Pengembangan Model Deep Learning Hierarchical Classification Untuk Deteksi Penyakit Skabies Dan Dermatitis Menggunakan Citra Makroskopis Dan Data Klinis Pasien

Sain, Anabela Aulia (2026) Pengembangan Model Deep Learning Hierarchical Classification Untuk Deteksi Penyakit Skabies Dan Dermatitis Menggunakan Citra Makroskopis Dan Data Klinis Pasien. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5049221008-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (24MB) | Request a copy

Abstract

Skabies adalah penyakit kulit infeksius akibat infestasi tungau Sarcoptes scabiei var. hominis yang menimbulkan gatal hebat dan ruam kemerahan, terutama pada malam hari. Di Indonesia, prevalensi skabies masih tinggi pada populasi berisiko, sedangkan kesalahan diagnosis awal dapat terjadi karena gejalanya mirip dengan dermatitis dan tanda khas berupa terowongan tungau sering tidak terlihat, terutama pada kasus awal atau lesi yang telah berubah akibat kortikosteroid. Kondisi ini menyulitkan penegakan diagnosis apabila hanya mengandalkan pemeriksaan visual. Penelitian ini bertujuan mengembangkan model deep learning dengan pendekatan hierarchical classification untuk mendeteksi Skabies, Dermatitis Atopik, dan Dermatitis Kontak menggunakan citra makroskopis dan data klinis pasien. Model dirancang dalam dua level, yaitu Level 1 untuk membedakan Skabies dan Non-Skabies, serta Level 2 untuk mengklasifikasikan Non-Skabies menjadi Dermatitis Atopik atau Dermatitis Kontak. Citra makroskopis diolah menggunakan CNN, sedangkan data klinis pasien diolah menggunakan ANN setelah dikonversi ke format numerik. Kedua sumber data kemudian diintegrasikan melalui model penggabungan dan dievaluasi. Hasil evaluasi pada test set menunjukkan bahwa model penggabungan memberikan kinerja lebih baik dibandingkan CNN-only dan ANN-only. Pada dataset primerv2_47, model penggabungan memperoleh akurasi 0,8000 dan F1-macro tiga kelas 0,7222. Hasil tersebut menunjukkan bahwa integrasi data visual dan klinis dapat meningkatkan kinerja klasifikasi. Model yang dikembangkan berpotensi digunakan sebagai alat bantu klasifikasi penyakit kulit, tetapi tidak menggantikan diagnosis klinis oleh dokter.
==================================================================================================================================
Scabies is an infectious skin disease caused by infestation with the mite Sarcoptes scabiei var. hominis, characterized by severe itching and erythematous rashes, particularly at night. In Indonesia, the prevalence of scabies remains high among at-risk populations, while initial misdiagnosis may occur because its clinical manifestations resemble dermatitis and the characteristic sign of mite burrows is often not visible, especially in early-stage cases or in lesions altered by corticosteroid use. This condition complicates diagnosis when it relies solely on visual examination. This study aimed to develop a deep learning model using a hierarchical classification approach to detect scabies, atopic dermatitis, and contact dermatitis based on macroscopic images and patient clinical data. The model was designed in two levels: Level 1 differentiated scabies from non-scabies, while Level 2 classified non-scabies cases into atopic dermatitis or contact dermatitis. Macroscopic images were processed using a Convolutional Neural Network, whereas patient clinical data were processed using an Artificial Neural Network after being converted into a numerical format. Information from both data was subsequently integrated within a multimodal classification framework and evaluated. The evaluation results on the test set showed that the integrated model achieved better performance than CNN-only and ANN-only models. In the primerv2_47 dataset, the integrated model obtained an accuracy of 0.8000 and a three-class macro F1-score of 0.7222. These findings indicate that integrating visual and clinical data can improve classification performance. The developed model has the potential to be used as an assistive tool for skin disease classification but is not intended to replace clinical diagnosis by physicians.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deep Learning, Dermatitis Atopik, Dermatitis Kontak, Hierarchical Classification, Skabies, Deep Learning, Dermatitis Atopic, Dermatitis Contact, Hierarchical Classification, Scabies
Subjects:	R Medicine > RL Dermatology T Technology > T Technology (General) > T58.6 Management information systems
Divisions:	Faculty of medicine and health (MEDICS) > Medical Technology > 11503-(S1) Undergraduate Thesis
Depositing User:	Anabela Aulia Sain
Date Deposited:	03 Aug 2026 06:48
Last Modified:	03 Aug 2026 06:48
URI:	http://repository.its.ac.id/id/eprint/142060

Actions (login required)

View Item