nodules. In this study, a novel deep learning-based approach is introduced, aimed at an improved multi-class classification, discerning various subtypes of lung cancer from CT images. Leveraging a pre-trained ResNet model, lung tissue images were classified into three distinct classes, two of which denote malignancy and one benign. Employing a dataset comprising 15,000 lung CT images sourced from the LC25000 histopathological images, the ResNet50 model was trained on 10,200 images, validated on 2,550 images, and tested on the remaining 2,250 images. Through the incorporation of custom layers atop the ResNet architecture and meticulous hyperparameter fine-tuning, a remarkable test accuracy of 98.8% was recorded. This represents a notable enhancement over the performance of prior models on the same dataset.

Lung Cancer Classification from CT Images Using ResNet

TL;DR

The paper tackles multiclass lung cancer classification by adapting a pre-trained ResNet-50 for lung tissue images from the LC25000 dataset. By replacing the final layer and adding two dense layers, the model learns a 3-class distinction among benign, adenocarcinoma, and squamous cell carcinoma using 15,000 lung-tissue images (10,200 train, 2,550 val, 2,250 test) resized to 224×224 and trained for 18 epochs with a 25-sample batch and 0.001 learning rate. The approach achieves a test accuracy of 98.8%, with benign samples classified perfectly and high F1-scores for malignant classes, while misclassifications occur primarily between the two malignant categories. Saliency maps are used to interpret learned features, and the results outperform several prior LC25000 studies, indicating strong generalization and potential clinical utility. The work highlights the effectiveness of transfer learning with customized head layers for histopathological lung tissue classification and suggests avenues for scaling to larger datasets and additional cancer subtypes.

Abstract

Lung cancer, a malignancy originating in lung tissues, is commonly diagnosed and classified using medical imaging techniques, particularly computed tomography (CT). Despite the integration of machine learning and deep learning methods, the predictive efficacy of automated systems for lung cancer classification from CT images remains below the desired threshold for clinical adoption. Existing research predominantly focuses on binary classification, distinguishing between malignant and benign lung nodules. In this study, a novel deep learning-based approach is introduced, aimed at an improved multi-class classification, discerning various subtypes of lung cancer from CT images. Leveraging a pre-trained ResNet model, lung tissue images were classified into three distinct classes, two of which denote malignancy and one benign. Employing a dataset comprising 15,000 lung CT images sourced from the LC25000 histopathological images, the ResNet50 model was trained on 10,200 images, validated on 2,550 images, and tested on the remaining 2,250 images. Through the incorporation of custom layers atop the ResNet architecture and meticulous hyperparameter fine-tuning, a remarkable test accuracy of 98.8% was recorded. This represents a notable enhancement over the performance of prior models on the same dataset.
Paper Structure (11 sections, 4 figures, 3 tables)

This paper contains 11 sections, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Cancer Classification and Detection
  4. Lung Cancer Classification
  5. Methodology
  6. The Dataset
  7. The Deep Learning Model
  8. Results and Discussion
  9. Model Evaluation
  10. Comparative Analysis
  11. Conclusion

Figures (4)

  • Figure 1: Image samples in the dataset.
  • Figure 2: Saliency maps showing some of the features extracted by the deep learning model from CT images of Top-left (Lung adenocarcinoma); Top-right (Lung squamous cell carcinoma); bottom (lung benign).
  • Figure 3: Training and validation curves of the lung cancer prediction model. Top: accuracy curves. Bottom: Loss curves.
  • Figure 4: Confusion matrix of the lung cancer prediction model.