Transfer learning approach in pre-treatment CT images to predict therapeutic response in advanced malignant pleural mesothelioma

Our plain-English summary of Transfer learning approach in pre-treatment CT images to predict therapeutic response in advanced malignant pleural mesothelioma, by A. Fanizzi, A. Catino, S. Bove, et al (Frontiers in Oncology, 2024).

The study

• Authors: A. Fanizzi, A. Catino, S. Bove, M.C. Comes, M. Montrone, A. Sicolo, R. Signorile, P. Perrotti, P. Pizzutilo, D. Galetta, R. Massafra
• Journal: Frontiers in Oncology
• Published: 2024;14:1432188
• Cited by: 3 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Introduction: Malignant pleural mesothelioma (MPM) is a poor-prognosis disease. Owing to the recent availability of new therapeutic options, there is a need to better assess prognosis. The initial clinical response could represent a useful parameter.

Methods: We proposed a transfer learning approach to predict an initial treatment response starting from baseline CT scans of patients with advanced/unresectable MPM undergoing first-line systemic therapy. The therapeutic response has been assessed according to the mRECIST criteria by CT scan at baseline and after two to three treatment cycles. We used three slices of baseline CT scan as input to the pre-trained convolutional neural network as a radiomic feature extractor. We identified a feature subset through a double feature selection procedure to train a binary SVM classifier to discriminate responders (partial response) from non-responders (stable or disease progression).

Results: The performance of the prediction classifiers was evaluated with an 80:20 hold-out validation scheme. We have evaluated how the developed model was robust to variations in the slices selected by the radiologist. In our dataset, 25 patients showed an initial partial response, whereas 13 patients showed progressive or stable disease. On the independent test, the proposed model achieved a median AUC and accuracy of 86.67% and 87.50%, respectively.

Conclusions: The proposed model has shown high performance even by varying the reference slices. Novel tools could help to improve the prognostic assessment of patients with MPM and to better identify subgroups of patients with different therapeutic responsiveness.

Copyright © 2024 Fanizzi, Catino, Bove, Comes, Montrone, Sicolo, Signorile, Perrotti, Pizzutilo, Galetta and Massafra.

In plain English

This study tested whether a computer model could use CT scans taken before treatment to predict how patients with advanced or unresectable malignant pleural mesothelioma would respond to first-line systemic therapy. The authors used scans and clinical data from 38 consecutive patients collected from July 2017 to October 2021. Asbestos exposure was recorded as yes for 12 patients, no for 9, and not available for 17. Response was judged by mRECIST criteria after two to three treatment cycles. 25 patients (65.79%) had a partial response and the rest had stable or progressive disease. Three CT slices per patient were fed to a pre-trained image network (AlexNet), and selected image features were used to train a classifier. On a separate test set of 8 patients, the model reached an AUC of 86.67% and an accuracy of 87.50%. When the chosen slices were shifted, test AUC was 73.33% and accuracy 75.00%. The authors described their data as preliminary but promising. They named the small number of patients as the main limitation and said validation in a larger group is needed.

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: Retrospective machine-learning (radiomics, transfer learning) study of pre-treatment CT images; original research
• Who was studied: 38 consecutive patients with advanced/unresectable pleural mesothelioma undergoing first-line systemic treatment, collected July 2017 to October 2021 at IRCCS Istituto Tumori 'Giovanni Paolo II', Bari, Italy. The sample was split into a training set of 30 patients and a test set of 8 patients.
• Exposure: Not an asbestos exposure study. The 'input' studied was radiomic features extracted by a pre-trained convolutional neural network (AlexNet) from three baseline CT slices per patient. Asbestos exposure was recorded only as a patient characteristic.
• Main results:
  • Patient table: asbestos exposure was recorded as yes in 12 of 38 patients (31.58%), no in 9 (23.68%), and not available in 17 (44.74%).
  • Of the 38 patients, 25 (65.79%) had an initial partial response to therapy. The remaining 13 had stable or progressive disease.
  • The 38 patients were split into a training set of 30 and an independent test set of 8.
  • Training set (30 patients), using slices chosen by the radiologist and 7 selected features: AUC 78.50%, accuracy 80.00%, sensitivity 87.50%, specificity 75% (medians).
  • Test set (8 patients, 5 with partial response), using radiologist-chosen slices: AUC 86.67%, accuracy 87.50% (as given in Table 2 and the abstract; the running text misprints it as 876.67%), sensitivity 80.00%, specificity 100%.
  • With slices randomly shifted two slices up or down (4 selected features): training set AUC 77.88% and accuracy 85.00%; test set AUC 73.33% and accuracy 75.00%.
• The authors' conclusion: The authors concluded that their data on predicting initial response, though preliminary, are promising. They also concluded that the model seemed to remain highly performing when the reference slices were varied, and that further validation in a larger sample is needed.
• Limitations the authors state:
  • The small number of patients was the main limitation.
  • The whole CT slice was used instead of a specific segmented region of interest.
  • Because the features came from a pre-trained network, their meaning cannot be explained.
  • Further validation on a larger sample population is needed.

Topics

Medical terms in this study, explained in our Lexicon: Pleural Mesothelioma, Prognosis, CT Scan, Unresectable, Systemic Therapy, Disease Progression.

Related studies

• 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)
• Recent Advances in Mesothelioma Treatment: Immunotherapy, Advanced Cell Therapy, and Other Innovative Therapeutic Modalities — R. Suraya, T. Nagano, et al. (2025)
• Accelerating Whole-Body Diffusion-weighted MRI with Deep Learning-based Denoising Image Filters — K. Zormpas-Petridis, N. Tunariu, et al. (2021)
• Asbestos Detection with Fluorescence Microscopy Images and Deep Learning — C. Cai, T. Nishimura, et al. (2021)
• Construction of Asbestos Slate Deep-Learning Training-Data Model Based on Drone Images — S.C. Baek, K.H. Lee, et al. (2023)
• A histomorphological atlas of resected mesothelioma discovered by self-supervised learning from 3446 whole-slide images — F. Seyedshahi, K. Rakovic, et al. (2025)

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