Skripsi

EXPLAINABLE AI (XAI) UNTUK DETEKSI PNEUMONIA DARI CITRA X-RAY MENGGUNAKAN GRAD-CAM : STUDI KOMPARATIF ARSITEKTUR CNN

---

Pneumonia remains a leading cause of mortality among toddlers globally, necessitating rapid and accurate early detection methods. While Convolutional Neural Networks (CNN) have proven reliable in medical image classification, their "black-box" nature often hinders clinical adoption due to a lack of transparency. This study aims to bridge this gap through an Explainable AI (XAI) approach using Gradient-weighted Class Activation Mapping (Grad-CAM). A comparative study was conducted on four CNN architectures—VGG16, DenseNet, MobileNet, and EfficientNet-B0—to detect pneumonia from chest X-ray images. The methodology included collecting a public X-ray dataset from Kaggle, preprocessing, model training, and a dual evaluation covering classification performance metrics and explainability scores based on visual assessments by medical experts. The results showed that all models achieved over 85% accuracy, with MobileNet recording the highest accuracy of 99.6% and DenseNet 99.4%. However, the explainability evaluation revealed that EfficientNet-B0, despite a lower accuracy of 86.9%, provided the best visual interpretation quality with an explainability score of 71.25%. Conversely, models with the highest accuracy often focused on activation areas that were less clinically relevant. In conclusion, EfficientNet-B0 is recommended as the most balanced model between accuracy performance and explainability to support trustworthy medical decisions.

---

Availability

Detail Information

Series Title

-

Call Number

T1893032025

Publisher

Palembang : Prodi Magister Ilmu Komputer, Fakultas Ilmu Komputer Universitas Sriwijaya., 2025

Collation

xvii, 161 hlm.; ilus.; tab.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

006.320 7

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Prodi Magister Ilmu Komputer
Jaringan Saraf--Neural Networks

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment