| Autor: |
Rachmad, Aeri, Husni, Hutagalung, Juniar, Hapsari, Dian, Hernawati, Suci, Syarief, Mohammad, Rochman, Eka Mala Sari, Asmara, Yuli Panca |
| Předmět: |
|
| Zdroj: |
Mathematical Modelling of Engineering Problems; Oct2024, Vol. 11 Issue 10, p2664-2670, 7p |
| Abstrakt: |
Tuberculosis (TB) remains a significant global issue, particularly in countries with low economic status and limited healthcare systems. One of the primary challenges is accurate early diagnosis, especially through microscopic examination of sputum samples. However, subjective interpretation and variations in microscopic image quality often hinder diagnostic accuracy. In recent years, the use of Convolutional Neural Networks (CNN) has increased to enhance TB diagnosis effectiveness. This study utilizes the EfficientNet architecture to understand the model's effectiveness in detecting TB in medical images. The dataset used consists of 1266 images, divided into training and testing data with a ratio of 70:30. Additionally, a median filter technique was applied for image preprocessing. Several optimization algorithms are used in this research, namely RMSprop, Stochastic Gradient Descent (SGD), Adam, and Stochastic Gradient Descent with Momentum (SGDM), to find the best scenario. The test results show that Adam optimization provides the best performance compared to the others. The results showed excellent performance, with a low loss rate (9.20%) and high accuracy (98.03%). The relatively fast model training time (122.81 seconds) also adds to the model's efficiency value. This confirms that EfficientNet B0 is an attractive choice for TB classification, with the hope that further development will improve accuracy and efficiency in diagnosing this disease. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
