Isodoses-a set theory-based patient-specific QA measure to compare planned and delivered isodose distributions in photon radiotherapy.

BACKGROUND: The gamma index and dose-volume histogram (DVH)-based patient-specific quality assurance (QA) measures commonly applied in radiotherapy planning are unable to simultaneously deliver detailed locations and magnitudes of discrepancy between isodoses of planned and delivered dose distributions. By exploiting statistical classification performance measures such as sensitivity or specificity, compliance between a planned and delivered isodose may be evaluated locally, both for organs-at-risk (OAR) and the planning target volume (PTV), at any specified isodose level. Thus, a patient-specific QA tool may be developed to supplement those presently available in clinical radiotherapy. MATERIALS AND METHODS: A method was developed to locally establish and report dose delivery errors in three-dimensional (3D) isodoses of planned (reference) and delivered (evaluated) dose distributions simultaneously as a function the dose level and of spatial location. At any given isodose level, the total volume of delivered dose containing the reference and the evaluated isodoses is locally decomposed into four subregions: true positive-subregions within both reference and evaluated isodoses, true negative-outside of both of these isodoses, false positive-inside the evaluated isodose but not the reference isodose, and false negatives-inside the reference isodose but not the evaluated isodose. Such subregions may be established over the whole volume of delivered dose. This decomposition allows the construction of a confusion matrix and calculation of various indices to quantify the discrepancies between the selected planned and delivered isodose distributions, over the complete range of values of dose delivered. The 3D projection and visualization of the spatial distribution of these discrepancies facilitates the application of the developed method in clinical practice. RESULTS: Several clinical photon radiotherapy plans were analyzed using the developed method. In some plans at certain isodose levels, dose delivery errors were found at anatomically significant locations. These errors were not otherwise highlighted-neither by gamma analysis nor by DVH-based QA measures. A specially developed 3D projection tool to visualize the spatial distribution of such errors against anatomical features of the patient aids in the proposed analysis of therapy plans. CONCLUSIONS: The proposed method is able to spatially locate delivery errors at selected isodose levels and may supplement the presently applied gamma analysis and DVH-based QA measures in patient-specific radiotherapy planning.

Zernike Moment Based Classification of Cosmic Ray Candidate Hits from CMOS Sensors.

Reliable tools for artefact rejection and signal classification are a must for cosmic ray detection experiments based on CMOS technology. In this paper, we analyse the fitness of several feature-based statistical classifiers for the classification of particle candidate hits in four categories: spots, tracks, worms and artefacts. We use Zernike moments of the image function as feature carriers and propose a preprocessing and denoising scheme to make the feature extraction more efficient. As opposed to convolution neural network classifiers, the feature-based classifiers allow for establishing a connection between features and geometrical properties of candidate hits. Apart from basic classifiers we also consider their ensemble extensions and find these extensions generally better performing than basic versions, with an average recognition accuracy of 88%.

Are gamma passing rate and dose-volume histogram QA metrics correlated?

PURPOSE: The quality of a measured distribution of dose delivered against its corresponding radiotherapy plan is routinely assessed by gamma index (GI) and dose-volume histogram (DVH) metrics. Any correlation between error detection rates, as based on either of these approaches, while argued, has never been convincingly demonstrated. The dependence of the strength of correlation between the GI passing rate ( γ P ) and DVH quality assurance (QA) metrics on various elements of the therapy plan has not been systematically investigated. METHODS: A formal analysis of the relation between γ P and DVH metrics has been undertaken, leading to a relationship which may partly approximate γ P with respect to the DVH. This relationship was further validated by studying examples of simulated clinical radiotherapy plans and by studying the correlation between γ P and the derived relationship using a simple two-dimensional representations of the planning target volume (PTV) and organs at risk (OAR), where penumbra regions, distance-to-agreement tolerances and dose delivery errors were systematically varied. RESULTS: It is shown formally that there cannot be any correlation between γ P and other commonly applied DVH-derived QA measures. However, γ P may be partly approximated given the planned and measured DVH. The derived γ P approximation (the " γ -slope indicator") may be clinically useful in some practical cases of radiotherapy plan QA. CONCLUSIONS: In formal terms, there cannot be any correlation between γ P and any common DVH-calculated patient-specific measures, with respect to PTV or OAR. However, as demonstrated analytically and further confirmed in our simulation studies, the γ P approximation derived in this study (the " γ -slope indicator") may in some cases offer a degree of correlation between γ P and the PTV and OAR DVH QA metrics in measured and planned patient-specific dose distributions-which may be potentially useful in clinical practice.

CNN-Based Classifier as an Offline Trigger for the CREDO Experiment.

Gamification is known to enhance users' participation in education and research projects that follow the citizen science paradigm. The Cosmic Ray Extremely Distributed Observatory (CREDO) experiment is designed for the large-scale study of various radiation forms that continuously reach the Earth from space, collectively known as cosmic rays. The CREDO Detector app relies on a network of involved users and is now working worldwide across phones and other CMOS sensor-equipped devices. To broaden the user base and activate current users, CREDO extensively uses the gamification solutions like the periodical Particle Hunters Competition. However, the adverse effect of gamification is that the number of artefacts, i.e., signals unrelated to cosmic ray detection or openly related to cheating, substantially increases. To tag the artefacts appearing in the CREDO database we propose the method based on machine learning. The approach involves training the Convolutional Neural Network (CNN) to recognise the morphological difference between signals and artefacts. As a result we obtain the CNN-based trigger which is able to mimic the signal vs. artefact assignments of human annotators as closely as possible. To enhance the method, the input image signal is adaptively thresholded and then transformed using Daubechies wavelets. In this exploratory study, we use wavelet transforms to amplify distinctive image features. As a result, we obtain a very good recognition ratio of almost 99% for both signal and artefacts. The proposed solution allows eliminating the manual supervision of the competition process.

Classification Algorithm for Person Identification and Gesture Recognition Based on Hand Gestures with Small Training Sets.

Classification algorithms require training data initially labelled by classes to build a model and then to be able to classify the new data. The amount and diversity of training data affect the classification quality and usually the larger the training set, the better the accuracy of classification. In many applications only small amounts of training data are available. This article presents a new time series classification algorithm for problems with small training sets. The algorithm was tested on hand gesture recordings in tasks of person identification and gesture recognition. The algorithm provides significantly better classification accuracy than other machine learning algorithms. For 22 different hand gestures performed by 10 people and the training set size equal to 5 gesture execution records per class, the error rate for the newly proposed algorithm is from 37% to 75% lower than for the other compared algorithms. When the training set consists of only one sample per class the new algorithm reaches from 45% to 95% lower error rate. Conducted experiments indicate that the algorithm outperforms state-of-the-art methods in terms of classification accuracy in the problem of person identification and gesture recognition.

Statistical approach to the selection of the tolerances for distance to agreement improves the quality control of the dose delivery in radiotherapy.

In the study, a local approach to setting reference tolerance values for the distance-to-agreement (DTA) component of the gamma index is proposed. The reference tolerance values are calculated in simulations, following a dose delivery model presented in a previous work. An analytical model for determining the quantiles of DTA distribution is also proposed and verified. It is shown that the distributions of DTA values normalized with either quantiles or standard deviation of DTA distributions are universal over analyzed plans and points within a single plan. This enables statistically sound inference about the quality of dose delivery. In particular, based on the normalized distributions the comparison of planned and delivered doses can be formulated within the framework of statistical inference as a problem of multiple statistical testing. For every evaluated point P of a plan, one may formulate and test a null hypothesis that there is no delivery error against an alternative hypothesis that there is a delivery error in P. It is also shown that the proposed approach is more sensitive than the current standard approach to shift errors in high dose gradient regions.

A simulation-based method for evaluating geometric tests of a linac c-arm in quality control in radiotherapy.

PURPOSE: Assessment of the accuracy of geometric tests of a linac used in external beam therapy is crucial for ensuring precise dose delivery. In this paper, a new simulation-based method for assessing accuracy of such geometric tests is proposed and evaluated on a set of testing procedures. METHODS: Linac geometry testing methods used in this study are based on an established design of a two-module phantom. Electronic portal imaging device (EPID) images of fiducial balls contained in these modules can be used to automatically reconstruct linac geometry. The projection of the phantom modules fiducial balls onto the EPID detector plane is simulated for assumed nominal geometry of a linac. Then, random errors are added to the coordinates of the projections of the centers of the fiducial balls and the linac geometry is reconstructed from these data. RESULTS: Reconstruction is performed for a set of geometric test designs and it is shown how the dispersion of the reconstructed values of geometric parameters depends on the design of a geometric test. Assuming realistic accuracy of EPID image analysis, it is shown that for selected testing plans the reconstruction accuracy of geometric parameters can be significantly better than commonly used action thresholds for these parameters. CONCLUSIONS: Proposed solution has the potential to improve geometric testing design and practice. It is an important part of a fully automated geometric testing solution.

Application of Computational Intelligence Methods for the Automated Identification of Paper-Ink Samples Based on LIBS.

Laser-induced breakdown spectroscopy (LIBS) is an important analysis technique with applications in many industrial branches and fields of scientific research. Nowadays, the advantages of LIBS are impaired by the main drawback in the interpretation of obtained spectra and identification of observed spectral lines. This procedure is highly time-consuming since it is essentially based on the comparison of lines present in the spectrum with the literature database. This paper proposes the use of various computational intelligence methods to develop a reliable and fast classification of quasi-destructively acquired LIBS spectra into a set of predefined classes. We focus on a specific problem of classification of paper-ink samples into 30 separate, predefined classes. For each of 30 classes (10 pens of each of 5 ink types combined with 10 sheets of 5 paper types plus empty pages), 100 LIBS spectra are collected. Four variants of preprocessing, seven classifiers (decision trees, random forest, k-nearest neighbor, support vector machine, probabilistic neural network, multi-layer perceptron, and generalized regression neural network), 5-fold stratified cross-validation, and a test on an independent set (for methods evaluation) scenarios are employed. Our developed system yielded an accuracy of 99.08%, obtained using the random forest classifier. Our results clearly demonstrates that machine learning methods can be used to identify the paper-ink samples based on LIBS reliably at a faster rate.