Brain Tumor Segmentation and Classification from Sensor-Based Portable Microwave Brain Imaging System Using Lightweight Deep Learning Models

Abstract

Automated brain tumor segmentation from reconstructed microwave (RMW) brain images and image classification is essential for the investigation and monitoring of the progression of brain disease. The manual detection, classification, and segmentation of tumors are extremely time-consuming but crucial tasks due to the tumor's pattern. In this paper, we propose a new lightweight segmentation model called MicrowaveSegNet (MSegNet), which segments the brain tumor, and a new classifier called the BrainImageNet (BINet) model to classify the RMW images. Initially, three hundred (300) RMW brain image samples were obtained from our sensors-based microwave brain imaging (SMBI) system to create an original dataset. Then, image preprocessing and augmentation techniques were applied to make 6000 training images per fold for a 5-fold cross-validation. Later, the MSegNet and BINet were compared to state-of-the-art segmentation and classification models to verify their performance. The MSegNet has achieved an Intersection-over-Union (IoU) and Dice score of 86.92% and 93.10%, respectively, for tumor segmentation. The BINet has achieved an accuracy, precision, recall, F1-score, and specificity of 89.33%, 88.74%, 88.67%, 88.61%, and 94.33%, respectively, for three-class classification using raw RMW images, whereas it achieved 98.33%, 98.35%, 98.33%, 98.33%, and 99.17%, respectively, for segmented RMW images. Therefore, the proposed cascaded model can be used in the SMBI system.

Keywords: Self-ONN; antenna sensor; brain tumor segmentation; classification; deep learning; sensor-based microwave brain imaging system.

Figure 1

Graphic diagram of the 3D antenna sensor: (a) top view, (b) bottom view, (c) perspective view.

Figure 2

Fabricated prototype of the SNTSRR MTM-loaded 3D antenna: (a) top view, (b) side view, (c) perspective view.

Figure 3

Measurement setup and resultant outcomes of the antenna sensor: (a) PNA setup, (b) measured and simulated reflection coefficient, (c) gain, (d) efficiency.

Figure 4

Phantom’s composition process using fabricated four tissues.

Figure 5

Experimental setup for sensors-based microwave brain imaging system [46].

Figure 6

Reflected scattering bio signals received by the receiving antenna sensors: (a) without tumor, (b) with tumor.

Figure 7

Comparision imaging results with simulated as considering real situations and formulated tissue−imitating phantom models with reconstructed images: (a) non-tumor, (b,c) single tumor, (d) double tumors.

Figure 7

Comparision imaging results with simulated as considering real situations and formulated tissue−imitating phantom models with reconstructed images: (a) non-tumor, (b,c) single tumor, (d) double tumors.

Figure 8

The complete methodology flow chart of the research work.

Figure 9

The RMW brain image samples and their corresponding ground truth masks from the original dataset: (a) non-tumor, (b) single tumor, (c) double tumors.

Figure 10

Augmented sample of training set: (a–c) pre-processed non-tumor, single tumor, and double tumor images, (d–f) images after rotation by 20 degrees counterclockwise and clockwise for non-tumor, single tumor, and double tumors, (g–i) images after three percent horizontal, five percent vertical and horizontal, and five percent horizontal and three percent vertical translation for non-tumor, single tumor, and double tumors.

Figure 11

Proposed lightweight MicrowaveSegNet (MSegNet) model for tumor segmentation.

Figure 12

The training results graph: (a) the DSC graph, (b) loss graph.

Figure 13

Proposed BrainImageNet using Self-ONN.

Figure 14

Proposed MSegNet model’s tumor segmentation results with ground truth masks, generated masks, and resultant segmented tumor images for: (a) non-tumor class, (b) single tumor class, (c) double tumors class.

Figure 14

Proposed MSegNet model’s tumor segmentation results with ground truth masks, generated masks, and resultant segmented tumor images for: (a) non-tumor class, (b) single tumor class, (c) double tumors class.

Figure 15

The confusion matrix of the proposed BINet classification model for: (a) the raw RMW brain images, (b) the segmented RMW brain images.

Figure 16

Some misclassified images by the BINet model for the raw RMW images: (a) non-tumor images were misclassified as a single tumor class, (b) single tumor images were misclassified as a non-tumor class, (c) single tumor images were misclassified as a double tumor class, (d) double tumor images were misclassified as a single tumor class.

Figure 16

Some misclassified images by the BINet model for the raw RMW images: (a) non-tumor images were misclassified as a single tumor class, (b) single tumor images were misclassified as a non-tumor class, (c) single tumor images were misclassified as a double tumor class, (d) double tumor images were misclassified as a single tumor class.