Breast Cancer (BC) has become a critical illness with a high mortality rate during the previous decade. It is considered the women’s most common cancer. In this paper, we propose two optimum automated BC classification approaches based on a hybridization of the Whale Optimization Algorithm (WOA) and Dragonfly Algorithm (DA), with Radial Basis Function Kernel Support Vector Machines (RBF-SVM), to increase the accuracy of BC classification (CA) by determining the optimum SVM parameters. The effectiveness of the proposed WOA-SVM and DA-SVM algorithms is tested on the Wisconsin Diagnosis Breast Cancer (WDBC) databases and the Wisconsin Breast Cancer Database (WBCD). Various metric parameters such as CA, confusion matrix, the area under the ROC curve (AUC), sensitivity, and specificity are utilized to assess and consider the effectiveness of the proposed approaches. The results are compared not only to the most common optimizers, Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), that are used for training SVM and artificial neural networks (ANN) classifiers, but also to other classification models. The WOA-SVM and DA-SVM are also explored for feature selection, and their findings are compared to the offered models. According to the experimental results, the proposed WOA-SVM method outperforms previous classification approaches on the WBCD dataset. On the WDBC dataset, however, the proposed DA-SVM algorithm outperforms the previous classification algorithms. Using typical datasets’ partition, the resultant CA is as high as 99.65% and 100% for WDBC and WBCD, respectively. However, using a 10-fold cross-validation datasets’ partition, the mean resultant CA are 97.89% and 99.27%, respectively.

In this paper, a technique for the early detection of brain cancer was proposed. The technique depends upon the use of an antenna and a model for the human head. The reflection coefficient (S ₁₁ ) was found for two cases: with and without tumor in the head model. The large increase of S ₁₁ due to the presence of the tumor provides a good indication for tumor detection. It also gives an idea about the size of the tumor. The antenna used was a reconfigurable four-element linear array of squared microstrip patches. Two arrays were designed one circularly polarized, the other linearly polarized. The antenna operates at Industrial Scientific and Medical (ISM) frequency 2.4 GHz. It was designed on FR-4 (lossy) substrate of relative permittivity 4.3 and thickness of 1.6 mm. To feed the array, a corporate feeding network was designed. The reconfigurability of the array was achieved using three single pole double throw (SPDT) switches. Two models of the human head were used; a specific anthropomorphic mannequin (SAM) model, and a 3-D head model consisting of four different head layers: skin, fat, skull and brain. The simulation calculates the reflection coefficient (S ₁₁ ) with and without tumor for circularly polarized (CP) and linearly polarized (LP) linear array. Calculations were taken for four sizes of the array. The best results were obtained with the four-element circularly polarized array. An increase in S ₁₁ of 1188% was obtained. Tumors as small as 5 millimeters (four-layer model) and 2.5 mm (SAM model) can be detected. Specific absorption rate (SAR) was calculated and found to be within the safe limit. A circularly polarized four-element linear antenna array was fabricated. The measured S ₁₁ and radiation pattern are in excellent agreement with simulated ones.

The use of a circularly polarized patch antenna array to detect kidney cancer by microwave techniques is proposed in this paper. A four-element linear antenna array is designed and fabricated at the ISM frequency of 2.4 GHz. The dimensions of the antenna array are 200 mm $\times78$ mm. The single element is a square patch with side length of 30 mm. The distance between patches is chosen to be 20 mm which ensures that mutual coupling between any two adjacent patches is less than 20 dBs. The substrate is a FR-4 material of relative permittivity 4.3 and thickness 1.6 mm. The circular polarization has an axial ratio of 0.8 dB at 2.4 GHz. The bandwidth at S ₁₁ = −10 dB is 7.23 %. Renal system phantom consisting of kidney cortex, renal capsule, ureter, adrenal gland, muscle, fat, and skin is used. Four stages of renal cancer tumors are considered depending upon the tumor size in each stage. The presence of a tumor causes an increase in the reflection coefficient (S ₁₁ ) and a shift in resonance frequency, which can be used to identify cancer. The increase in reflection coefficient and the shift in resonance frequency are calculated for each stage of the cancer tumors. The shift in resonance frequency for the early stages is too small. Therefore, detection depends mainly on the increase in S ₁₁ . The shift in resonance frequency and increase in S ₁₁ are large for advanced stages of the tumor, which makes detection easier. Computed specific absorption rate (SAR) is found to be less than the safety levels, which means this technique is safe to use. Overall, this work suggests a new simple detection technique of kidney cancer. The advantages of this technique are: safe, compact, fast, inexpensive, comfortable examination, non-invasive, and finally non- ionizing radiation during measurement.

Automated and accurate classification of magnetic resonance images (MRIs) of the brain has great importance for medical analysis and interpretation. This paper presents a hybrid optimized classification method to classify the brain tumor by classifying the given magnetic resonance brain image as normal or abnormal. The proposed system implements a gray wolf optimizer (GWO) combined with a supervised artificial neural network (ANN) classifier to achieve enhanced MRI classification accuracy via selecting the optimal parameters of ANN. The introduced GWO–ANN classification system performance is compared to the traditional neural network (NN) classifier using receiver operating characteristic analysis. Experimental results obviously indicate that the presented system achieves a high classification rate and performs much better than the traditional NN classifier.