Diagnosis of mesothelioma with deep learning

Our plain-English summary of Diagnosis of mesothelioma with deep learning, by X. Hu, Z. Yu (Oncology Letters, 2019).

The study

• Authors: X. Hu, Z. Yu
• Journal: Oncology Letters
• Published: 2019;17(2):1483-1490
• Cited by: 10 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Malignant mesothelioma (MM) is a rare but aggressive cancer. The definitive diagnosis of MM is critical for effective treatment and has important medicolegal significance. However, the definitive diagnosis of MM is challenging due to its composite epithelial/mesenchymal pattern. The aim of the current study was to develop a deep learning method to automatically diagnose MM. A retrospective analysis of 324 participants with or without MM was performed. Significant features were selected using a genetic algorithm (GA) or a ReliefF algorithm performed in MATLAB software. Subsequently, the current study constructed and trained several models based on a backpropagation (BP) algorithm, extreme learning machine algorithm and stacked sparse autoencoder (SSAE) to diagnose MM. A confusion matrix, F-measure and a receiver operating characteristic (ROC) curve were used to evaluate the performance of each model. A total of 34 potential variables were analyzed, while the GA and ReliefF algorithms selected 19 and 5 effective features, respectively. The selected features were used as the inputs of the three models. SSAE and GA+SSAE demonstrated the highest performance in terms of classification accuracy, specificity, F-measure and the area under the ROC curve. Overall, the GA+SSAE model was the preferred model since it required a shorter CPU time and fewer variables. Therefore, the SSAE with GA feature selection was selected as the most accurate model for the diagnosis of MM. The deep learning methods developed based on the GA+SSAE model may assist physicians with the diagnosis of MM.

In plain English

This study built and compared computer programs meant to help diagnose malignant mesothelioma (MM). It did not study what causes the disease. The authors used an existing public dataset from the University of California, Irvine machine learning database. It covered 324 participants, described as 97 patients with MM and 227 healthy participants. For each person it recorded 34 variables, including age, asbestos exposure, duration of asbestos exposure, symptoms, and blood and pleural fluid tests. Two selection methods picked out smaller sets of variables (19 and 5). Three types of model were trained on 70% of the data and tested on 30%. A deep learning method called a stacked sparse autoencoder (SSAE) scored 100% on accuracy, specificity, F-measure and area under the curve. It did so both with all 34 variables and with the 19 chosen by a genetic algorithm (GA). The weakest combination had 65.30% accuracy. Using the 19 variables cut the SSAE's mean computing time from 33.2 to 28.8 seconds. The authors concluded that a GA+SSAE-based system may contribute to the definitive diagnosis of MM.

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: Computer modelling study: development and comparison of machine learning diagnostic models, with and without variable (feature) selection, using an existing public dataset
• Who was studied: 324 participants from a dataset in the University of California (Irvine) machine learning database: 97 patients with MM and 227 healthy participants. The article does not say where or when the data were collected. The data were split at random into training (70%) and testing (30%) sets.
• Exposure: The study did not measure exposure itself. Asbestos exposure and duration of asbestos exposure were two of the 34 recorded variables used as model inputs. The others included age, sex, city, symptoms, smoking, and blood and pleural fluid measurements. No fiber types are named.
• Main results:
  • The genetic algorithm (GA) selected 19 of the 34 variables. The ReliefF algorithm selected five.
  • The SSAE model using all 34 variables and the GA+SSAE model using 19 variables each achieved 100% accuracy, specificity, F-measure and AUC.
  • With the 19 GA-selected variables, the backpropagation (GA+BP) and extreme learning machine (GA+ELM) models each reached 98.00% accuracy.
  • The ReliefF+ELM model performed worst: accuracy 65.30%, F-measure 5.48% and AUC 50.00%.
  • Over ten runs, the mean computing (CPU) time of the SSAE model was 33.2 seconds with all variables and 28.8 seconds after GA feature selection.
• The authors' conclusion: The authors concluded that the GA+SSAE algorithm performed best on all evaluation criteria and needed the fewest variables. They concluded that a diagnosis support system based on it may contribute to the definitive diagnosis of MM.

Topics

Medical terms in this study, explained in our Lexicon: 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)
• Deep learning-based classification of mesothelioma improves prediction of patient outcome — P. Courtiol, C. Maussion, et al. (2019)
• Fully automated volumetric measurement of malignant pleural mesothelioma by deep learning AI: validation and comparison with modified RECIST response criteria — A.C. Kidd, O. Anderson, et al. (2022)
• Accelerating Whole-Body Diffusion-weighted MRI with Deep Learning-based Denoising Image Filters — K. Zormpas-Petridis, N. Tunariu, et al. (2021)
• Deep-learning based classification distinguishes sarcomatoid malignant mesotheliomas from benign spindle cell mesothelial proliferations — J.R. Naso, A.B. Levine, et al. (2021)
• A deep learning-based model (DeepMPM) to help predict survival in patients with malignant pleural mesothelioma — W. Li, M. Zhang, et al. (2023)

Browse more research in Oncology Letters, or return to medical research.