Most of neuroimaging applications tend to still rely on expert knowledge in determining anatomies of the brain. For example in Parkinson's disease surgery, detection of the anterior commissure (AC) and posterior commissure (PC) are still done manually by doctors. Previously, various methods have been developed related to the automatic detection of AC and PC. However, the majority of previously methods have several drawbacks, such as only compatible on T1-W or T2-W, only compatible for data with the same matrix size, and requires a time-consuming training process. This study proposes a new strategy by combining a multilevel thresholding and morphological relationships approach for automatic detection of AC and PC. The process divided into 4 main stages: preprocessing, multilevel thresholding, segmentation, and detection of AC and PC. The segmentation is performed on several anatomies of the brain including corpus callosum, fornix, and colliculus. From the experiment, it can be concluded that the use of multilevel thresholding and morphological relationship was successfully detecting AC and PC with the mean error were 1.02 mm and 1.06 mm, respectively. The proposed method can perform an automatic detection of AC and PC with simply algorithm, does not require a large of diverse data sets for the training process, without training process that takes up time, and reliable on the diversity of MRI since it is compatible for T1-W and T2-W with various matrix sizes of 256 x 256 and 512 x 512 pixels which cannot be handled by previous researches.

3) ABSTRAK Memahami makna utama yang terkandung dalam beberapa dokumen tentu tidak mudah dan membutuhkan waktu yang cukup lama. Menanggapi masalah tersebut, penelitian terkait peringkasan teks dokumen secara otomatis menjadi perhatian khusus dalam beberapa tahun terakhir. Penelitian ini mengusulkan metode peringkasan teks multi-dokumen yang dapat meningkatkan relevansi antar kalimat dengan menggunakan metode sentence extraction dan word sense disambiguation. Metode sentence extraction yang digunakan didasarkan pada sentence distribution dan part of speech (POS) tagging. Berdasarkan pengujian peringkasan teks dengan metode yang diusulkan, nilai rata-rata ROUGE-1 adalah 0,712, 0,163, 0,247 pada recall, precision, f-measure secara berurutan. Sedangkan hasil pengujian peringkasan teks multi-dokumen tanpa menggunakan word sense disambiguation mendapatkan nilai rata-rata ROUGE-1 sebesar 0,685, 0,139, 0,216 pada recall, precision, fmeasure secara berurutan. Hasil penelitian menunjukkan bahwa penggunaan metode sentence extraction dan word sense disambiguation pada peringkasan teks multi-dokumen dapat meningkatkan kualitas hasil peringkasan teks.

A business process is a set of activities that needs to be considered in organizations or companies. Linear temporal logic (LTL) can models relationships of activities; however, the existing LTL does not consider occurrences probability of relationships of activities based on the event log. Weighted Linear Temporal Logic (W-LTL) extends the existing LTL by giving weights based on the occurrences probabilities. This paper proposes a new similarity method that combines Weighted-Linear Temporal Logic (W-LTL) Tree and Weighted Directed Acyclic Graph (wDAG) that modifies the original wDAG similarity, so it can distinguish the similarity value of two wDAGs that have two branches with opposite weight values. The proposed method (W-LTLDAG) will be verified by comparing with the original wDAG similarity, TPED, Cosine-TDP, and WGED. Based on the comparison, wDAG and WGED gives similarity value of 1 for all experiments, shows that both cannot distinguish weight between 2 graphs. TPED only concerns on relation without giving regards to the number of traces, Cosine-TDP and proposed method are able to distinguish parallel relations that have different occurrence probability of activity relations, but proposed method is proven to give a better calculation by giving a high similarity value, 0.976 for graphs with a small difference value of weights between branches, and low similarity value, 0.327 for graphs with a large difference value of weights between branches.

In interactive image segmentation, distance calculation between regions and sequence of region merging is being an important thing that needs to be considered to obtain accurate segmentation results. Region merging without regard to label in Hierarchical Clustering Analysis causes the possibility of two different labels merged into a cluster and resulting errors in segmentation. This study proposes a new multi-class region merging strategy for interactive image segmentation using the Hierarchical Clustering Analysis. Marking is given to regions that are considered as objects and background, which are then referred as classes. A different label for each class is given to prevent any classes with different label merged into a cluster. Based on experiment, the mean value of ME and RAE for the results of segmentation using the proposed method are 0.035 and 0.083, respectively. Experimental results show that giving the label on each class is effectively used in multi-class region merging.