Modelling prognostic power of cardiac tests using rough sets

Rough sets (Pawlak Z. Rough Sets: Theoretical Aspects of Reasoning about Data, Dordrecht: Kluwer Academic Publishers, 1991) is a relatively new approach to representing and reasoning with incomplete and uncertain knowledge. This article introduces the basic concepts of rough sets and Boolean reasoning (Brown FM. Boolean Reasoning: The Logic of Boolean Equations, Dordrecht: Kluwer Academic Publishers, 1990). A rough set framework is then set up to investigate the prognosis of cardiac events in a set of patients with chest pain that was earlier studied by Geleijnse et al. (J Am Coll Cardiol 1996;28(2):447-454). That study used logistic regression to find that the single most important independent predictor for future hard cardiac events (cardiac death or non-fatal myocardial infarction) was an abnormal scintigraphic scan pattern. However, performing a scintigraphic scan is a relatively expensive procedure, and may for some patients not really be fully necessary as knowledge of the outcome of the scan may be redundant with respect to making a prognosis. Using an approach based on rough sets, this paper explores how a patient group in need of a scintigraphic scan can be identified for subsequent modelling. Identification of such patients may potentially contribute to lowering the cost of medical care and to improving its quality since, virtually without loss of information, fewer patients may be referred for this procedure.

Using Boolean reasoning to anonymize databases

This paper investigates how Boolean reasoning can be used to make the records in a database anonymous. In a medical setting, this is of particular interest due to privacy issues and to prevent the possible misuse of confidential information. As electronic medical records and medical data repositories get more common and widespread, the issue of making sensitive data anonymous becomes increasingly important. A theoretically well-founded algorithm is proposed that via cell suppression can be used to make a database anonymous before releasing or sharing it to the outside world. The degree of anonymity can be tailored according to the specific needs of the recipient, and according to the amount of trust we place in the recipient. Furthermore, the required measure of anonymity can be specified as far down as to the individual objects in the database. The algorithm can also be used for anonymization relative to a particular piece of information, effectively blocking deterministic inferences about sensitive database fields.

Rough sets: a knowledge discovery technique for multifactorial medical outcomes

Rough sets is a fairly new and promising technique for data mining and knowledge discovery from databases. This tutorial article presents the fundamentals of rough set theory in a nontechnical manner and outlines how the technique can be used to extract minimal if-then rules from tables of empirical data that either fully or approximately describe given example classifications. An example application for prediction of ambulation for patients with spinal cord injury is given. Because such rules are readily interpretable, they can be inspected to yield possible new insight into how various contributing factors interact and, thus, serve as hypothesis generators for further research. Additionally, the set of mined rules may function as a classifier of new, unseen cases.

π* → n Fluorescence Transition in Formaldehyde in Aqueous Solution: A Combined Quantum Chemical Statistical Mechanical Study

The solvent shift to the fluorescence transition pi --> n in formaldehyde in aqueous solution is theoretically analyzed. The solvent model has explicit representation of the solvent and uses the complete active space state interaction (CASSI) method to obtain a description of the wave function of the solute similar to what the complete active space self-consistent-field (CASSCF) method would give. In the description of the solute-solvent interaction the discrete set of solvent molecules perturb the solute not only through an electrostatic perturbation but also through a nonelectrostatic operator. The latter describes in a way analogous to pseudopotential theory the effect the Pauli principle has on the solute embedded in the solvent. This way the exchange repulsion between solute and solvent is accounted for which therefore can be anisotropic. The best estimate of the average shift is a blue shift of 0.003 eV, and for the current transition the nonelectrostatic perturbation broadens the distribution but has no significant effect on the average shift.

p-Benzoquinone in aqueous solution: Stark shifts in spectra, asymmetry in solvent structure

Results from a simulation of p-benzoquinone (PBQ) in water is presented. An explicit solvent representation is used together with a multiconfigurational ab initio quantum chemical method. The electronic n --> pi* transitions are studied in aqueous solution and the two such transitions are both blue-shifted but to different degree. Both non-equilibrium and many-body effects are found to have decisive influence on the solvation: despite stronger hydrogen bonding between solute and solvent in an excited state than in the ground state, there is a blue-shift, and the solvent structure around the non-polar PBQ is asymmetric, which is argued to come from special many-body effects. The unusual result of strengthened hydrogen bonds in the excited state that follows from an excitation of a non-bonding electron on a proton acceptor, is explained by the near-linear Stark shift that is present in the transition.

Planar or nonplanar: what is the structure of urea in aqueous solution?

A combined quantum chemical statistical mechanical method has been used to study the solvation of urea in water, with emphasis on the structure of urea. The model system consists of three parts: a Hartree-Fock quantum chemical core, 99 water molecules described with a polarizable force-field, and a dielectric continuum. A free-energy profile along the transition of urea from planar to a nonplanar structure is calculated. This mode in aqueous solution is found to be floppy. That is, the structure of urea in water is not well-defined because the planar to nonplanar transition requires an energy of the order of the thermal energy at room temperature. We discuss the implications of this finding for simulation studies of urea in polar environments like water and proteins.

Simulations of the Absorption and Fluorescence of Indole in Aqueous Solution and at a Nonpolar/Polar Interface

Theoretical results are presented on the absorption and fluorescence of indole in aqueous solution as well as at the air/water surface. We use a combined quantum chemical statistical mechanical model with explicit solvent. An approximate ab initio complete active space self-consistent field description of the indole molecule is used, coupled to a discrete polarizable water medium. From the bulk simulations, strong support is found for the interchange mechanism, which explains the unusual solvent shift of the fluorescence of indole or tryptophan in a polar surrounding by a solvent induced switch of the fluorescing state. Two mechanisms are given to explain the different shifts for indole at the interface. First, a dielectric depletion effect, which is expected from the reduction of the amount of polar media. Second, an interface-specific effect, which derives from the stronger hydrogen bond formation at the surface. The latter effect acts to increase the shift for both absorption and emission at the surface as compared to the bulk. From these results, the intrinsic probe photophysics of tryptophan in proteins is discussed in terms of the properties of the protein/solvent interface and the orientation of the amino acid.

A monitor for neutron flux measurements up to 20 MeV

A liquid scintillation detector aimed for neutron energy and fluence measurements in the energy region <20 MeV has been calibrated using monoenergetic and white spectrum neutron fields. Careful measurements of the proton light output function and the response matrix have been performed allowing for the application of unfolding techniques using existing codes. The response matrix is used to characterise monoenergetic neutron fields produced by the T(d,n) at a low-energy deuteron accelerator installed at the Swedish Defense Research Agency (FOI).

Improving machine learning performance by removing redundant cases in medical data sets

Neural network models and other machine learning methods have successfully been applied to several medical classification problems. These models can be periodically refined and retrained as new cases become available. Since training neural networks by backpropagation is time consuming, it is desirable that a minimum number of representative cases be kept in the training set (i.e., redundant cases should be removed). The removal of redundant cases should be carefully monitored so that classification performance is not significantly affected. We made experiments on data removal on a data set of 700 patients suspected of having myocardial infarction and show that there is no statistical difference in classification performance (measured by the differences in areas under the ROC curve on two previously unknown sets of 553 and 500 cases) when as many as 86% of the cases are randomly removed. A proportional reduction in the amount of time required to train the neural network model is achieved.

Comparison of multiple prediction models for ambulation following spinal cord injury

Few studies have properly compared predictive performance of different models using the same medical data set. We developed and compared 3 models (logistic regression, neural networks, and rough sets) in the in prediction of ambulation at hospital discharge following spinal cord injury. We used the multi-center Spinal Cord Injury Model System database. All models performed well and had areas under the receiver operating characteristic curve in the 0.88-0.91 range. All models had sensitivity, specificity, and accuracy greater than 80% at ideal thresholds. The performance of neural network and logistic regression methods was not statistically different (p = 0.48). The rough sets classifier performed statistically worse than either the neural network or logistic regression models (p-values 0.002 and 0.015 respectively).

Kerma coefficients for neutron scattering on 12C and 16O at 96 MeV

Recently, many new applications of fast neutrons are emerging or under development, like dose effects due to cosmic ray neutrons for airplane crew, fast neutron cancer therapy, studies of electronics failure induced by cosmic ray neutrons and accelerator-driven incineration of nuclear waste and energy production technologies. In radiation treatment, the kerma (Kinetic energy release in matter) coefficient, which describes the average energy transferred from neutrons to charged particles, is widely used. The kerma coefficient can be calculated from microscopic nuclear data. Nuclear data above 20 MeV are rather scarce, and more complete nuclear data libraries are needed in order to improve the understanding of the processes occurring on a cellular level. About half the dose in human tissue due to fast neutrons comes from proton recoils in neutron-proton (np) scattering, 10-15% from nuclear recoils due to elastic and inelastic neutron scattering and the remaining 35-40% from neutron-induced emission of light ions. Experimental data on elastic and inelastic neutron scattering at 96 MeV from (12)C and (16)O have been obtained recently at The Svedberg Laboratory in Uppsala, Sweden. These data are shown to be relevant for the determination of nuclear recoil kerma coefficients from elastic and inelastic neutron scattering at intermediate energies.

Modelling cardiac patient set residuals using rough sets

Many medical studies deal with the assessment of the prognostic or diagnostic power of some particular test with respect to some particular medical condition. However, even though a test is deemed to be powerful in this respect, the test may not be strictly needed to perform for everyone. If the test is costly or invasive, this issue is of particular interest. This paper presents a methodology based on rough set theory and Boolean reasoning that can be used to identify those patients for whom performing the test is redundant or superfluous. Furthermore, the methodology enables one to automatically construct a set of descriptive and minimal if-then rules that model the patient group in need of the test. A reanalysis of a previously published real-world dataset of patients with chest pain is used as a case study.

Building manageable rough set classifiers

An interesting aspect of techniques for data mining and knowledge discovery is their potential for generating hypotheses by discovering underlying relationships buried in the data. However, the set of possible hypotheses is often very large and the extracted models may become prohibitively complex. It is therefore typically desirable to only consider the "strongest" hypotheses, so that smaller models can be obtained that also retain good classificatory capabilities. This paper outlines how rule-based classifiers based on rough set theory and Boolean reasoning that are both small and perform well can be developed. Applied to a real-world medical dataset, the final models are shown to exhibit good performance using only a subset of the available information. Furthermore, the number of resulting rules is low and enables practical a posteriori inspection and interpretation of the models.

Precision measurements of the np scattering differential cross section in the intermediate energy region

In fast neutron cancer therapy, approximately 50% of the cell damage is caused by recoil protons from neutron-proton (np) scattering. In the intermediate energy region, there is a need for unambiguous np scattering data with good precision in both the shape of the angular distribution and the absolute normalisation. The normalisation techniques have been reviewed for np scattering measurements as well as recent experimental results, particularly the data obtained at The Svedberg Laboratory at 96 and 162 MeV. In addition, to what extent systematic uncertainties in the np differential cross section might affect the determination of proton recoil kerma coefficients is investigated.

Elastic neutron scattering studies at 96 MeV for transmutation

Elastic neutron scattering from (12)C, (14)N, (16)O, (28)Si, (40)Ca, (56)Fe, (89)Y and (208)Pb has been studied at 96 MeV in the10-70 degrees interval, using the SCANDAL (SCAttered Nucleon Detection AssembLy) facility. The results for (12)C and (208)Pb have recently been published, while the data on the other nuclei are under analysis. The achieved energy resolution, 3.7 MeV, is about an order of magnitude better than for any previous experiment above 65 MeV incident energy. A novel method for normalisation of the absolute scale of the cross section has been used. The estimated normalisation uncertainty, 3%, is unprecedented for a neutron-induced differential cross section measurement on a nuclear target. Elastic neutron scattering is of utmost importance for a vast number of applications. Besides its fundamental importance as a laboratory for tests of isospin dependence in the nucleon-nucleon, and nucleon-nucleus, interaction, knowledge of the optical potentials derived from elastic scattering come into play in virtually every application where a detailed understanding of nuclear processes is important. Applications for these measurements are dose effects due to fast neutrons, including fast neutron therapy, as well as nuclear waste incineration and single event upsets in electronics. The results at light nuclei of medical relevance ((12)C, (14)N and (16)O) are presented separately. In the present contribution, results on the heavier nuclei are presented, among which several are of relevance to shielding of fast neutrons.

Light charged-particle production in 96 MeV neutron-induced reactions on carbon and oxygen

In recent years, an increasing number of applications involving fast neutrons have been developed or are under consideration, e.g. radiation treatment of cancer, neutron dosimetry at commercial aircraft altitudes, soft-error effects in computer memories, accelerator-driven transmutation of nuclear waste and energy production and determination of the response of neutron detectors. Data on light-ion production in light nuclei such as carbon, nitrogen and oxygen are particularly important in calculations of dose distributions in human tissue for radiation therapy at neutron beams, and for dosimetry of high-energy neutrons produced by high-energy cosmic radiation interacting with nuclei (nitrogen and oxygen) in the atmosphere. When studying neutron dose effects, it is especially important to consider carbon and oxygen, since they are, by weight, the most abundant elements in human tissue. Preliminary experimental double-differential cross sections of inclusive light-ion (p, d, t, (3)He and alpha) production in carbon induced by 96-MeV neutrons have been presented. Energy spectra were measured at eight laboratory angles: 20, 40, 60, 80, 100, 120, 140 and 160 degrees. Measurements were performed at The Svedberg Laboratory (TSL), Uppsala, using the dedicated MEDLEY experimental setup. The authors have earlier reported experimental double-differential cross sections of inclusive light-ion production in oxygen. In this paper, the deduced kerma coefficients for oxygen has been presented and compared with reaction model calculations.

Neutron-induced light-ion production from Fe, Pb and U at 96 MeV

Double-differential cross-sections for light-ion production (up to A = 4) induced by 96 MeV neutrons have been measured for Fe, Pb and U. The experiments have been performed at The Svedberg Laboratory in Uppsala, using two independent devices, MEDLEY and SCANDAL. The recorded data cover a wide angular range (20 degrees -160 degrees ) with low energy thresholds. The data have been normalised to obtain cross-sections using np elastic scattering events. The latter have been recorded with the same setup, and results for this measurement are reported. The work was performed within the HINDAS collaboration with the primary aim of improving the database for three of the most important nuclei for incineration of nuclear waste with accelerator-driven systems. The obtained cross-section data are of particular interest for the understanding of the so-called pre-equilibrium stage in a nuclear reaction and will be compared with model calculations.