Oops, you're using an old version of your browser so some of the features on this page may not be displaying properly.

MINIMAL Requirements: Google Chrome 24+Mozilla Firefox 20+Internet Explorer 11Opera 15–18Apple Safari 7SeaMonkey 2.15-2.23

Poster session 18

1912P - AI-based early prediction of radiation pneumonitis in stage III NSCLC patients

Date

14 Sep 2024

Session

Poster session 18

Topics

Radiological Imaging;  Cancer Intelligence (eHealth, Telehealth Technology, BIG Data)

Tumour Site

Non-Small Cell Lung Cancer

Presenters

Samaantha Bove

Citation

Annals of Oncology (2024) 35 (suppl_2): S1115-S1121. 10.1016/annonc/annonc1613

Authors

S. Bove1, M.C. Comes2, A. Fanizzi3, V. Gregorc4, A. Catino5, D. Galetta6, M. Montrone7, R. Massafra8

Author affiliations

  • 1 Biostatistics And Bioinformatics Lab, IRCCS Istituto Tumori Giovanni Paolo II, 70124 - Bari/IT
  • 2 Biostatistics And Bionformatics Laboratory, IRCCS Istituto Tumori Giovanni Paolo II, 70124 - BARI/IT
  • 3 Biostatistics And Bionformatics Laboratory, IRCCS Istituto Tumori Bari, 70124 - bari/IT
  • 4 Oncology Dept., IRCCS Ospedale San Raffaele, 20132 - Milan/IT
  • 5 Thoracic Oncology Department, Istituto Tumori Giovanni Paolo II, 70124 - Bari/IT
  • 6 Medical Thoracic Oncology Unit, Istituto Tumori Bari Giovanni Paolo II - IRCCS, 70124 - Bari/IT
  • 7 Medical Area Department, Thoracic Oncology Unit, Istituto Tumori Bari Giovanni Paolo II - IRCCS, 70124 - Bari/IT
  • 8 Biostatistics And Bionformatics Laboratory, Istituto Tumori Bari Giovanni Paolo II - IRCCS, 70124 - Bari/IT

Resources

Login to get immediate access to this content.

If you do not have an ESMO account, please create one for free.

Abstract 1912P

Background

Treatment approaches for NSCLC patients differ depending on stage, histology, genetic alterations, and patient’s condition. Stage III NSCLC patients are non-surgical candidates and currently treated with chemoradiotherapy eventually followed by immunotherapy. For these patients, the occurrence of treatment-related diseases such as radiation pneumonitis could be observed. Thus, an early identification of which patients are more prone to develop radiation pneumonitis via AI-based models could be crucial to define personalized treatment approaches and improve patients’ prognosis.

Methods

We designed an AI approach to identify Stage III NSCLC patients at high risk of developing radiation pneumonitis. For this purpose, we analyzed pre-treatment CT images belonging to 54 Stage III NSCLC patients afferent to Istituto Tumori “Giovanni Paolo II” of Bari (Italy), in which radiation pneumonitis was observed for 24 patients. For each patient we identified the CT slide presenting the largest tumor area. Starting from this CT slide, we selected other four CT slides, the two immediately preceding and the two immediately following the one with largest tumor area. Then, for all the selected CT slides, we defined a bounding box around the extremal points of the tumour in the four planar x-y dimensions. Finally, we exploited these images to train the last few layers of a pre-trained Convolutional Neural Network, that is ResNet50, adopting a 5-fold cross-validation scheme after identifying each time a validation set containing the 15% of the starting dataset.

Results

The performances of the designed model was summarized in terms of mean and 95% confidence interval. On the test and validation set, our model achieved respectively an AUC value of 90.60% [85.08-93.99] and 62.06% [61.25-67.38], a Sensitivity value of 94.12% [73.03-99.96] and 66.67% [60.02-75.69], a Specificity value of 76.31% [69.43-91.17] and 75.20% [59.52-75.20], and an Accuracy value of 81.81% [73.23-93.63] and 65.50% [62.50-67.88].

Conclusions

The proposed AI-based model represents a promising procedure for early identifying Stage III NSCLC patients at high risk of developing radiation pneumonitis, with the purpose of defining personalized treatment approaches and improving patients’ prognosis.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

R. Massafra.

Funding

Ministry of Health.

Disclosure

All authors have declared no conflicts of interest.

This site uses cookies. Some of these cookies are essential, while others help us improve your experience by providing insights into how the site is being used.

For more detailed information on the cookies we use, please check our Privacy Policy.

Customise settings
  • Necessary cookies enable core functionality. The website cannot function properly without these cookies, and you can only disable them by changing your browser preferences.