Michele Berlingerio

Time-Annotated Sequences for Medical Data Mining

The problem of transfer learning, where information gained in one learning task is used to improve performance in another related task, is an important new area of research. While previous work has studied the supervised version of this problem, we study the more challenging case of unsupervised transductive transfer learning, where no labeled data from the target domain are available at training. We describe some current state-of-the-art inductive and transductive approaches and then adapt these models to the problem of transfer learning for protein name extraction. In the process, we introduce a novel maximum entropy based technique, iterative feature transformation (IFT), and show that it achieves comparable performance with state-of-the-art transductive SVMs. We also show how simple relaxations, such as providing additional information like the proportion of positive examples in the test data, can significantly improve the performance of some of the transductive transfer learners.A typical structure of medical data is a sequence of observations of clinical parameters taken at different time moments. In this kind of contexts, the temporal dimension of data is a fundamental variable that should be taken into account in the mining process and returned as part of the extracted knowledge. Therefore, the classical and well established framework of sequential pattern mining is not enough, because it only focuses on the sequentiality of events, without extracting the typical time elapsing between two particular events. Time-annotated sequences (IAS) is a novel mining paradigm that solves this problem. Recently defined in our laboratory [4] together with an efficient algorithm for extracting them, TAS are sequential patterns where each transition between two events is annotated with a typical transition time that is found frequent in the data. In this paper we report a real-world medical case study, in which the TAS mining paradigm is applied to clinical data regarding a set of patients in the follow-up of a liver transplantation. The aim of the data analysis is that of assessing the effectiveness of the extracorporeal photopheresis (ECP) as a therapy to prevent rejection in solid organ transplantation. We believe that this case study does not only show the interestingness of extracting TAS patterns in this particular context but, more ambitiously, it suggests a general methodology for clinical data mining, whenever the time dimension is an important variable of the problem under investigation.

Mining Clinical Data with a Temporal Dimension: A Case Study

Clinical databases store large amounts of information about patients and their medical conditions. Data mining techniques can extract relationships and patterns holding in this wealth of data, and thus be helpful in understanding the progression of diseases and the efficacy of the associated therapies. A typical structure of medical data is a sequence of observations of clinical parameters taken at different time moments. In this kind of contexts, the temporal dimension of data is a fundamental variable that should be taken in account in the mining process and returned as part of the extracted knowledge. Therefore, the classical and well established framework of sequential pattern mining is not enough, because it only focuses on the sequentiality of events, without extracting the typical time elapsing between two particular events. Time-annotated sequences (IAS), is a novel mining paradigm that solves this problem. Recently defined in our laboratory together with an efficient algorithm for extracting them, IAS are sequential patterns where each transition between two events is annotated with a typical transition time that is found frequent in the data. In this paper we report a real-world medical case study, in which the IAS mining paradigm is applied to clinical data regarding a set of patients in the follow-up of a liver transplantation. The aim of the data analysis is that of assessing the effectiveness of the extracorporeal photopheresis (ECP) as a therapy to prevent rejection in solid organ transplantation. For each patient, a set of biochemical variables is recorded at different time moments after the transplantation. The IAS patterns extracted show the values of interleukins and other clinical parameters at specific dates, from which it is possible for the physician to assess the effectiveness of the ECP therapy. We believe that this case study does not only show the interestingness of extracting IAS patterns in this particular context but, more ambitiously, it suggests a general methodology for clinical data mining, whenever the time dimension is an important variable of the problem in analysis.

Mining the Temporal Dimension of the Information Propagation

In the last decade, Social Network Analysis has been a field in which the effort devoted from several researchers in the Data Mining area has increased very fast. Among the possible related topics, the study of the information propagation in a network attracted the interest of many researchers, also from the industrial world. However, only a few answers to the questions How does the information propagates over a network, why and how fast? have been discovered so far. On the other hand, these answers are of large interest, since they help in the tasks of finding experts in a network, assessing viral marketing strategies, identifying fast or slow paths of the information inside a collaborative network. In this paper we study the problem of finding frequent patterns in a network with the help of two different techniques: TAS (Temporally Annotated Sequences) mining, aimed at extracting sequential patterns where each transition between two events is annotated with a typical transition time that emerges from input data, and Graph Mining, which is helpful for locally analyzing the nodes of the networks with their properties. Finally we show preliminary results done in the direction of mining the information propagation over a network, performed on two well known email datasets, that show the power of the combination of these two approaches.

Mining Clinical, Immunological, and Genetic Data of Solid Organ Transplantation

Clinical databases store large amounts of information about patients and their medical conditions. Data mining techniques can extract relationships and patterns implicit in this wealth of data, and thus be helpful in understanding the progression of diseases and the efficacy of the associated therapies. In this perspective, in Pisa (Italy) we have started an important data collection and analysis project, where a very large number of epidemiological, clinical, immunological and genetic variables collected before the transplantation of a solid organ, and during the follow-up assessment of the patients, are stored in a datawarehouse for future mining. This on-going data collection involves all liver, kidney, pancreas and kidney-pancreas transplantations of the last five years of one of the largest (as to number of transplantations) centers in Europe. The project ambitious goal is to gain deeper insights in all the phenomena related to solid organ transplantation, with the aim of improving the donor-recipient matching policy used nowadays. In this chapter we report in details two different data mining activities developed within this project. The first analysis involves mining genetic data of patients affected by terminal hepatic cirrhosis with viral origin (HCV and HBV) and patients with terminal hepatic cirrhosis with non-viral origin (autoimmune): the goal is to assess the influence of the HLA antigens on the course of the disease. In particular, we have evaluated if some genetic configurations of the class I and class II HLA are significantly associated with the triggering causes of the hepatic cirrhosis. The second analysis involves clinical data of a set of patients in the follow-up of a liver transplantation. The aim of the data analysis is that of assessing the effectiveness of the extracorporeal photopheresis (ECP) as a therapy to prevent rejection in solid organ transplantation. For both analyses we describe in details, the medical context and goal, the nature and structure of the data. We also discuss which kind of data mining technique is the most suitable for our purposes, and we describe the details of the knowledge discovery process followed and extracted knowledge.