Perancangan Alat Bantu Diagnosis Tumor Otak Menggunakan Transfer Learning ResNet101V2
Rhendiya Maulana Zein, Ir. Nazrul Effendy, S.T., M.T., Ph.D., IPM; dr. Endro Basuki. S, Sp.BS(K), M.Kes
2023 | Skripsi | TEKNIK NUKLIR
Brain tumor treatment is carried out by doctors and is required to be fast and accurate. This causes problems such as high workloads, requiring a lot of time, and human error. Diagnostic tool using DL is solution because they can reduce or eliminate human error, are efficient, produce unbiased decisions, and are always available. The DL model was designed using transfer learning ResNet101V2 to classify meningioma, glioma, pituitary, and normal tumors from MRI images. This study will develop a computer-aided diagnostic tool with DL.
The design is carried out in the following stages. They are, first, data collection that addresses implementation, related rules, and data availability. Second, data processing to make the model able to study the data. Third, model engineering designs the architecture and the algorithms so that they can generalize data well. Finally, evaluate the model, which will assess the performance of the model for decision-making for the best model or redesign.
This study produces three models with consideration of model evaluation. Model 3 is the best model with accuracy on the training data, validation, and test sequentially is 98,38%, 96,97%, and 96,15%, while the loss is 0,05, 0,09, and 0,11. Diagnostic tools can be used to predict tumors from an MRI image or a set of MRI images of a patient.
Kata Kunci : Tumor otak, MRI, Transfer Learning, ResNet101V2