Core Concepts
This study presents different neural network-based classifier algorithms for diagnosing and classifying Anemia, comparing their performance with established models like Feed Forward Neural Network (FFNN), Elman network, and Non-linear Auto-Regressive Exogenous model (NARX).
Abstract
The study focuses on developing an efficient system for the diagnosis and classification of Anemia using Artificial Neural Networks (ANNs). It compares the performance of different ANN models, including Feed Forward Neural Network (FFNN), Elman network, and Non-linear Auto-Regressive Exogenous model (NARX), in accurately identifying and categorizing various types of Anemia.
The key highlights and insights are:
The study uses data from clinical laboratory test results for 230 patients, with 9 input features (age, gender, RBC, HGB, HCT, MCV, MCH, MCHC, WBCs) and 1 output.
The proposed ANN models demonstrate high accuracy in detecting the presence of Anemia, with the Elman model achieving the best performance with a 93.99% accuracy on the test set.
The Elman model outperforms the FFNN and NARX models in terms of precision, recall, and F1 score, indicating its superior ability to correctly identify positive Anemia cases.
The study suggests that the proposed ANN-based approach can be seamlessly integrated into clinical laboratories for automated generation of Anemia patient reports, providing a cost-effective and efficient solution.
Future research directions include exploring the integration of image processing techniques and ANNs to further enhance the diagnostic and classificatory capabilities for Anemia.
Stats
RBC, HGB, HCT, MCV, MCH, MCHC, and WBCs are the key parameters used to diagnose and classify different types of Anemia.
The dataset consisted of 230 samples, with 147 in the training set and 83 in the test set.
The training set included 105 Anemia-positive and 42 Anemia-negative cases, further categorized into 26 microcytic, 40 normocytic, and 39 macrocytic Anemia cases.
Quotes
"The proposed neural network rapidly and accurately detects the presence of the disease."
"The suggested method is affordable and can be deployed on hardware at low costs."