International Journal of Data Mining and Bioinformatics (5 papers in press)

Regular Issues

• Early Sepsis Recognition Based on Infrared Thermography  
  by Hasanain Al-Sadr, Mihail Popescu, James Keller 
  Abstract: Systematic screening is crucial for the early diagnosis of sepsis. Detecting abnormal body temperature patterns can accurately predict sepsis before any other symptoms of infection. Therefore, we suggest using thermography as a non-invasive tool capable of continuously measuring body temperature patterns and detecting abnormalities. One such pattern is the temperature difference (ETD) between body extremities such as the outer ear or the tip of the nose, and the core temperature such as the inside of the ear or the inner corner of eye. Specifically, in sepsis, the ETD decreases from about 5-6
  Keywords: Thermography; Sepsis; Ear Localization; Eye-Nose Localization; FCM; KLT Tracking.
  
  
• OpenHI: Open platform for histopathological image annotation  
  by Pargorn Puttapirat, Zhang Haichuan, Jingyi Deng, Yuxin Dong, Jiangbo Shi, Peiliang Lou, Chunbao Wang, Lixia Yao, Xiangrong Zhang, Chen Li 
  Abstract: Consolidating semantically rich annotation on digital histopathological images known as whole-slide images requires a software capable of handling such type of biomedical data with support for procedures which align with existing pathological protocols. Demands for large-scale annotated histopathological datasets are on the raise since they are needed for developments of artificial intelligence techniques to promote automated diagnosis, mass screening, phenotype-genotype association study, etc. This paper presents an open platform for efficient collaborative histopathological image annotation with standardized semantic enrichment at a pixel-level precision named OpenHIOpen Histopathological Image. The frameworks responsive processing algorithm can perform large-scale histopathological image annotation and serve as biomedical data infrastructure for digital pathology. Its web-based design is highly configurable and could be extended to annotate histopathological image of various oncological types. The framework is open-source and fully documented. It is publicly available at https://gitlab.com/BioAI/OpenHI.
  Keywords: OpenHI; digital pathology; whole-slide image; WSI; image annotation; virtual slide; virtual magnification; histopathology; cancer diagnosis; cancer grading; genotype-phenotype association.
  
  
• Identify protein complexes based on PageRank algorithm and architecture on dynamic PPI networks  
  by Xiujuan Lei, Jing Liang, Ling Guo 
  Abstract: Protein-Protein Interactions(PPI) are dynamic in cellular organization. Protein complexes play significant roles in cells. Thus, detecting protein complexes from dynamic PPI networks is realistic. In this paper, we proposed a novel protein complexes prediction algorithm based on core-attachment structure and Pagerank algorithm(PRCA), which run in dynamic PPI networks. This method is divided into three steps. Firstly, calculating the weight value of every protein in dynamic PPI networks to obtain seed proteins. Second, considering triangular structures, cores of protein complexes are acquired. Third, calculating the PageRank value of the adjacent proteins of protein complexes, attachments of protein complexes are appended to their corresponding cores to form protein complexes. This method identifies protein complexes in dynamic PPI networks of DIP, MIPS and Krogan dataset. The experimental results show that PRCA algorithm outperforms other algorithms in precision, recall, f-measure and p-value.
  Keywords: PageRank algorithm; protein complex; PPI network; core-attachment.
  
  
• Unsupervised Representation Learning and Anomaly Detection in ECG Sequences  
  by João Pereira, Margarida Silveira 
  Abstract: While the big data revolution takes place, large amounts of electronic health records, such as electrocardiograms (ECGs) and vital signs data, have become available. These signals are often recorded as time series of observations and are now easier to obtain. In particular, with the arise of smart devices that can perform ECG, there is the quest for developing novel approaches that allow to monitor these signals efficiently, and quickly detect anomalies. However, since most data generated remains unlabelled, the task of anomaly detection is still very challenging.rnUnsupervised representation learning using deep generative models (e.g., variational autoencoders) has been used to learn expressive feature representations of sequences that can make downstream tasks, such as anomaly detection, easier to execute and more accurate.rnWe propose an approach for unsupervised representation learning of ECG sequences using a variational autoencoder parameterised by recurrent neural networks, and use the learned representations for anomaly detection using multiple detection strategies. We tested our approach on the ECG5000 electrocardiogram dataset of the UCR time series classification archive. Our results show that the proposed approach is able to learn expressive representations of ECG sequences, and to detect anomalies with scores that outperform other both supervised and unsupervised methods.
  Keywords: deep learning; representation learning; data mining; bioinformatics; variational autoencoders; recurrent neural networks; time series; anomaly detection; clustering; healthcare; electrocardiogram; unsupervised learning.
  
  
• Tackling imbalance radiomics in acoustic neuroma  
  by Natascha Claudia D'Amico, Mario Merone, Rosa Sicilia, Ermanno Cordelli, Federico D'Antoni, Isa Bossi Zanetti, Giovanni Valbusa, Enzo Grossi, Giancarlo Beltramo, Deborah Fazzini, Giuseppe Scotti, Giulio Iannello, Paolo Soda 
  Abstract: Acoustic neuroma is a primary intracranial tumor of the myelin-forming cells of the 8th cranial nerve. Although it is a slow growing benign tumor, symptoms in the advanced phase can be serious. Hence, controlling tumor growth is essential and stereotactic radiosurgery, which can be performed with the CyberKnife robotic device, has proven effective for managing this disease. However, this approach may have side effects and a follow-up is necessary to assess its efficacy. To optimize the administration of this treatment,in this work we present a machine learning-based radiomics approach that first computes quantitative biomarkers from MR images routinely collected before the CyberKnife treatment and then predicts the treatment response. To tackle the challenge of class imbalance observed in the available dataset we present a cascade of cost-sentitive decision trees. We also experimentally compare the proposed approach with several approaches suited for learning under class skew.The results achieved demonstrate that radiomics has a great potential in predicting patients response to radiosurgery prior to the treatment that, in turns, can reflect into great advantages in therapy planning, sparing radiation toxicity and surgery when unnecessary.
  Keywords: Radiomics; Machine Learning; Imbalance; Learning; Acoustic Neuromas; CyberKnife.