Virtual Library

Start Your Search

Suresh Krishan Yogeswaran



Author of

  • +

    FP02 - Health Services Research/Health Economics (ID 120)

    • Event: WCLC 2020
    • Type: Posters (Featured)
    • Track: Health Services Research/Health Economics
    • Presentations: 1
    • Moderators:
    • Coordinates: 1/28/2021, 00:00 - 00:00, ePoster Hall
    • +

      FP02.07 - Predicting the Future: Using AI to Predict Treatment Outcomes in Lung Cancer (ID 3831)

      00:00 - 00:00  |  Author(s): Suresh Krishan Yogeswaran

      • Abstract
      • Presentation
      • Slides

      Introduction

      In recent years, artificial intelligence (AI) has gained popularity as a tool for processing large datasets. In medicine, computational approaches such as machine learning are increasingly being used in fields such as serum analysis and the assessment of radiological data, dermoscopic images of melanoma, and histopathological specimens. In lung cancer research, a number of studies have already shown that machine learning can be a valuable tool for identifying early-stage disease choosing personalized treatments. However, there is a lack of data on the use of machine learning to predict treatment outcomes and prognosis. In this study, we use machine learning to create prediction models based on our patient data from the International Association for the Study of Lung Cancer (IASLC) staging project database.

      Methods

      Data of 464 lung cancer patients (309 male and 155 female) with a mean age of 64.9±9.4 years (range 32 – 83) were included in this study. Mortality was used as outcome variable for the prediction models. Furthermore, a total of 39 different training variables (patient history, comorbidities, staging data, treatment outcomes, and lab values) were used to train the machine learning models in this study. Random Forest, “Extreme Gradient Boosting”, and regularized regression models were applied to create prediction models for this dataset. Area under the receiver operating characteristic (ROC) curve was used to assess model performance. In addition, the variables of importance per model were analyzed as well.

      Results

      A total of 5 different prediction models were obtained using our training algorithms. The area under the ROC curve values ranged between 0.71 and 0.75, depending on the method of model tuning. Most common variables of importance were: resection margin, N-status, resection extent, and tumor histology.

      Conclusion

      Machine learning models can be used to create prediction models based on patient databases. In the future, these models may provide valuable information in the clinical decision-making process. More prospective studies based on larger datasets are needed to validate these prediction models.

      Only Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login, select "Add to Cart" and proceed to checkout. If you would like to become a member of IASLC, please click here.

      Only Active Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login or select "Add to Cart" and proceed to checkout.