Strategies to develop radiomics and machine learning models for lung cancer stage and histology prediction using small data samples

Abstract

Predictive models based on radiomics and machine-learning (ML) need large and annotated datasets for training, often difficult to collect. We designed an operative pipeline for model training to exploit data already available to the scientific community. The aim of this work was to explore the capability of radiomic features in predicting tumor histology and stage in patients with non-small cell lung cancer (NSCLC).We analyzed the radiotherapy planning thoracic CT scans of a proprietary sample of 47 subjects (L-RT) and integrated this dataset with a publicly available set of 130 patients from the MAASTRO NSCLC collection (Lung1). We implemented intra-and inter-sample cross-validation strategies (CV) for evaluating the ML predictive model performances with not so large datasets.We carried out two classification tasks: histology classification (3 classes) and overall stage classification (two classes: stage I and II). In the first task, the best performance was obtained by a Random Forest classifier, once the analysis has been restricted to stage I and II tumors of the Lung1 and L-RT merged dataset (AUC = 0.72 ± 0.11). For the overall stage classification, the best results were obtained when training on Lung1 and testing of L-RT dataset (AUC = 0.72 ± 0.04 for Random Forest and AUC = 0.84 ± 0.03 for linear-kernel Support Vector Machine).According to the classification task to be accomplished and to the heterogeneity of the available dataset(s), different CV strategies have to be explored and compared to make a robust assessment of the potential of a predictive model based on radiomics and ML.

Radiomics and deep learning methods for the prediction of 2-year overall survival in LUNG1 dataset

“…Nevertheless, the LUNG1 dataset continues to play a pivotal role: possibly, better models could be achieved when focusing on a different outcome, e.g., stage 21 , histochemical type 14 . Moreover, more accurate models for 2-year OS prediction could be built on other datasets, and could still rely on LUNG1 for their external validation 21 , 26 . This is the advantage of publicly sharing large datasets with the research community 10 .…”

“…In conclusion, the wide range of methods and approaches investigated in this work produced further evidence that the image-based prediction of the 2-year OS on the LUNG1 dataset is a challenging task, with limited margins of improvements. Nevertheless, the LUNG1 dataset continues to play a pivotal role: possibly, better models could be achieved when focusing on a different outcome, e.g., stage 21 , histochemical type 14 . Moreover, more accurate models for 2-year OS prediction could be built on other datasets, and could still rely on LUNG1 for their external validation 21 , 26 .…”

Development and validation of a feature extraction-based logical anthropomorphic diagnostic system for early gastric cancer: A case-control study

“…In clinical practice, the endoscopic images are usually complex and diverse, and some features of the images may escape the naked eye. 28 Even among expert endoscopists, diagnosis often varies widely. 17 With the segmentation of MS and MV in the M-IEE images, ENDOANGEL-LA eliminated the influence of interference information on the endoscopists' observation and simplified the complex endoscopic image.…”

“… 17 With the segmentation of MS and MV in the M-IEE images, ENDOANGEL-LA eliminated the influence of interference information on the endoscopists' observation and simplified the complex endoscopic image. 28 By extracting the vital diagnostic information and marking the most characteristic MV for reference, endoscopists can notice meaningful features of the images, which can enhance the endoscopists’ confidence in diagnosis and may improve the accuracy of diagnosis.…”

Automated Meniscus Segmentation and Tear Detection of Knee MRI with a 3D Mask-RCNN

“…Computer-assisted diagnostic of medical image systems has been proposed to detect abnormalities in the knee joint for early diagnosis and treatment purposes. Numerous radiomics algorithms and deep learning algorithms are employed for the intelligent diagnosis of meniscus lesions in medical imaging due to the advancements in arti cial intelligence (AI) driven by the rise in computing power and improvement in big data management [9][10][11][12][13][14] .…”