Exploring long-term breast cancer survivors' care trajectories using dynamic time warping-based unsupervised clustering.

OBJECTIVES: Long-term breast cancer survivors (BCS) constitute a complex group of patients, whose number is estimated to continue rising, such that, a dedicated long-term clinical follow-up is necessary. MATERIALS AND METHODS: A dynamic time warping-based unsupervised clustering methodology is presented in this article for the identification of temporal patterns in the care trajectories of 6214 female BCS of a large longitudinal retrospective cohort of Spain. The extracted care-transition patterns are graphically represented using directed network diagrams with aggregated patient and time information. A control group consisting of 12 412 females without breast cancer is also used for comparison. RESULTS: The use of radiology and hospital admission are explored as patterns of special interest. In the generated networks, a more intense and complex use of certain healthcare services (eg, radiology, outpatient care, hospital admission) is shown and quantified for the BCS. Higher mortality rates and numbers of comorbidities are observed in various transitions and compared with non-breast cancer. It is also demonstrated how a wealth of patient and time information can be revealed from individual service transitions. DISCUSSION: The presented methodology permits the identification and descriptive visualization of temporal patterns of the usage of healthcare services by the BCS, that otherwise would remain hidden in the trajectories. CONCLUSION: The results could provide the basis for better understanding the BCS' circulation through the health system, with a view to more efficiently predicting their forthcoming needs and thus designing more effective personalized survivorship care plans.

A system-level analysis of patient disease trajectories based on clinical, phenotypic and molecular similarities.

MOTIVATION: Incorporating the temporal dimension into multimorbidity studies has shown to be crucial for achieving a better understanding of the disease associations. Furthermore, due to the multifactorial nature of human disease, exploring disease associations from different perspectives can provide a holistic view to support the study of their aetiology. RESULTS: In this work, a temporal systems-medicine approach is proposed for identifying time-dependent multimorbidity patterns from patient disease trajectories, by integrating data from electronic health records with genetic and phenotypic information. Specifically, the disease trajectories are clustered using an unsupervised algorithm based on dynamic time warping and three disease similarity metrics: clinical, genetic and phenotypic. An evaluation method is also presented for quantitatively assessing, in the different disease spaces, both the cluster homogeneity and the respective similarities between the associated diseases within individual trajectories. The latter can facilitate exploring the origin(s) in the identified disease patterns. The proposed integrative methodology can be applied to any longitudinal cohort and disease of interest. In this article, prostate cancer is selected as a use case of medical interest to demonstrate, for the first time, the identification of temporal disease multimorbidities in different disease spaces. AVAILABILITY AND IMPLEMENTATION: https://gitlab.com/agiannoula/diseasetrajectories. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

comoRbidity: an R package for the systematic analysis of disease comorbidities.

Motivation: The study of comorbidities is a major priority due to their impact on life expectancy, quality of life and healthcare cost. The availability of electronic health records (EHRs) for data mining offers the opportunity to discover disease associations and comorbidity patterns from the clinical history of patients gathered during routine medical care. This opens the need for analytical tools for detection of disease comorbidities, including the investigation of their underlying genetic basis. Results: We present comoRbidity, an R package aimed at providing a systematic and comprehensive analysis of disease comorbidities from both the clinical and molecular perspectives. comoRbidity leverages from (i) user provided clinical data from EHR databases (the clinical comorbidity analysis) and (ii) genotype-phenotype information of the diseases under study (the molecular comorbidity analysis) for a comprehensive analysis of disease comorbidities. The clinical comorbidity analysis enables identifying significant disease comorbidities from clinical data, including sex and age stratification and temporal directionality analyses, while the molecular comorbidity analysis supports the generation of hypothesis on the underlying mechanisms of the disease comorbidities by exploring shared genes among disorders. The open-source comoRbidity package is a software tool aimed at expediting the integrative analysis of disease comorbidities by incorporating several analytical and visualization functions. Availability and implementation: https://bitbucket.org/ibi_group/comorbidity. Supplementary information: Supplementary data are available at Bioinformatics online.

Identifying temporal patterns in patient disease trajectories using dynamic time warping: A population-based study.

Time is a crucial parameter in the assessment of comorbidities in population-based studies, as it permits to identify more complex disease patterns apart from the pairwise disease associations. So far, it has been, either, completely ignored or only, taken into account by assessing the temporal directionality of identified comorbidity pairs. In this work, a novel time-analysis framework is presented for large-scale comorbidity studies. The disease-history vectors of patients of a regional Spanish health dataset are represented as time sequences of ordered disease diagnoses. Statistically significant pairwise disease associations are identified and their temporal directionality is assessed. Subsequently, an unsupervised clustering algorithm, based on Dynamic Time Warping, is applied on the common disease trajectories in order to group them according to the temporal patterns that they share. The proposed methodology for the temporal assessment of such trajectories could serve as the preliminary basis of a disease prediction system.

Reconstructing 3-D maps of the local viscoelastic properties using a finite-amplitude modulated radiation force.

A modulated acoustic radiation force, produced by two confocal tone-burst ultrasound beams of slightly different frequencies (i.e. 2.0 MHz ± Δf/2, where Δf is the difference frequency), can be used to remotely generate modulated low-frequency (Δf ≤ 500 Hz) shear waves in attenuating media. By appropriately selecting the duration of the two beams, the energy of the generated shear waves can be concentrated around the difference frequency (i.e., Δf ± Δf/2). In this manner, neither their amplitude nor their phase information is distorted by frequency-dependent effects, thereby, enabling a more accurate reconstruction of the viscoelastic properties. Assuming a Voigt viscoelastic model, this paper describes the use of a finite-element-method model to simulate three-dimensional (3-D) shear-wave propagation in viscoelastic media containing a spherical inclusion. Nonlinear propagation is assumed for the two ultrasound beams, so that higher harmonics are developed in the force and shear spectrum. Finally, an inverse reconstruction algorithm is used to extract 3-D maps of the local shear modulus and viscosity from the simulated shear-displacement fields based on the fundamental and second-harmonic component. The quality of the reconstructed maps is evaluated using the contrast between the inclusion and the background and the contrast-to-noise ratio (CNR). It is shown that the shear modulus can be accurately reconstructed based on the fundamental component, such that the observed contrast deviates from the true contrast by a root-mean-square-error (RMSE) of only 0.38 and the CNR is greater than 30 dB. If the second-harmonic component is used, the RMSE becomes 1.54 and the corresponding CNR decreases by approximately 10-15 dB. The reconstructed shear viscosity maps based on the second harmonic are shown to be of higher quality than those based on the fundamental. The effects of noise are also investigated and a fusion operation between the two spectral components is applied to enhance the reconstruction quality. Finally, a modified shear-wave spectroscopy technique, shown to be more robust to noise, is described for the estimation of the viscoelastic properties inside and outside the spherical inclusion under conditions of increased noise.

Analysis of the modulated acoustic radiation-force profile for a dual-beam confocal geometry.

A localized modulated radiation force can be produced when two confocal ultrasound beams of nearly equal frequencies interfere in an attenuating medium such as tissue. It is well-established that this force generates both shear and longitudinal waves. By scanning the focal point over a plane and observing the propagation of these waves, the mechanical properties of the medium can be imaged. In this paper, the modulated radiation force is analytically derived in the case of attenuating media, by expanding on the theory of ultrasound-stimulated-vibro-acoustography (USVA) for lossless media. Furthermore, weak nonlinearities are considered in the formulation, since higher source pressures may prove to be necessary to improve the radiation-force profile - only the fundamental component is, however, studied in this paper. An analysis of the generated radiation force is performed and the effects of various parameters are investigated on its amplitude and spatial distribution. It will be shown that by carefully selecting the confocal geometry of the beams, as well as, the source pressure and center frequency, the spatial profile of the radiation force can be optimized. This, subsequently, could improve not only the resolution of the point-spread-function in USVA, but also, the profile of the shear waves in elastography applications.

Estimating the local viscoelastic properties from dispersive shear waves using time-frequency ridge analysis.

Modulated low-frequency shear waves can be non-invasively generated locally within a medium, by the oscillatory acoustic radiation force resulting from the interference of two focused quasi-CW ultrasound beams of slightly different frequencies. The propagation of such shear waves within a viscoelastic medium is known to be affected by the dispersive effects of viscosity. Specifically, a low-frequency (LF) spectral component was shown to arise with increased viscosities and higher modulation frequencies and appear as a 'slow' wave at the end of the shear waveform. In this paper, the shear dispersion characteristics are studied based on the Pseudo-Wigner-Ville distribution (PWVD) in the time-frequency domain. The ridges of the PWVD are then extracted and used to calculate the frequency-dependent shear speed, by identifying the LF dispersive component both in time and frequency. Using numerical simulations, it is shown that this way of estimating the shear dispersion is more efficient and robust than the conventional phase-delay Fourier method. Thus, more accurate estimates of the local shear modulus and viscosity of the propagating medium could be achieved. The effects of noise on the proposed method are also discussed.

A low memory cost model based reconstruction algorithm exploiting translational symmetry for photoacustic microscopy.

A model based reconstruction algorithm that exploits translational symmetries for photoacoustic microscopy to drastically reduce the memory cost is presented. The memory size needed to store the model matrix is independent of the number of acquisitions at different positions. This helps us to overcome one of the main limitations of previous algorithms. Furthermore, using the algebraic reconstruction technique and building the model matrix "on the fly", we have obtained fast reconstructions of simulated and experimental data on both two- and three-dimensional grids using a traditional dark field photoacoustic microscope and a standard personal computer.

Classification-based adaptive filtering for multiframe blind image restoration.

In this paper, the blind restoration of a scene is investigated, when multiple degraded (blurred and noisy) acquisitions are available. An adaptive filtering technique is proposed, where the distorted images are filtered, classified and then fused based upon the classification decisions. Finite normal-density mixture (FNM) models are used to model the filtered outputs at each iteration. For simplicity, fixed number of Gaussian components (classes) is, initially, considered for each degraded frame and the selection of the optimal number of classes is performed according to the global relative entropy criterion. However, there exist cases where dynamically varying FNM models should be considered, where the optimal number of classes is selected according to the Akaike information criterion. The iterative application of classification and fusion, followed by optimal adaptive filtering, converges to a global enhanced representation of the original scene in only a few iterations. The proposed restoration method does not require knowledge of the point-spread-function support size or exact alignment of the acquired frames. Simulation results on synthetic and real data, using both fixed and dynamically varying FNM models, demonstrate its efficiency under both noisy and noise-free conditions.

Mapping the local shear modulus and viscosity using a transient finite-amplitude modulated radiation force.

Localized narrowband low-frequency shear waves can be non-invasively generated within tissue, by a modulated finite-amplitude radiation force, resulting from the interference of two focused quasi-CW ultrasound beams of slightly different frequencies. Assuming a Voigt viscoelastic model, this paper describes the use of a finite-element-method model, to simulate two-dimensional shear-wave propagation in viscoelastic media, containing circular inclusions (lesions). Using this model, an inverse approach is used to extract maps of the local shear modulus and viscosity. The performance is evaluated based on three metrics: the lesion contrast, the contrast-transfer-efficiency (CTE), and the contrast-to-noise ratio (CNR). Modified definitions of these metrics are proposed and used in order to account for the time-varying nature of the shear waves and the inverse reconstruction algorithm. In the absence of any noise, it is shown that accurate reconstruction can be achieved not only with the fundamental, but also with the higher harmonics, as well as, with a low-frequency component that occurs for high viscosity and high modulation frequencies. For low-viscosity conditions, the lesion contrast, CTE, and CNR are shown to exhibit very good performance not only for the fundamental, but also, for the higher harmonics. In the case of increased viscosities and modulation frequencies, the generated low-frequency component is shown to provide superior contrast performance even when compared to that of the fundamental. The effects of noise on the reconstruction quality are examined. Depending on the lesion and background properties, it is shown that noise can seriously degrade reconstruction from the higher harmonics.

Compressed sensing in diffuse optical tomography.

Diffuse optical tomography (DOT) allows tomographic (3D), non-invasive reconstructions of tissue optical properties for biomedical applications. Severe under-sampling is a common problem in DOT which leads to image artifacts. A large number of measurements is needed in order to minimize these artifacts. In this work, we introduce a compressed sensing (CS) framework for DOT which enables improved reconstructions with under-sampled data. The CS framework uses a sparsifying basis, ℓ1-regularization and random sampling to reduce the number of measurements that are needed to achieve a certain accuracy. We demonstrate the utility of the CS framework using numerical simulations. The CS results show improved DOT results in comparison to "traditional" linear reconstruction methods based on singular-value decomposition (SVD) with ℓ2-regularization and with regular and random sampling. Furthermore, CS is shown to be more robust against the reduction of measurements in comparison to the other methods. Potential benefits and shortcomings of the CS approach in the context of DOT are discussed.

Propagation of shear waves generated by a modulated finite amplitude radiation force in a viscoelastic medium.

An effective way to generate localized narrowband low-frequency shear waves within tissue noninvasively, is by the modulated radiation force, resulting from the interference of two confocal quasi-CW ultrasound beams of slightly different frequencies. By using approximate viscoelastic Green's functions, investigations of the properties of the propagated shear-field component at the fundamental modulation frequency were previously reported by our group. However, high-amplitude source excitations may be needed to increase the signal-to-noise-ratio for shear-wave detection in tissue. This paper reports a study of the generation and propagation of dynamic radiation force components at harmonics of the modulation frequency for conditions that generally correspond to diagnostic safety standards. We describe the propagation characteristics of the resulting harmonic shear waves and discuss how they depend on the parameters of nonlinearity, focusing gain, and absorption. For conditions of high viscosity (believed to be characteristic of soft tissue) and higher modulation frequencies, the approximate shear wave Green's function is inappropriate. A more exact viscoelastic Green's function is derived in k-space, and using this, it is shown that the lowpass and dispersive effects, associated with a Voigt model of tissue, are more accurately represented. Finally, it is shown how the viscoelastic properties of the propagating medium can be estimated, based on several spectral components of the shearwave spectrum.

Narrowband shear wave generation by a Finite-Amplitude radiation force: The fundamental component.

A highly localized source of low-frequency shear waves can be created by the modulated radiation force resulting from two intersecting quasi-continuous-wave ultrasound beams of slightly different frequencies. In contrast to most other radiation force-based methods, these shear waves can be narrowband. Consequently, different frequency-dependent effects will not significantly affect their spectrum as they propagate within a viscoelastic medium, thereby enabling the viscoelastic shear properties of the medium to be determined at any given modulation frequency. This can be achieved by tracking the shear wave phase delay and change in amplitude over a specific distance. In this paper we explore the properties of short duration (dynamic) low-frequency shear wave propagation and study how the shear displacement field depends on the excitation conditions. Our investigations make use of the approximate Green's functions for viscoelastic media, and the evolution of such waves is studied in the spatiotemporal domain from a theoretical perspective. Although nonlinearities are included in our confocal source model, just the properties of the fundamental shear component are examined in this paper. We examine how the shear wave propagation is affected by the shear viscosity, the coupling wave, the spatial distribution of the force, the shear speed, and the duration of the modulated wave. A method is proposed for estimating the shear viscosity of a viscoelastic medium. In addition, it is shown how the Voigt model paremeters can be extracted from the frequency-dependent speed and attenuation.