Deteksi Tuberkulosis Pada Radiografi Dada Menggunakan Deep Learning DenseNet121

Rahmadina, Azzahra Sekar (2026) Deteksi Tuberkulosis Pada Radiografi Dada Menggunakan Deep Learning DenseNet121. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5027221035-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Tuberkulosis (TB) merupakan penyakit infeksi paru yang masih menjadi tantangan dalam sistem kesehatan global, sehingga diperlukan metode deteksi yang cepat dan akurat. Penelitian ini mengembangkan model deep learning menggunakan arsitektur DenseNet121 untuk mengklasifikasikan citra radiografi dada ke dalam tiga kelas, yaitu Normal, Tuberculosis, dan Non-Tuberculosis. Penelitian ini bertujuan menilai kinerja model, memeriksa konsistensi pelatihan, dan menguji kesesuaian prediksi terhadap label ground truth sebagai representasi expert judgement. Metode penelitian meliputi tahap preprocessing dataset, pelatihan model menggunakan transfer learning, serta evaluasi performa menggunakan metrik akurasi, precision, recall, F1-score, dan AUC-ROC. Untuk menilai stabilitas dan konsistensi model, dilakukan evaluasi menggunakan Stratified K-Fold Cross Validation. Analisis statistik mencakup uji Shapiro–Wilk untuk menilai distribusi akurasi dan uji Chi-Square untuk menilai hubungan antara prediksi model dan label aktual. Pengujian citra individual melalui sistem berbasis web dilakukan untuk mengevaluasi performa model pada konteks penggunaan nyata. Hasil penelitian menunjukkan bahwa DenseNet121 mencapai akurasi 95,1% dengan AUC 0,99–1,00 pada seluruh kelas. Uji Chi-Square menghasilkan p < 0,001 dan Cramer’s V = 0,911 yang menunjukkan hubungan sangat kuat antara prediksi model dan ground truth. Uji Shapiro–Wilk menunjukkan bahwa akurasi berdistribusi normal sehingga performa model dinilai stabil. DenseNet121 terbukti efektif dan memiliki potensi sebagai sistem bantu diagnosis dalam skrining Tuberculosis berbasis radiografi dada.
===================================================================================================================================
Tuberculosis (TB) is a pulmonary infectious disease that remains a major global health challenge, highlighting the need for diagnostic approaches that are faster and more accurate. This study proposes the use of a deep learning method based on the DenseNet121 architecture to classify chest radiography images into three categories: Normal, Tuberculosis, and Non-Tuberculosis. The aim of this research is to evaluate the performance of the model, assess the consistency of its training results, and examine the agreement between model predictions and ground-truth labels representing expert judgement. The methodology includes dataset preprocessing, model training using transfer learning, and performance evaluation based on accuracy, precision, recall, F1-score, and AUC-ROC metrics. To assess model stability and consistency, Stratified K-Fold Cross Validation is applied. Statistical analysis consists of the Shapiro–Wilk test to assess the distribution of accuracy and the Chi-Square test to evaluate the association between predicted and actual labels. Individual image testing through a web-based system is conducted to validate the model’s performance in real-use scenarios. Additional testing was conducted through a web-based system to validate the model’s performance on individual chest X-ray inputs.
The results show that the model achieved an accuracy of 95.1%, with AUC values ranging from 0.99 to 1.00 across all classes, indicating excellent discriminative ability. The Chi-Square test produced p < 0.001 with a Cramer’s V value of 0.911, demonstrating a very strong association between model predictions and ground-truth labels. The normality tests confirmed that the accuracy values followed a normal distribution, indicating stable performance. Overall, DenseNet121 proved to be effective and has strong potential to be used as a clinical decision-support tool for TB screening based on chest radiography.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Tuberkulosis, Radiografi Dada, DenseNet121, Deep learning, Klasifikasi Citra, Chest Radiography, Image Classification
Subjects:	A General Works > AI Indexes (General) A General Works > AI Indexes (General) R Medicine > R Medicine (General) > R858 Deep Learning R Medicine > RZ Other systems of medicine T Technology > T Technology (General) > T57.5 Data Processing T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing. T Technology > TA Engineering (General). Civil engineering (General) > TA174 Computer-aided design.
Divisions:	Faculty of Information and Communication Technology > Information Technology > 59201-(S1) Undergraduate Thesis
Depositing User:	Azzahra Sekar Rahmadina
Date Deposited:	22 Jan 2026 02:41
Last Modified:	22 Jan 2026 02:41
URI:	http://repository.its.ac.id/id/eprint/130032

Actions (login required)

View Item