Radiomics machine learning study with a small sample size: Single random training-test set split may lead to unreliable results

Abstract

This study aims to determine how randomly splitting a dataset into training and test sets affects the estimated performance of a machine learning model and its gap from the test performance under different conditions, using real-world brain tumor radiomics data. We conducted two classification tasks of different difficulty levels with magnetic resonance imaging (MRI) radiomics features: (1) “Simple” task, glioblastomas [n = 109] vs. brain metastasis [n = 58] and (2) “difficult” task, low- [n = 163] vs. high-grade [n = 95] meningiomas. Additionally, two undersampled datasets were created by randomly sampling 50% from these datasets. We performed random training-test set splitting for each dataset repeatedly to create 1,000 different training-test set pairs. For each dataset pair, the least absolute shrinkage and selection operator model was trained and evaluated using various validation methods in the training set, and tested in the test set, using the area under the curve (AUC) as an evaluation metric. The AUCs in training and testing varied among different training-test set pairs, especially with the undersampled datasets and the difficult task. The mean (±standard deviation) AUC difference between training and testing was 0.039 (±0.032) for the simple task without undersampling and 0.092 (±0.071) for the difficult task with undersampling. In a training-test set pair with the difficult task without undersampling, for example, the AUC was high in training but much lower in testing (0.882 and 0.667, respectively); in another dataset pair with the same task, however, the AUC was low in training but much higher in testing (0.709 and 0.911, respectively). When the AUC discrepancy between training and test, or generalization gap, was large, none of the validation methods helped sufficiently reduce the generalization gap. Our results suggest that machine learning after a single random training-test set split may lead to unreliable results in radiomics studies especially with small sample sizes.

Developing Machine Learning Algorithms to Support Patient-centered, Value-based Carpal Tunnel Decompression Surgery

“…The authors have not utilized a validation set and as such the proposed ML models are likely overfitted to the test set and report a misleading higher area under the curve performance. 5…”

Ten simple rules for reporting machine learning methods implementation and evaluation on biomedical data

“…Recent work has shown that using ML methods after a single random training-validation dataset split may yield unreliable results. 17 Thus, to assess the method's out-ofsample performance and to minimize the risk of overfitting, cross-validation offers a simple way to test the accuracy of the method for new (independent) data. For instance, k-fold cross validation procedures can provide unbiased estimates of the true generalization performance for feature selection, model comparison, or classification accuracy.…”

“…Recent work has shown that using ML methods after a single random training‐validation dataset split may yield unreliable results 17 . Thus, to assess the method's out‐of‐sample performance and to minimize the risk of over‐fitting, cross‐validation offers a simple way to test the accuracy of the method for new (independent) data.…”

<strong>Radiomics in Antineoplastic Agents Development: Application and Challenge in Response Evaluation</strong>

“…[7] In addition, a study demonstrated that even with cross-validation, studies with a small sample size are still unreliable. [38] Therefore, multi-center research is necessary to obtain sufficient sample size and avoid unstable and suboptimal results.…”