Search

PURPOSE Literature on clinical note mining has highlighted the superiority of machine learning (ML) over hand-crafted rules. Nevertheless, most studies assume the availability of large training sets, which is rarely the case. For this reason, in the clinical setting, rules are still common. We suggest 2 methods to leverage the knowledge encoded in pre-existing rules to inform ML decisions and obtain high performance, even with scarce annotations. METHODS We collected 501 prostate pathology reports from 6 American hospitals. Reports were split into 2,711 core segments, annotated with 20 attributes describing the histology, grade, extension, and location of tumors. The data set was split by institutions to generate a cross-institutional evaluation setting. We assessed 4 systems, namely a rule-based approach, an ML model, and 2 hybrid systems integrating the previous methods: a Rule as Feature model and a Classifier Confidence model. Several ML algorithms were tested, including logistic regression (LR), support vector machine (SVM), and eXtreme gradient boosting (XGB). RESULTS When training on data from a single institution, LR lags behind the rules by 3.5% (F1 score: 92.2% v 95.7%). Hybrid models, instead, obtain competitive results, with Classifier Confidence outperforming the rules by +0.5% (96.2%). When a larger amount of data from multiple institutions is used, LR improves by +1.5% over the rules (97.2%), whereas hybrid systems obtain +2.2% for Rule as Feature (97.7%) and +2.6% for Classifier Confidence (98.3%). Replacing LR with SVM or XGB yielded similar performance gains. CONCLUSION We developed methods to use pre-existing handcrafted rules to inform ML algorithms. These hybrid systems obtain better performance than either rules or ML models alone, even when training data are limited.

PURPOSE Metal-oxide-semiconductor field-effect transistors (MOSFETs) serve as a helpful tool for organ radiation dosimetry and their use has grown in computed tomography (CT). While different approaches have been used for MOSFET calibration, those using the commonly available 100 mm pencil ionization chamber have not incorporated measurements performed throughout its length, and moreover, no previous work has rigorously evaluated the multiple sources of error involved in MOSFET calibration. In this paper, we propose a new MOSFET calibration approach to translate MOSFET voltage measurements into absorbed dose from CT, based on serial measurements performed throughout the length of a 100-mm ionization chamber, and perform an analysis of the errors of MOSFET voltage measurements and four sources of error in calibration. METHODS MOSFET calibration was performed at two sites, to determine single calibration factors for tube potentials of 80, 100, and 120 kVp, using a 100-mm-long pencil ion chamber and a cylindrical computed tomography dose index (CTDI) phantom of 32 cm diameter. The dose profile along the 100-mm ion chamber axis was sampled in 5 mm intervals by nine MOSFETs in the nine holes of the CTDI phantom. Variance of the absorbed dose was modeled as a sum of the MOSFET voltage measurement variance and the calibration factor variance, the latter being comprised of three main subcomponents: ionization chamber reading variance, MOSFET-to-MOSFET variation and a contribution related to the fact that the average calibration factor of a few MOSFETs was used as an estimate for the average value of all MOSFETs. MOSFET voltage measurement error was estimated based on sets of repeated measurements. The calibration factor overall voltage measurement error was calculated from the above analysis. RESULTS Calibration factors determined were close to those reported in the literature and by the manufacturer (~3 mV/mGy), ranging from 2.87 to 3.13 mV/mGy. The error σV of a MOSFET voltage measurement was shown to be proportional to the square root of the voltage V: σV=cV where c = 0.11 mV. A main contributor to the error in the calibration factor was the ionization chamber reading error with 5% error. The usage of a single calibration factor for all MOSFETs introduced an additional error of about 5-7%, depending on the number of MOSFETs that were used to determine the single calibration factor. The expected overall error in a high-dose region (~30 mGy) was estimated to be about 8%, compared to 6% when an individual MOSFET calibration was performed. For a low-dose region (~3 mGy), these values were 13% and 12%. CONCLUSIONS A MOSFET calibration method was developed using a 100-mm pencil ion chamber and a CTDI phantom, accompanied by an absorbed dose error analysis reflecting multiple sources of measurement error. When using a single calibration factor, per tube potential, for different MOSFETs, only a small error was introduced into absorbed dose determinations, thus supporting the use of a single calibration factor for experiments involving many MOSFETs, such as those required to accurately estimate radiation effective dose.

PURPOSE Accurate, patient-specific radiation dosimetry for CT scanning is critical to optimize radiation doses and balance dose against image quality. While Monte Carlo (MC) simulation is often used to estimate doses from CT, comparison of estimates to experimentally measured values is lacking for advanced CT scanners incorporating novel design features. We aimed to compare radiation dose estimates from MC simulation to doses measured in physical anthropomorphic phantoms using metal-oxide semiconductor field-effect transistors (MOSFETs) in a 256-slice CT scanner. METHODS Fifty MOSFETs were placed in organs within tissue-equivalent anthropomorphic adult and pediatric radiographic phantoms, which were scanned using a variety of chest, cardiac, abdomen, brain, and whole-body protocols on a 256-slice system. MC computations were performed on voxelized CT reconstructions of the phantoms using a highly parallel MC tool developed specifically for diagnostic X-ray energies and rapid computation. Doses were compared between MC estimates and physical measurements. RESULTS The average ratio of MOSFET to MC dose in the in-field region was close to 1 (range, 0.96-1.12; mean ± SD, 1.01 ± 0.04), indicating outstanding agreement between measured and simulated doses. The difference between measured and simulated doses tended to increase with distance from the in-field region. The error in the MC simulations due to the limited number of simulated photons was less than 1%. The errors in the MOSFET dose determinations in the in-field region for a single scan were mainly due to the calibration method and were typically about 6% (8% if the error in the reading of the ionization chamber that was used for the MOSFET calibration was included). CONCLUSIONS Radiation dose estimation using a highly parallelized MC method is strongly correlated with experimental measurements in physical adult and infant anthropomorphic phantoms for a wide range of scans performed on a 256-slice CT scanner. Incorporation into CT scanners of radiation-dose distribution estimation, employing the scanner's reconstructed images of the patient, may offer the potential for accurate patient-specific CT dosimetry.

In prostate cancer external beam radiation therapy (EBRT), intra-fraction prostate drifts may compromise the treatment efficacy by underdosing the target and/or overdosing the organs at risk. In this study, a recently developed real-time adaptive planning strategy for intensity-modulated radiation therapy (IMRT) for prostate cancer was evaluated in hypofractionated regimes against traditional treatment planning based on a treatment volume margin expansion. The proposed workflow makes use of a &ldquo, library of plans&rdquo, corresponding to possible intra-fraction prostate positions. During delivery, at each beam end, the plan prepared for the position of the prostate closest to the current one is selected and the corresponding beam delivered. This adaptive planning strategy was compared with the traditional approach on a clinical prostate cancer case where different prostate shift magnitudes were considered. Five, six and fifteen fraction hypofractionated schemes were considered for each of these scenarios. When shifts larger than the treatment margin were present, using the traditional approach the seminal vesicles were underdosed by 3&ndash, 4% of the prescribed dose. The adaptive approach instead allowed for correct target dose coverage and lowered the dose on the rectum for each dosimetric endpoint on average by 3&ndash, 4% in all the fractionation schemes. Standard intensity-modulated radiation therapy planning did not always guarantee a correct dose distribution on the seminal vesicles and the rectum. The adaptive planning strategy proposed resulted insensitive to the intra-fraction prostate drifts, produced a dose distribution in agreement with the dosimetric requirements in every case analysed and significantly lowered the dose on the rectum.

The purpose of this study is to determine the conversion factors that enable accurate estimation of the effective dose (ED) used for cardiac 64-MDCT angiography performed for children.Anthropomorphic phantoms representative of 1- and 10-year-old children, with 50 metal oxide semiconductor field-effect transistor dosimeters placed in organs, underwent scanning performed using a 64-MDCT scanner with different routine clinical cardiac scan modes and x-ray tube potentials. Organ doses were used to calculate the ED on the basis of weighting factors published in 1991 in International Commission on Radiological Protection (ICRP) publication 60 and in 2007 in ICRP publication 103. The EDs and the scanner-reported dose-length products were used to determine conversion factors for each scan mode. The effect of infant heart rate on the ED and the conversion factors was also assessed.The mean conversion factors calculated using the current definition of ED that appeared in ICRP publication 103 were as follows: 0.099 mSv · mGyConversion factors determined using the definition of ED in ICRP publication 103 and cardiac, rather than chest, scan coverage suggest that the radiation doses that children receive from cardiac CT performed using a contemporary 64-MDCT scanner are higher than the radiation doses previously reported when older chest conversion factors were used. Additional up-to-date pediatric cardiac CT conversion factors are required for use with other contemporary CT scanners and patients of different age ranges.

Recently, spectral CT has been drawing a lot of attention in a variety of clinical applications primarily due to its capability of providing quantitative information about material properties. The quantitative integrity of the reconstructed data depends on the accuracy of the data corrections applied to the measurements. Scatter correction is a particularly sensitive correction in spectral CT as it depends on system effects as well as the object being imaged and any residual scatter is amplified during the non-linear material decomposition. An accurate way of removing scatter is subtracting the scatter estimated by Monte Carlo simulation. However, to get sufficiently good scatter estimates, extremely large numbers of photons are required, which may lead to unexpectedly high computational costs. Other approaches model scatter as a convolution operation using kernels derived using empirical methods. These techniques have been found to be insufficient in spectral CT due to their inability to sufficiently capture object dependence. In this work, we develop a deep residual learning framework to address both issues of computation simplicity and object dependency. A deep convolution neural network is trained to determine the scatter distribution from the projection content in training sets. In test cases of a digital anthropomorphic phantom and real water phantom, we demonstrate that with much lower computing costs, the proposed network provides sufficiently accurate scatter estimation.

In this paper we present a theoretical description of the accessibility of nucleosomal DNA to proteins. We reassess the classical analysis of Polach and Widom (1995) who demonstrated that proteins (in their case restriction enzymes) gain access to buried binding sites inside a nucleosome through spontaneous unwrapping of DNA from the protein spool. We introduce a straightforward nucleosome model the predictions of which show good agreement with experimental data. By fitting the model to the data we obtain the values of two quantities: the adsorption energy to the histone octamer per length of DNA and the extra length that the DNA needs to unwrap beyond the binding site of an enzyme before the enzyme can act as effectively as on bare DNA. Our results indicate that the effective binding energy is surprisingly low which suggests that the nucleosomal parameters are tuned such that two large energies, the DNA bending energy and the pure adsorption energy, nearly cancel. This paper is based on a lecture presented at the summer school "DNA and Chromosomes 2009: Physical and Biological Applications". We follow the lecture as closely as possible which is why we spend more time than usual on issues that are already well-known in the field, and why we discuss some well-known results from a different perspective.

e18586Background: A typical pathology report contains diagnoses of tissue types. Most pathology departments capture such information in free text, which makes it challenging to extract the informat...

We consider the force acting on a polymer part of whose length is configurationally confined in a tube and the rest of which is free. This situation arises in many different physical contexts, including a flexible synthetic polymer partially confined in a nanopore and a stiff viral genome partially ejected from its capsid. In both cases the force acting to pull the chain molecule out of its confinement is argued to be constant once a few persistence lengths are "free"/"outside". We present Brownian dynamics simulations that confirm the constancy of the force for different chain lengths and illustrate the dependence of the force on the strength of tube confinement. Experimental results are reported for genome ejection from viral capsids, from which we estimate the pulling force to be a few tenths of a piconewton.

We present a theory of the dependence on sequence of the three-dimensional size of large single-stranded (ss) RNA molecules. The work is motivated by the fact that the genomes of many viruses are large ssRNA molecules—often several thousand nucleotides long—and that these RNAs are spontaneously packaged into small rigid protein shells. We argue that there has been evolutionary pressure for the genome to have overall spatial properties—including an appropriate radius of gyration, R g —that facilitate this assembly process. For an arbitrary RNA sequence, we introduce the (thermal) average maximum ladder distance (〈MLD〉) and use it as a measure of the “extendedness” of the RNA secondary structure. The 〈MLD〉 values of viral ssRNAs that package into capsids of fixed size are shown to be consistently smaller than those for randomly permuted sequences of the same length and base composition, and also smaller than those of natural ssRNAs that are not under evolutionary pressure to have a compact native form. By mapping these secondary structures onto a linear polymer model and by using 〈MLD〉 as a measure of effective contour length, we predict the R g values of viral ssRNAs are smaller than those of nonviral sequences. More generally, we predict the average 〈MLD〉 values of large nonviral ssRNAs scale as N 0.67±0.01 , where N is the number of nucleotides, and that their R g values vary as 〈MLD〉 0.5 in an ideal solvent, and hence as N 0.34 . An alternative analysis, which explicitly includes all branches, is introduced and shown to yield consistent results.

We theoretically investigate under what conditions the director field in a spindle-shaped nematic droplet or tactoid obtains a twisted, parity-broken structure. By minimizing the sum of the bulk elastic and surface energies, we show that a twisted director field is stable if the twist and bend elastic constants are small enough compared to the splay elastic constant, but only if the droplet volume is larger than some minimum value. We furthermore show that the transition from an untwisted to a twisted director-field structure is a sharp function of the various control parameters. We predict that suspensions of rigid, rod-like particles cannot support droplets with a parity broken structure, whereas they could possibly occur in those of semi-flexible, worm-like particles., Comment: 20 pages, 9 figures, submitted to Journal of Physics: Condensed Matter

We have examined the early stages of surfactant dissolution and mesophase formation using a dimer-solvent model with phase behaviour representative of surfactant-water systems. We use an orientational order parameter to characterize systematically the appearance of mesophases. We find the process is diffusion controlled, and the appearance of mesophases is governed adiabatically by the equilibrium phase diagram from the earliest point at which the orientational order parameter can reliably distinguish between mesophases, when only a few repeat spacings of the mesophase microstructures are present. In real systems, such a stage would correspond to times of the order of a few microseconds after the initial contact.

Fluorescence resonance energy transfer (FRET) measurements allow one to observe site exposure in nucleosomes, i.e. the transient unwrapping of a part of the wrapped DNA from the histone octamer. In such experiments one can typically distinguish between a closed state and an open state but in principle one might hope to detect several states, each corresponding to a certain number of open binding sites. Here we show that even in an ideal FRET setup it would be hard to detect unwrapping states with intermediate levels of FRET efficiencies. As the unwrapped DNA molecule, modelled here as a wormlike chain, has a finite stiffness, shape fluctuations smear out FRET signals completely from such intermediate states.

Under conditions of low ionic strength and a pH between about 3.5 and 5.0, solutions of purified coat proteins of cowpea chlorotic mottle virus (CCMV) form spherical multi-shell structures in the absence of viral RNA. The native protein shell, which has an outer diameter of about 28 nm and is built up from 180 coat proteins, forms the inner shell and is surrounded by one or two larger concentric shells. We show that counterion release is the main force stabilizing these multi-shell structures, arguing that this compensates for the outer shells not being able to adopt their smaller optimal radius of curvature. This explains why the structures are only stable at low enough ionic strengths and why we never see the larger outer shells separately in solution. The inner surfaces of the shells are positively charged over a large range of pH values, with most of the positive charge residing on the last 10 nm of the N-terminus side of the protein, which sticks out into the interior. Our calculations show that the multi-shell structures are only stable when the outer surface of the shells is negatively charged, which is above a pH of about 3.5. The fact that we do not see multi-shells at high pH (>5.0) is explained by looking at the stability of the pentamers and hexamers of protein dimers, the building blocks of the shells. Although we focus on CCMV multi-shells, structures of this kind should arise generally in any situation where an attraction between concentric spherical (or cylindrical) shells competes with a preferred spontaneous curvature.

A theory is presented for lambda_C, the coefficient of the first-order correction in the density of the collective diffusion coefficient, for protein spheres interacting by electrostatic and adhesive forces. An extensive numerical analysis of the Stokesian hydrodynamics of two moving spheres is given so as to gauge the precise impact of lubrication forces. An effective stickiness is introduced and a simple formula for lambda_C in terms of this variable is put forward. A precise though more elaborate approximation for lambda_C is also developed. These and numerically exact expressions for lambda_C are compared with experimental data on lysozyme at pH 4.5 and a range of ionic strengths between 0.05 M and 2 M., 8 pages, 3 figures, 1 table

We investigate theoretically the fluid-crystal coexistence of solutions of globular charged nanoparticles like proteins and inorganic colloids. The thermodynamic properties of the fluid phase are computed via the optimized Baxter model. This is done specifically for lysozyme and silicotungstates for which the bare adhesion parameters are evaluated via the experimental second virial coefficients. The electrostatic free energy of the crystal is approximated by supposing the cavities in the interstitial phase between the particles are spherical in form. In the salt-free case a Poisson-Boltzmann equation is solved to calculate the effective charge on a particle and a Donnan approximation is used to derive the chemical potential and osmotic pressure in the presence of salt. The coexistence data of lysozyme and silicotungstates are analyzed within this scheme, especially with regard to the ionic-strength dependence of the chemical potentials. The latter agree within the two phases provided some upward adjustment of the effective charge is allowed for., 15 pages, 9 figures

We perform Monte Carlo simulations on the hard-core attractive Yukawa system to test the Optimized Baxter Model that was introduced in [P.Prinsen and T. Odijk, J. Chem. Phys. 121, p.6525 (2004)] to study a fluid phase of spherical particles interacting through a short-range pair potential. We compare the chemical potentials and pressures from the simulations with analytical predictions from the Optimized Baxter Model. We show that the model is accurate to within 10 percent over a range of volume fractions from 0.1 to 0.4, interaction strengths up to three times the thermal energy and interaction ranges from 6 to 20 % of the particle diameter, and performs even better in most cases. We furthermore establish the consistency of the model by showing that the thermodynamic properties of the Yukawa fluid computed via simulations may be understood on the basis of one similarity variable, the stickiness parameter defined within the Optimized Baxter Model. Finally we show that the Optimized Baxter Model works significantly better than an often used, naive method determining the stickiness parameter by equating the respective second virial coefficients based on the attractive Yukawa and Baxter potentials., Comment: 11 pages, 8 figures

A theory is set up of spherical proteins interacting by screened electrostatics and constant adhesion, in which the effective adhesion parameter is optimized by a variational principle for the free energy. An analytical approach to the second virial coefficient is first outlined by balancing the repulsive electrostatics against part of the bare adhesion. A theory similar in spirit is developed at nonzero concentrations by assuming an appropriate Baxter model as the reference state. The first-order term in a functional expansion of the free energy is set equal to zero which determines the effective adhesion as a function of salt and protein concentrations. The resulting theory is shown to have fairly good predictive power for the ionic-strength dependence of both the second virial coefficient and the osmotic pressure or compressibility of lysozyme up to about 0.2 volume fraction., Comment: 40 pages, 9 figures

Tactoids are droplets of a nematic phase that under suitable conditions form in dispersions of elongated colloidal particles. We theoretically study the shape and the director-field configuration of such droplets for the case where a planar anchoring of the director field to the interface is favored. A minimum of four regimes can be identified in which the droplets have a different structure. Large droplets tend to be nearly spherical with a director field that is bipolar if the surface tension is strongly anisotropic and homogeneous if this is not so. Small droplets can become very elongated and spindlelike if the surface tension is sufficiently anisotropic. Depending on the anchoring strength, the director field is then either homogeneous or bipolar. We find that the more elongated the tactoid, the more strongly it resists the crossing over from a homogeneous to a bipolar structure. This should have implications for the nucleation rate of the nematic phase. Our calculations qualitatively describe the size dependence of the aspect ratio of tactoids found in recent experiments.

Based on a simple linear polymer model we derive a scaling relationship between the “maximum-ladder-distance” characterizing the secondary structure of folded, single stranded, RNA and the radius of gyration characterizing its spatial dimensions. Secondary structures are calculated for a large number of viral RNA sequences as well as non-viral sequences of comparable length and nucleotide composition. The results show that viral RNAs fold into significantly smaller 3D structures than non-viral sequences, consistent with an evolutionary pressure to package the viral sequences inside small rigid protein capsids. We also present a theory explaining why the two ends of a folded RNA are generally found close to each other, independent of nucleotide sequence and length.