Identifikasi Stroke Menggunakan Metode Transfer learning Arsitektur Convolutional Neural Network Pada Citra CT-scan Kepala
DOI:
https://doi.org/10.26418/pf.v11i3.65242Abstract
Stroke menjadi penyebab terbesar atas kecatatan dan kematian pada masyarakat Indonesia. Tingkat penderita stroke yang tertinggi di wilayah Asia Tenggara adalah Indonesia. Hal tersebut yang menjadi perhatian pada penelitian ini untuk dapat mengidentifikasi citra kepala hasil dari proses pencitraan medis yaitu CT-scan. Salah satu pemanfaatan teknologi IT dalam bidang medis adalah menggantikan fungsi kerja manusia dalam merepresentasikan hasil citra CT-scan dengan kinerja mesin. Teknologi tersebut memanfaatkan arsitektur Convolutional Neural Network (CNN) metode transfer learning. Pada penelitian ini, dataset yang akan digunakan adalah citra CT-scan kepala dari website www.kaggle.com. Beberapa variasi arsitektur yang digunakan adalah arsitektur AlexNet, VGG16, dan GoogLeNet serta variasi optimizer Stochastic Gradient Descent (SGD), AdaGrad, RMSProp, dan Adam. Hasil yang diperoleh adalah Akurasi AlexNet dengan menggunakan optimizer SGD adalah 75%, AdaGrad sebesar 93%, RMSProp menghasilkan 90%, dan Adam sebesar 85%. Hasil akurasi pada arsitektur VGG16 dengan menggunakan optimizer SGD adalah 73%, AdaGrad sebesar 88%, RMSProp menghasilkan 68%, dan Adam sebesar 91. Arsitektur GoogLeNet menghasilkan nilai akurasi dengan menggunakan optimizer SGD sebesar 65%, AdaGrad, RMSProp dan Adam masing masing menghasilkan akurasi sebesar 65%, 84%, 93% dan 85%.Arsitektur GoogLeNet dan AlexNet dengan optimizer yang berbeda yaitu AdaGrad dan RMSProp berhasil memperoleh akurasi tertinggi diantara arsitektur lainnya dengan akurasi sebesar 93%. Perbedaannya hanya waktu yang dibutuhkan kedua arsitektur ini untuk melakukan proses pelatihan yaitu 4 menit 15 detik untuk arsitektur AlexNet optimizer AdaGrad, dan 12 menit 26 detik untuk arsitektur GoogLeNet optimizer RMSProp.
Kata Kunci: Convolutional_Neural_Network_(CNN), CT-scan, Optimizer, Stroke, Transfer_learning.
References
Mushtaq, M.F., Shahroz, M., Aseere, A.M., Shah, H., Majeed, R., Shehzad, D., and Samad, A., 2021, BHCNet: Neural Network-Based Brain Hemorrhage Classification Using Head CT Scan, IEEE Access, 9, 113901–113916.
Venketasubramanian, N., Yudiarto, F.L., and Tugasworo, D., 2022, Stroke Burden and Stroke Services in Indonesia, Cerebrovasc. Dis. Extra, 12 (1), 53–57.
Sakinah, N., Badriyah, T., and Syarif, I., 2020, Analisis Kinerja Algoritma Mesin Pembelajaran untuk Klarifikasi Penyakit Stroke Menggunakan Citra CT Scan, J. Teknol. Inf. dan Ilmu Komput., 7 (4), 833.
Setiawan, W., 2019, Perbandingan Arsitektur Convolutional Neural Network Untuk Klasifikasi Fundus, J. Simantec, 7 (2), 48–53.
Lewicki, T., Kumar, M., Hong, R., and Wu, W., 2020, Intracranial Hemorrhage Detection in CT Scans using Deep Learning, In: IEEE 6th International Conference on Big Data Computing Service and Applications, Big Data Service 2020. IEEE, 169–172.
Gautam, A. and Raman, B., 2021, Towards effective classification of brain hemorrhagic and ischemic stroke using CNN, Biomed. Signal Process. Control, 63 (August 2020), 102178.
Vani, S. and Rao, T.V.M., 2019, An Experimental Approach Towards the Performance Assessment of Various Optimizers on Convolutional Neural Network, Proc. Int. Conf. Trends Electron. Informatics, ICOEI 2019, (Icoei), 331–336.
Poojary, R. and Pai, A., 2019, Comparative Study of Model Optimization Techniques in Fine-Tuned CNN Models, In: nternational Conference on Electrical and Computing Technologies and Applications, ICECTA 2019. ICECTA, 22–25.
Sajja, T.K., Devarapalli, R.M., and Kalluri, H.K., 2019, Lung cancer detection based on CT scan images by using deep transfer learning, Trait. du Signal, 36 (4), 339–344.
Downloads
Published
Issue
Section
License
Copyright Notice using Creative Commons License for an Online Journal
The online journal Prisma Fisika operates under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0). This copyright notice serves to inform authors, readers, and users of the terms and conditions regarding the use and distribution of the journal's content. By submitting or accessing the content, you agree to the following:
1. Content License:
a. All articles, images, and supplementary materials published in Prisma Fisika are licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
b. This license permits users to copy, distribute, and adapt the content for non-commercial purposes, provided appropriate credit is given to the original authors and the journal.
c. If you remix, transform, or build upon the content, you must distribute your contributions under the same license as the original.
2. Attribution:
a. When using or sharing content from Prisma Fisika, you must properly attribute the original authors and the journal.
b. Include the title of the article, the names of all authors, Prisma Fisika, the year of publication, the volume/issue number, and the specific page numbers.
c. Provide a hyperlink to the original article on Prisma Fisika's website, if applicable.
3. Non-Commercial Use:
a. The content published in Prisma Fisika may not be used for commercial purposes without obtaining explicit permission from the copyright holders.
b. Commercial use includes, but is not limited to, reproduction or distribution of the content for financial gain, selling or licensing the content, or using the content in products or services that are sold or monetized.
4. ShareAlike:
a. If you adapt or remix the content from Prisma Fisika, you must distribute your derivative work under the same Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
b. This provision ensures that any modifications, adaptations, or remixes of the content maintain the same open licensing framework.
5. Disclaimer:
a. The views expressed in articles published in Prisma Fisika are solely those of the authors and do not necessarily reflect the opinions or policies of the journal, its editorial board, or affiliated institutions.
b. Prisma Fisika and its editors are not liable for any errors, omissions, or damages arising from the use or interpretation of the published content.
Please note that this Copyright Notice does not supersede any additional permissions or restrictions that may be granted by individual authors or copyright holders for their specific content within Prisma Fisika. For any questions or requests regarding the use of the journal's content, please contact the editorial team at prismafisika@physics.untan.ac.id.
By accessing or contributing to Prisma Fisika, you agree to comply with the terms of this Creative Commons license. Violation of these terms may result in legal consequences as governed by applicable copyright laws.
Date: 24-05-2023
Prisma Fisika
Email: prismafisika@physics.untan.ac.id
