Multiple deep learning models based on MRI images in discriminating glioblastoma from solitary brain metastases: a multicentre study

Abstract

Objective: Development of a deep learning model for accurate preoperative identification of glioblastoma and solitary brain metastases by combining multi-centre and multi-sequence magnetic resonance images and comparison of the performance of different deep learning models.

Methods: Clinical data and MR images of a total of 236 patients with pathologically confirmed glioblastoma and single brain metastases were retrospectively collected from January 2019 to May 2024 at Provincial Hospital of Shandong First Medical University, and the data were randomly divided into a training set and a test set according to the ratio of 8:2, in which the training set contained 197 cases and the test set contained 39 cases; the images were preprocessed and labeled with the tumor regions. The images were pre-processed and labeled with tumor regions, and different MRI sequences were input individually or in combination to train the deep learning model 3D ResNet-18, and the optimal sequence combinations were obtained by five-fold cross-validation enhancement of the data inputs and training of the deep learning models 3D Vision Transformer (3D Vit), 3D DenseNet, and 3D VGG; the working characteristic curves (ROCs) of subjects were plotted, and the area under the curve (AUC) was calculated. The area under the curve (AUC), accuracy, precision, recall and F1 score were used to evaluate the discriminative performance of the models. In addition, 48 patients with glioblastoma and single brain metastases from January 2020 to December 2022 were collected from the Affiliated Cancer Hospital of Shandong First Medical University as an external test set to compare the discriminative performance, robustness and generalization ability of the four deep learning models.

Results: In the comparison of the discriminative effect of different MRI sequences, the three sequence combinations of T1-CE, T2, and T2-Flair gained discriminative effect, with the accuracy and AUC values of 0.8718 and 0.9305, respectively; after the four deep learning models were inputted into the aforementioned sequence combinations, the accuracy and AUC of the external validation of the 3D ResNet-18 model were 0.8125, respectively, 0.8899, all of which are the highest among all models.

Conclusions: A combination of multi-sequence MR images and a deep learning model can efficiently identify glioblastoma and solitary brain metastases preoperatively, and the deep learning model 3D ResNet-18 has the highest efficacy in identifying the two types of tumours.

Keywords: Companion diagnostic system; Deep learning; Glioblastoma; Magnetic resonance imaging; Solitary brain metastasis.

Fig. 1

Patient enrolment flowchart

Fig. 2

Workflow diagram. Note: LN: Layer Normalization: Mean Long Self-Attention Layer. MVF: Mean Value Filter. LAS: Long Attention Layer FFN: Feedforward Neural Network. 3D ResNet-18, 3D DenseNet, 3D VGG are all based on the CNN network structure, and the schematic diagrams are shown in Fig

Fig. 3

Radar chart of results from the internal test set of four deep learning models. Note: In the internal test set, 3D ResNet-18 performs well and balanced compared to the other three models in all evaluation metrics. This advantage stems from its residual linkage structure that effectively mitigates the gradient vanishing problem, the 3D convolutional layer’s ability to capture voxel-level spatial features, and the parameter efficiency advantage (the number of parameters is only 42% of that of 3D ViT), which ensures the stability of the training and at the same time realizes the optimal balance between feature extraction and classification performance

Fig. 4

Internal validation of ROC curves for different deep learning models

Fig. 5

Radar chart of external validation results for four deep learning models. Note: 3D ResNet-18 has a balanced performance in all evaluation indexes, and its stability and accuracy are better than other models; 3D Vit model Recall is the highest, indicating that its tumor capturing ability is stronger, while its learning ability and identification efficiency are much lower than other models

Fig. 6

External validation ROC curves for different deep learning models