A novel decision support system based on feature extraction, feature ranking, feature selection for accurate diagnosis of diseases using medical images

33rd International Symposium on Pharmaceutical and Biomedical Analysis, Ankara, Türkiye, 2 - 06 Temmuz 2023, ss.64-67

The continuing shortage of medical professionals around

the world steadily increases their workload while preventing

patients from accessing appropriate, prompt, and affordable

health services. Therefore, it becomes more and more important

to develop appropriate decision support systems that will reduce

the workload of medical professionals while ensuring that patients

receive appropriate and prompt health services. This study

proposes a new approach for rapid and accurate diagnosis of

diseases using medical images obtained through different medical

imaging modalities. The proposed approach is based on the

combined use of feature extraction using AlexNet and ResNet-101

neural networks, feature ranking with two-sample z-test, and

feature selection with binary genetic algorithm. From this respect,

the proposed approach is an innovative decision support system

that provides rapid and accurate classification of medical images

using much fewer features and can be implemented on low-cost

hardware. Performance of the proposed approach was evaluated

using three different datasets by implementing k-Nearest

Neighbours (KNN) and Support Vector Machine (SVM)

classifiers with 4-fold cross validation. Obtained results show

that the proposed approach is very successful in highly accurate

classification of medical images despite utilizing much fewer

features leading to significantly lower computational cost.