Evaluation and Exploration of Machine Learning and Convolutional Neural Network Classifiers in Detection of Lung Cancer from Microarray Gene-A Paradigm Shift

Our plain-English summary of Evaluation and Exploration of Machine Learning and Convolutional Neural Network Classifiers in Detection of Lung Cancer from Microarray Gene-A Paradigm Shift, by K. M S, H. Rajaguru, A.R. Nair (Bioengineering, 2023).

The study

• Authors: K. M S, H. Rajaguru, A.R. Nair
• Journal: Bioengineering
• Published: 2023;10(8)
• Cited by: 6 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Microarray gene expression-based detection and classification of medical conditions have been prominent in research studies over the past few decades. However, extracting relevant data from the high-volume microarray gene expression with inherent nonlinearity and inseparable noise components raises significant challenges during data classification and disease detection. The dataset used for the research is the Lung Harvard 2 Dataset (LH2) which consists of 150 Adenocarcinoma subjects and 31 Mesothelioma subjects. The paper proposes a two-level strategy involving feature extraction and selection methods before the classification step. The feature extraction step utilizes Short Term Fourier Transform (STFT), and the feature selection step employs Particle Swarm Optimization (PSO) and Harmonic Search (HS) metaheuristic methods. The classifiers employed are Nonlinear Regression, Gaussian Mixture Model, Softmax Discriminant, Naive Bayes, SVM (Linear), SVM (Polynomial), and SVM (RBF). The two-level extracted relevant features are compared with raw data classification results, including Convolutional Neural Network (CNN) methodology. Among the methods, STFT with PSO feature selection and SVM (RBF) classifier produced the highest accuracy of 94.47%.

In plain English

This was a computer-analysis study of existing data. It did not study asbestos exposure. The authors used a published gene-activity (microarray) dataset, the Lung Harvard 2 dataset, containing 150 adenocarcinoma and 31 mesothelioma tissue samples, each described by 12,533 genes. They tested whether computer programs could sort the samples into these two cancer types. The data were first condensed with a mathematical method (Short-Time Fourier Transform), then narrowed further with one of two feature-selection methods (Particle Swarm Optimization or Harmony Search), and then passed to seven classifier programs. A convolutional neural network was also tried. The best accuracy, meaning the share of samples sorted correctly, was 94.47%. It came from combining the Fourier step, Particle Swarm Optimization and a support vector machine (RBF) classifier. With the Fourier step alone and no feature selection, the best of the seven classifiers reached 80.66%; the neural network approaches reached up to 91.66%. The authors noted that the dataset was imbalanced, that the neural networks were prone to overfitting, and that microarrays may not capture all gene activity. The article does not say where or when the samples were collected.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Computational (machine learning) classification study using an existing microarray gene expression dataset; no new patients were recruited
• Who was studied: Lung Harvard 2 (LH2) dataset, cited to Gordon et al. 2002: 150 adenocarcinoma and 31 mesothelioma tissue samples, each characterized by 12,533 genes. Location and dates of sample collection are not given.
• Exposure: Not studied. No asbestos or other exposure data were analyzed; asbestos is mentioned only in the introduction as a general lung cancer risk factor.
• Main results:
  • Fourier-transform features with Particle Swarm Optimization feature selection and the SVM (RBF) classifier: accuracy 94.47%, precision 97.94%, recall 95.33%, F1 score 96.62%, error rate 5.52%
  • Fourier-transform features with Harmony Search feature selection and the SVM (RBF) classifier: accuracy 90.05%, error rate 9.94%
  • Fourier-transform features without feature selection, among the seven conventional classifiers: the best (Gaussian Mixture Model) reached 80.66% accuracy
  • Convolutional neural network on raw data: the best classifier (SVM RBF) reached 90.607% accuracy
  • Convolutional neural network on Fourier-transform features: the best classifier (Softmax Discriminant) reached 91.66% accuracy
• The authors' conclusion: The authors concluded that the Fourier-transform step combined with Particle Swarm Optimization feature selection and the SVM (RBF) classifier produced the highest accuracy, 94.47%.
• Limitations the authors state:
  • The dataset was imbalanced (many more adenocarcinoma than mesothelioma samples), and the neural networks were prone to overfitting and underfitting
  • The training data were limited
  • Neural network results are hard to interpret (black-box nature)
  • Microarray data may not capture all gene activity
  • The method trades some accuracy for lower computational complexity

Topics

Medical terms in this study, explained in our Lexicon: Lung Cancer, Adenocarcinoma, Mesothelioma.

Related studies

• Deep learning with deep convolutional neural network using FDG-PET/CT for malignant pleural mesothelioma diagnosis — K. Kitajima, H. Matsuo, et al. (2021)
• Convolutional Neural Networks for Segmentation of Malignant Pleural Mesothelioma: Analysis of Probability Map Thresholds (CALGB 30901, Alliance) — M. Shenouda, E. Gudmundsson, et al. (2023)
• Informative Power Evaluation of Clinical Parameters to Predict Initial Therapeutic Response in Patients with Advanced Pleural Mesothelioma: A Machine Learning Approach — R. Massafra, A. Catino, et al. (2022)
• Identification of glycolysis genes signature for predicting prognosis in malignant pleural mesothelioma by bioinformatics and machine learning — Y. Xiao, W. Huang, et al. (2022)
• Cancer-Associated Fibroblasts Influence Survival in Pleural Mesothelioma: Digital Gene Expression Analysis and Supervised Machine Learning Model — S. Borchert, A. Mathilakathu, et al. (2023)
• Towards a potential pan-cancer prognostic signature for gene expression based on probesets and ensemble machine learning — D. Chicco, A. Alameer, et al. (2022)

Browse return to medical research.