Deep learning applied to dose prediction in external radiation therapy: A narrative review.

Over the last decades, the use of artificial intelligence, machine learning and deep learning in medical fields has skyrocketed. Well known for their results in segmentation, motion management and posttreatment outcome tasks, investigations of machine learning and deep learning models as fast dose calculation or quality assurance tools have been present since 2000. The main motivation for this increasing research and interest in artificial intelligence, machine learning and deep learning is the enhancement of treatment workflows, specifically dosimetry and quality assurance accuracy and time points, which remain important time-consuming aspects of clinical patient management. Since 2014, the evolution of models and architectures for dose calculation has been related to innovations and interest in the theory of information research with pronounced improvements in architecture design. The use of knowledge-based approaches to patient-specific methods has also considerably improved the accuracy of dose predictions. This paper covers the state of all known deep learning architectures and models applied to external radiotherapy with a description of each architecture, followed by a discussion on the performance and future of deep learning predictive models in external radiotherapy.

NMR-spectroscopic and solid-state investigations of cometal-free asymmetric conjugate addition: a dinuclear paracyclophaneimine zinc methyl complex.

We present herein the first indications for dimeric structures in cometal-free asymmetric conjugate addition reactions of dialkylzinc reagents with aldehydes. These are revealed by nonlinear effect (NLE) studies. A monomer-dimer equilibrium can be assumed which explains the increase of the ee value in the product over time. Also, DOSY NMR spectroscopic measurements indicate the existence of the catalyst as [LZnEt](n) complexes in solution. Additionally, the first X-ray structure of a zinc complex with a [2.2]paracyclophane ligand was determined. The structures of the zinc complexes are supported by DFT calculations of monomeric and dimeric species.

[Dosimetric impact of the 2D motion of a platform simulating breathing during a dynamic mode treatment]. / Impact dosimétrique du mouvement 2D d'une plateforme simulant la respiration lors d'un traitement en mode dynamique.

Breathing-adapted techniques in external radiotherapy lead to the improvement of the taken into account of the tumour motion during the patient treatment. Indeed, this motion involves dosimetric uncertainties, in particular during a dynamic treatment (intensity-modulated radiation therapy, dynamic wedge...). As tumoral movement is complex and is carried out in various directions of space, a dynamic platform moving in one or two plans was conceived. This article approaches the technical aspects of design and functioning of this prototype. A study of the dosimetric effects of the respiratory movement on one and two plans during a dynamic treatment without gating will be presented. Films were irradiated while varying the rates with wedged fields at various speeds. The penumbra of beams were compared with the static case and appeared twice broader in the majority of the cases. The results highlighted the contributions of the longitudinal and the axial components of the motion on the form of the dose distribution. These results were completed with gamma index measurements to determine an internal margin. Moreover, this platform proves to be a promising tool for breathing-adapted treatment, in particularly to test the synchronisation of RPM system in fluoroscopic mode in board imaging system.

A new Monte Carlo model of a Cyberknife® system for the precise determination of out-of-field doses.

In a previous work, a PENELOPE Monte Carlo model of a Cyberknife system equipped with fixed collimator was developed and validated for in-field dose evaluation. The aim of this work is to extend it to evaluate peripheral doses and to determine the precision of the treatment planning system (TPS) Multiplan in evaluating the off-axis doses. The Cyberknife® head model was completed with surrounding components based on manufacturer drawings. The contribution of the different head parts on the out-of-field dose was studied. To model the attenuation and the modification of particle energy caused by components not modelled, the photon transport was modified in one of the added components. The model was iteratively adjusted to fit dose profiles measured with EBT3 films and an ionization chamber for several collimator sizes. Finally, dose profiles were calculated using the two Multiplan TPS algorithms and were compared to our simulations. The contributions to out-of-field dose were identified as scattered radiation from the phantom and head leakage and scatter originating at the secondary collimator level. Particle transport in the additional pieces was modified to model this radiation. The maximum differences between simulated and measured doses are of 20.4%. Regarding the detector responses away from axis, EBT3 films and the Farmer chamber give similar response (less than 20% difference). The TPS Monte Carlo algorithm underestimates the doses away from axis more importantly for the smaller field sizes (up to 98%). Besides, RayTracing simplifies peripheral dose to a constant value with no inclusion of particle transport. A Monte Carlo model of a Cyberknife system for the determination of out-of-field doses up to 14 cm off-axis was successfully developed and validated for different depths and field sizes in comparison with measurements. This study also confirms that TPS algorithms do not model peripheral dose properly.

[Dosimetric influence of hip prosthesis during radiotherapeutic treatement]. / Impact dosimétrique des prothèses de hanche lors d'un traitement radiothérapeutique.

As the population become aged, many patients with hip prosthesis are treated for a pelvic cancer. The recommended ballistic must avoid to pass in the prosthesis, but sometimes it is inevitable. So it is essential to quantify with accuracy the dose modifications linked to the presence of metallic implant. The aim of this study is to analyze by Monte Carlo method these modifications in simple and complex models (anthropomorphic phantom) which take into account the geometry and the composition of the prosthesis and its coatings. Then, this methodology was used to study the behaviour of a treatment planning system in theses extreme conditions.

Characterisation of two new radiochromic gel dosimeters TruView™ and ClearView™ in combination with the vista™ optical CT scanner: A feasibility study.

PURPOSE: This study aims at characterising the properties of TruView™ and ClearView™ two new gel dosimeters (Modus Medical Devices Inc.) and at studying the feasibility of relative dosimetry using these dosimeters and the Vista™ Optical CT scanner to accurately evaluate dose. METHODS: In this work, we investigated key dosimetric aspects (dose response, energy and dose rate dependence) and stability of these radiochromic gels initiated in preliminary works (Huet et al., 2017; Colnot et al., 2017) using spectrophotometric measurements. Moreover, by mean of optical CT scanning (Vista™), their performances to measure relative depth dose (PDD) and cross profiles were analysed. RESULTS: TruView™ and ClearView™ present a linear dose response up to 20 Gy and up to 80 Gy respectively, independent of both photon beam energy (4-18 MV) and dose rate (up to 9.9 Gy/min) (Huet et al., 2017; Colnot et al., 2017). ClearView™ response proves to be stable for a week post-irradiation and uniform within the batch whereas TruView™ presents an unstable but uniform response. Optical CT scanning generates errors due to stray light that need to be corrected in order to use these gels; ClearView™ scanning particularly requires important precautions. After corrections, those gels used in combination with the Vista™ scanner show promising spatial and dosimetric precision (dose difference <5%). Finally, TruView™ is reusable and presents excellent reproducible response (maximum 3% difference) and the ClearView™ dosimeter presents good spatial stability (0.5% difference after 6 days). CONCLUSION: This study provides important knowledge about two gel dosimeters presenting interesting dosimetric properties. A study is ongoing to benchmark those promising candidates for clinical dose verification.

Calculations of dose distributions using a neural network model.

The main goal of external beam radiotherapy is the treatment of tumours, while sparing, as much as possible, surrounding healthy tissues. In order to master and optimize the dose distribution within the patient, dosimetric planning has to be carried out. Thus, for determining the most accurate dose distribution during treatment planning, a compromise must be found between the precision and the speed of calculation. Current techniques, using analytic methods, models and databases, are rapid but lack precision. Enhanced precision can be achieved by using calculation codes based, for example, on Monte Carlo methods. However, in spite of all efforts to optimize speed (methods and computer improvements), Monte Carlo based methods remain painfully slow. A newer way to handle all of these problems is to use a new approach in dosimetric calculation by employing neural networks. Neural networks (Wu and Zhu 2000 Phys. Med. Biol. 45 913-22) provide the advantages of those various approaches while avoiding their main inconveniences, i.e., time-consumption calculations. This permits us to obtain quick and accurate results during clinical treatment planning. Currently, results obtained for a single depth-dose calculation using a Monte Carlo based code (such as BEAM (Rogers et al 2003 NRCC Report PIRS-0509(A) rev G)) require hours of computing. By contrast, the practical use of neural networks (Mathieu et al 2003 Proceedings Journees Scientifiques Francophones, SFRP) provides almost instant results and quite low errors (less than 2%) for a two-dimensional dosimetric map.

Identification of (E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate as a major activator for human gammadelta T cells in Escherichia coli.

The gcpE and lytB gene products control the terminal steps of isoprenoid biosynthesis via the 2-C-methyl-D-erythritol 4-phosphate pathway in Escherichia coli. In lytB-deficient mutants, a highly immunogenic compound accumulates significantly, compared to wild-type E. coli, but is apparently absent in gcpE-deficient mutants. Here, this compound was purified from E. coli DeltalytB mutants by preparative anion exchange chromatography, and identified by mass spectrometry, (1)H, (13)C and (31)P NMR spectroscopy, and NOESY analysis as (E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate (HMB-PP). HMB-PP is 10(4) times more potent in activating human Vgamma9/Vdelta2 T cells than isopentenyl pyrophosphate.

Secondary structure of the IIB domain of the Escherichia coli mannose transporter, a new fold in the class of alpha/beta twisted open-sheet structures.

The mannose transporter of the Escherichia coli bacterial phosphotransferase system consists of three subunits: IIAB, IIC and IID. IIABMan transfers phosphoryl groups to the transported substrate via phosphohistidine intermediates. Its IIB domain was overexpressed and isotopically labelled with 13C, 15N and 2H. Heteronuclear 3D triple-resonance NMR experiments combined with a semi-automatic assignment procedure yielded the sequential assignment of the 1H, 13C and 15N backbone resonances. Based on the evaluation of conformationally sensitive parameters, the secondary structure of the IIBMan domain has been determined as an alpha/beta twisted open-sheet structure consisting of a six-stranded parallel beta-sheet with the novel strand order 3-2-4-1-5-6, six helices and a short two-stranded antiparallel beta-sheet.

A new multi-quantum version of the HBHA(CBCACO)NH experiment with enhanced sensitivity for partially deuterated samples.

A new multi-quantum version of the HBHA(CBCACO)NH experiment for partially deuterated protein samples is presented. The method is based on the significant reduction of the proton and carbon relaxation rates due to multi-quantum delays in highly deuterated proteins recently published by our group. The introduction of a multi-quantum period in the coherence transfer pathway of the HBHA(CBCACO)NH experiment yields a dramatic increase of sensitivity-on average 46% with a 75% deuterated sample of the homodimeric 31 kDa E. coli IIAMan domain. Additional resolution in the proton dimension can be achieved by a double time shared approach keeping the 1H single-quantum period at a minimum.

Evaluation of shape descriptors for the morphometric analysis of cell nuclei.

A series of shape descriptors were developed for the morphometrical analysis of cell nuclei. These included five descriptors measuring ellipticity, two measuring concavity, and one measuring the bending energy of a contour. These different shape descriptors were compared using a test sample of 1800 contours of cell nuclei.

Morphological, immunohistochemical, stereological and nuclear shape characteristics of proliferative Leydig cell alterations in rats.

Proliferative Leydig cell (LC) alterations (hyperplasia, adenoma) of laboratory rats often pose diagnostic problems because the progression from normal to hyperplasia to neoplasia is continuous. The LC compartments of 130 Wistar rats (kfm: WIST strain) of approximately 2 years of age were examined. Ten typical cases conventionally classified as being normal or as showing diffuse or focal hyperplasia or small or large adenomata were investigated in more detail. In large adenomata, areas with large and small LC nuclei were identified. Immunohistochemical characterization, EM examination, as well as stereologic and planimetric investigations were performed. Hyperplastic and neoplastic LC essentially retained their normal appearance and immunohistochemical characteristics, but were found to contain more lipid droplets, fibroblast-like cells and patches of collagen than normal LC at the EM level. LC proliferation was accompanied by significant LC hypertrophy. LC nuclei of hyperplastic LC compartments were slightly larger while those of LC adenoma were markedly larger than nuclei of normal LC. The values for circle-related and ellipticity factors indicated that the nuclei of normal and hyperplastic LC were more markedly oval than nuclei of neoplastic LC. Concavity factor and bending energy measurements revealed that the small and oval nuclei of normal and hyperplastic LC had significantly more and deeper indentations than the larger and somewhat rounder nuclei of neoplastic LC. It is concluded that LC proliferations conventionally diagnosed as hyperplasia or adenoma on the basis of their size were composed of cytologically different LC populations.

Study of dental prostheses influence in radiation therapy.

Dental prostheses made of high density material contribute to modify dose distribution in head and neck cancer treatment. Our objective is to quantify dose perturbation due to high density inhomogeneity with experimental measurements and Monte Carlo simulations. Firstly, measurements were carried in a phantom representing a human jaw with thermoluminescent detectors (GR200A) and EBT2 Gafchromic films in the vicinity of three samples: a healthy tooth, a tooth with amalgam and a Ni-Cr crown, irradiated in clinical configuration. Secondly, Monte Carlo simulations (BEAMnrc code) were assessed in an identical configuration. Experimental measurements and simulation results confirm the two well-known phenomena: firstly the passage from a low density medium to a high density medium induces backscattered electrons causing a dose increase at the interface, and secondly, the passage from a high density medium to a low density medium creates a dose decrease near the interface. So, the results show a 1.4% and 23.8% backscatter dose rise and attenuation after sample of 26.7% and 10.9% respectively for tooth with amalgam and crown compared to the healthy tooth. Although a tooth with amalgam has a density of about 12-13, the changes generated are not significant. However, the results for crown (density of 8) are very significant and the discordance observed may be due to calculation point size difference 0.8 mm and 0.25 mm respectively for TLD and Monte Carlo. The use of Monte Carlo simulations and experimental measurements provides objective evidence to evaluate treatment planning system results with metal dental prostheses.

LED based NMR illumination device for mechanistic studies on photochemical reactions--versatile and simple, yet surprisingly powerful.

An LED based illumination device for mechanistic studies on photochemical reactions by means of NMR spectroscopy is presented. The LEDs are directly switched by the NMR spectrometer with the help of a one-stage electronic circuit. This allows for continuous or alternatively pulsed operation of the LEDs. Continuous operation provides direct comparability with conditions in synthetic chemistry, in pulsed operation the short time light power can be enhanced ninefold. The LEDs are efficiently coupled to a 1000 µm core optical fiber guiding the light into the spectrometer by simply bringing it in close contact to the fiber. The tip of the fiber is roughened by sandblasting and thus emits light in a uniform and efficient way over the full length of the receiver coil. The combination of these techniques tremendously increases the amount of light brought into the NMR sample and makes LEDs an easy, versatile and handy light source for the in situ illumination of NMR samples allowing even for single millisecond time resolved Photo-CIDNP spectroscopy.

[Simulation of lung motions using an artificial neural network]. / Utilisation d'un réseau de neurones artificiels pour la simulation des mouvements pulmonaires.

PURPOSE: A way to improve the accuracy of lung radiotherapy for a patient is to get a better understanding of its lung motion. Indeed, thanks to this knowledge it becomes possible to follow the displacements of the clinical target volume (CTV) induced by the lung breathing. This paper presents a feasibility study of an original method to simulate the positions of points in patient's lung at all breathing phases. PATIENTS AND METHODS: This method, based on an artificial neural network, allowed learning the lung motion on real cases and then to simulate it for new patients for which only the beginning and the end breathing data are known. The neural network learning set is made up of more than 600 points. These points, shared out on three patients and gathered on a specific lung area, were plotted by a MD. RESULTS: The first results are promising: an average accuracy of 1mm is obtained for a spatial resolution of 1 × 1 × 2.5mm(3). CONCLUSION: We have demonstrated that it is possible to simulate lung motion with accuracy using an artificial neural network. As future work we plan to improve the accuracy of our method with the addition of new patient data and a coverage of the whole lungs.

Fluctuations of light absorption of healthy papillae repeatedly photographed over a long period of time.

In order to distinguish normal, age-related changes of the papilla from changes due to glaucoma, we investigated 7 normal papillae that had been photographed several times over a period of 11 years. By means of digital-image processing techniques, the red- and green-filtered, digitized pictures were compared directly using a subtraction method. The results seem to indicate that normal papillae do not blanch over a long period of time. The difficulties of comparison are discussed.

A Spin System Labeled and Highly Resolved ed-H(CCO)NH-TOCSY Experiment for the Facilitated Assignment of Proton Side Chains in Partially Deuterated Samples.

The introduction of deuterated and partially deuterated protein samples has greatly facilitated the 13C assignment of larger proteins. Here we present a new version of the HC(CO)NH-TOCSY experiment, the ed-H(CCO)NH-TOCSY experiment for partially deuterated samples, introducing a multi-quantum proton evolution period. This approach removes the main relaxation source (the dipolar coupling to the directly bound 13C spin) and leads to a significant reduction of the proton and carbon relaxation rates. Thus, the indirect proton dimension can be acquired with high resolution, combined with a phase labeling of the proton resonances according to the C-C spin system topology. This editing scheme, independent of the CHn multiplicity, allows to distinguish between proton side-chain positions occurring within a narrow chemical shift range. Therefore this new experiment facilitates the assignment of the proton chemical shifts of partially deuterated samples even of high molecular weights, as demonstrated on a 31 kDa protein.

Some fundamental aspects of morphometry in clinical pathology, demonstrated on a simple, multipurpose analysis system.

The aim of this study was (1) to investigate the value of morphometry, (2) to fix a set of parameters suitable for analyzing diagnostic problems, and (3) to create a general strategy for data storage and for user-friendly data management. The intrinsic value of morphometry lies in the fact that in contrast to other morphologic methods, it permits the presentation of findings in the form of numbers. The following set of morphometric parameters, in the broad sense of the term morphometry, is standard in our laboratory: planimetric parameters (shape descriptors), parameters of the gray value histogram (descriptors of the general gray value distribution), texture parameters (descriptors of the correlation between various image segments), invariant moments (descriptors of the size and localization of textural image segments) and densitometric parameters. The introduction of morphometric procedures into the daily routine is facilitated if data registration and evaluation are performed separately. Original data generated by direct measurement are primary or raw data, which are stored as such. In a separate, second step these raw data are used to compare more or less complex morphometric parameters, which are called "secondary data". A system designed for separate data registration and evaluation can easily be adapted to new methodologic developments. For instance, primary data on objects (gray values, coordinates of the contour) measured one time in the past can be reused at any other time for computing new features from these data. This procedure is comparable to the possibilities in immunohistochemical staining: new immunohistochemical stains can be applied to newly prepared sections of old tissue blocks.

[Morphology and function of macrophages in AIDS in broncho-alveolar lavages (BAL)]. / Beobachtungen zur Morphologie und Funktion der Makrophagen bei AIDS an broncho-alveolären Lavagen (BAL).

HIV infects susceptible T-cell and mononuclear phagocyte targets. Unlike the dramatic changes that occur with T-cells during HIV disease, changes in monocyte and macrophage phenotype and function are qualitatively minimal. The aim of the study was to quantitatively analyze the morphology of macrophages in broncho-alveolar lavages in patients without AIDS, with AIDS, with AIDS complicated by Pneumocystis carinii and in carcinoma patients. In patients with AIDS, the nuclei of macrophages demonstrate a higher integrated optical density (2P less than 0.10), are bigger in size (2P less than 0.05) and have less notches (2P less than 0.02).

Me(2)CuLi*LiCN in diethyl ether prefers a homodimeric core structure [Me(2)CuLi](2) and not a heterodimeric one [Me(2)CuLi*LiCN]: (1)H, (6)Li HOE and (1)H, (1)H NOE studies by NMR.

H-Li distances and (1)H-(1)H dipolar interactions in Me(2)CuLiLiCN and Me(2)CuLi in diethyl ether (Et(2)O), obtained by NMR spectroscopy, were used to gain structural information about the contact ion pair of the salt-containing organocuprate Me(2)CuLiLiCN in this solvent. The H-Li distances of Me(2)CuLiLiCN and Me(2)CuLi in Et(2)O, resulting from the initial buildup rates in conjunction with the motional correlation times, are almost identical, indicating a similar homodimeric core structure [Me(2)CuLi](2) for both samples. However, the H-Li distances obtained for Me(2)CuLiLiCN do not rigorously exclude a heterodimeric structure [Me(2)CuLiLiCN] as proposed by ab initio calculations. Therefore, (1)H-(1)H dipolar interactions were investigated by SYM-BREAK-NOE/ROE-HSQC experiments, which allow for the observation of NOEs between equivalent protons. Since these experiments showed similar (1)H-(1)H dipolar interactions of Me(2)CuLiLiCN and Me(2)CuLi, we propose that for Me(2)CuLiLiCN a homodimeric core structure [Me(2)CuLi](2) indeed is predominant in Et(2)O.