Sertan Girgin

A novel report generation approach for medical applications: The SISDS methodology and its applications

BACKGROUND Despite exciting innovation in information system technologies, the medical reporting has remained static for a long time. Structured reporting was established to address the deficiencies in report content but has largely failed in its adoption due to concerns over workflow and productivity. The methods used in medical reporting are insufficient in providing with information for statistical processing and medical decision making as well as high quality healthcare. OBJECTIVE The aim of this study is to introduce a novel method that enables professionals to efficiently produce medical reports that are less error-prone and can be used in decision support systems without extensive post-processing. METHODOLOGY We first present the formal definition of the proposed method, called SISDS, that provides a clear separation between the data, logic and presentation layers. It allows free-text like structured data entry in a structured form, and reduces the cognitive effort by inline editing and dynamically controlling the information flow based on the entered data. Then, we validate the usability and reliability of the method on a real-world testbed in the field of radiology. For this purpose, a sample esophagus report was constructed by a focus group of radiologists and real patient data have been collected using a web-based prototype; these data are then used to build a decision support system with off-the-shelf tools. The usability of the method is assessed by evaluating its acceptability by the users and the accuracy of the resulting decision support system. For reliability, we conducted a controlled experiment comparing the performance of the method to that of transcriptionist-oriented systems in terms of the rate of successful diagnosis and the total time required to enter the data. RESULT The most noticeable observation in the evaluation is that the rate of successful diagnosis improves significantly with the proposed method; in our case study, a success rate of 81.25% has been achieved by using the SISDS method compared to 43.75% for the transcriptionist-oriented system. In addition, the average time required to obtain the final approved reports decreased from 29 min to 14 min. Based on questionnaire responses, the acceptance rate of the SISDS methodology by users is also found to outperform the rates of the current methods. CONCLUSION The empirical results show that the method can effectively help to reduce medical errors, increase data quality, and lead to more accurate decision support. In addition, the dynamic hierarchical data entry model proves to provide a good balance between cognitive load and structured data collection.

Basis Expansion in Natural Actor Critic Methods

In reinforcement learning, the aim of the agent is to find a policy that maximizes its expected return. Policy gradient methods try to accomplish this goal by directly approximating the policy using a parametric function approximator; the expected return of the current policy is estimated and its parameters are updated by steepest ascent in the direction of the gradient of the expected return with respect to the policy parameters. In general, the policy is defined in terms of a set of basis functions that capture important features of the problem. Since the quality of the resulting policies directly depend on the set of basis functions, and defining them gets harder as the complexity of the problem increases, it is important to be able to find them automatically. In this paper, we propose a new approach which uses cascade-correlation learning architecture for automatically constructing a set of basis functions within the context of Natural Actor-Critic (NAC) algorithms. Such basis functions allow more complex policies be represented, and consequently improve the performance of the resulting policies. We also present the effectiveness of the method empirically.

Managing advertising campaigns — an approximate planning approach

We consider the problem of displaying commercial advertisements on web pages, in the “cost per click” model. The advertisement server has to learn the appeal of each type of visitor for the different advertisements in order to maximize the profit. Advertisements have constraints such as a certain number of clicks to draw, as well as a lifetime. This problem is thus inherently dynamic, and intimately combines combinatorial and statistical issues. To set the stage, it is also noteworthy that we deal with very rare events of interest, since the base probability of one click is in the order of 10−4. Different approaches may be thought of, ranging from computationally demanding ones (use of Markov decision processes, or stochastic programming) to very fast ones.We introduce NOSEED, an adaptive policy learning algorithm based on a combination of linear programming and multi-arm bandits. We also propose a way to evaluate the extent to which we have to handle the constraints (which is directly related to the computation cost). We investigate the performance of our system through simulations on a realistic model designed with an important commercial web actor.

A bilinear interpolation based approach for optimizing hematoxylin and eosin stained microscopical images

Hematoxylin & Eosin (HE this helps to ease the diagnosis process. However, usually the microscopic digital images obtained using this technique suffer from uneven lighting, i.e. poor Koehler illumination. The existing ad-hoc methods for correcting this problem generally work in RGB color model, and may result in both an unwanted color shift and loosing essential details in terms of the diagnosis. The aim of this study is to present an alternative method that remedies these deficiencies. We first identify the characteristics of uneven lighting in pathological images produced by using the H&E technique, and then show how the quality of these images can be improved by applying an interpolation based approach in the Lab color model without losing any important detail. The effectiveness of the proposed method is demonstrated on sample microscopic images.

Feature discovery in approximate dynamic programming

Feature discovery aims at finding the best representation of data. This is a very important topic in machine learning, and in reinforcement learning in particular. Based on our recent work on feature discovery in the context of reinforcement learning to discover a good, if not the best, representation of states, we report here on the use of the same kind of approach in the context of approximate dynamic programming. The striking difference with the usual approach is that we use a non parametric function approximator to represent the value function, instead of a parametric one. We also argue that the problem of discovering the best state representation and the problem of the value function approximation are just the two faces of the same coin, and that using a non parametric approach provides an elegant solution to both problems at once.

Developing Diagnostic DSSs Based on a Novel Data Collection Methodology

Although necessary information on prognostic implications is missing and reliable data are available in very few areas of medicine, there is an increasing demand for diagnostic decision support systems (DDSS), mainly due to the multitude of variables involved and highly complex relations between them. Unfortunately, existing approaches seem inadequate for providing accurate and high quality data --- a prerequisite for establishing a successful DDSS. In this paper, we demonstrate how SISDS methodology that aims to remedy the deficiencies of current systems in use can be utilized to ease the data collection process and provide opportunities to construct DDSSs without tedious pre-processing and data preparation steps. We also provide empirical results on a real-world testbed application in the field of radiology.

A Novel Multilingual Report Generation System for Medical Applications

There has been an increasing demand for high quality medical data that are in a standard electronic format and easily shared. Although a great amount of effort has been invested to ease the process, an effective solution has yet to be found. In this study, we first discuss necessary features of an effective data collection and reporting system, and then reveal the conceptual view of a novel method that aims to encompass these features. We also present the design and implementation details of a Web-based prototype.