Synthetic data set
Dotaz
Zobrazit nápovědu
OBJECTIVES: Class imbalance in datasets is one of the challenges of machine learning (ML) in medical image analysis. We employed synthetic data to overcome class imbalance when segmenting bitewing radiographs as an exemplary task for using ML. METHODS: After segmenting bitewings into classes, i.e. dental structures, restorations, and background, the pixel-level representation of implants in the training set (1543 bitewings) and testing set (177 bitewings) was 0.03 % and 0.07 %, respectively. A diffusion model and a generative adversarial network (pix2pix) were used to generate a dataset synthetically enriched in implants. A U-Net segmentation model was trained on (1) the original dataset, (2) the synthetic dataset, (3) on the synthetic dataset and fine-tuned on the original dataset, or (4) on a dataset which was naïvely oversampled with images containing implants. RESULTS: U-Net trained on the original dataset was unable to segment implants in the testing set. Model performance was significantly improved by naïve over-sampling, achieving the highest precision. The model trained only on synthetic data performed worse than naïve over-sampling in all metrics, but with fine-tuning on original data, it resulted in the highest Dice score, recall, F1 score and ROC AUC, respectively. The performance on other classes than implants was similar for all strategies except training only on synthetic data, which tended to perform worse. CONCLUSIONS: The use of synthetic data alone may deteriorate the performance of segmentation models. However, fine-tuning on original data could significantly enhance model performance, especially for heavily underrepresented classes. CLINICAL SIGNIFICANCE: This study explored the use of synthetic data to enhance segmentation of bitewing radiographs, focusing on underrepresented classes like implants. Pre-training on synthetic data followed by fine-tuning on original data yielded the best results, highlighting the potential of synthetic data to advance AI-driven dental imaging and ultimately support clinical decision-making.
- Klíčová slova
- Artificial intelligence, Dataset imbalance, Dentistry, Diffusion model, Generative adversarial network, Synthetic medical data,
- MeSH
- lidé MeSH
- počítačové zpracování obrazu * metody MeSH
- strojové učení * MeSH
- zubní implantáty MeSH
- Check Tag
- lidé MeSH
- Publikační typ
- časopisecké články MeSH
- Názvy látek
- zubní implantáty MeSH
AFM microscopy from its nature produces outputs with certain distortions, inaccuracies and errors given by its physical principle. These distortions are more or less well studied and documented. Based on the nature of the individual distortions, different reconstruction and compensation filters have been developed to post-process the scanned images. This article presents an approach based on machine learning - the involved convolutional neural network learns from pairs of distorted images and the ground truth image and then it is able to process pairs of images of interest and produce a filtered image with the artifacts removed or at least suppressed. What is important in our approach is that the neural network is trained purely on synthetic data generated by a simulator of the inputs, based on an analytical description of the physical phenomena causing the distortions. The generator produces training samples involving various combinations of the distortions. The resulting trained network seems to be able to autonomously recognize the distortions present in the testing image (no knowledge of the distortions or any other human knowledge is provided at the test time) and apply the appropriate corrections. The experimental results show that not only is the new approach better or at least on par with conventional post-processing methods, but more importantly, it does not require any operator's input and works completely autonomously. The source codes of the training set generator and of the convolutional neural net model are made public, as well as an evaluation dataset of real captured AFM images.
BACKGROUND: In European axial spondyloarthritis (axSpA) and psoriatic arthritis (PsA) clinical registries, we aimed to investigate commonalities and differences in (1) set-up, clinical data collection; (2) data availability and completeness; and (3) wording, recall period, and scale used for selected patient-reported outcome measures (PROMs). METHODS: Data was obtained as part of the EuroSpA Research Collaboration Network and consisted of (1) an online survey and follow-up interview, (2) upload of real-world data, and (3) selected PROMs included in the online survey. RESULTS: Fifteen registries participated, contributing 33,948 patients (axSpA: 21,330 (63%), PsA: 12,618 (37%)). The reported coverage of eligible patients ranged from 0.5 to 100%. Information on age, sex, biological/targeted synthetic disease-modifying anti-rheumatic drug treatment, disease duration, and C-reactive protein was available in all registries with data completeness between 85% and 100%. All PROMs (Bath Ankylosing Spondylitis Disease Activity and Functional Indices, Health Assessment Questionnaire, and patient global, pain and fatigue assessments) were more complete after 2015 (68-86%) compared to prior (50-79%). Patient global, pain and fatigue assessments showed heterogeneity between registries in terms of wording, recall periods, and scale. CONCLUSION: Important heterogeneity in registry design and data collection across fifteen European axSpA and PsA registries was observed. Several core measures were widely available, and an increase in data completeness of PROMs in recent years was identified. This study might serve as a basis for examining how differences in data collection across registries may impact the results of collaborative research in the future.
- Klíčová slova
- Clinical data collection, Collaborative research, European registries, Real-world evidence, Spondyloarthritis,
- MeSH
- ankylózující spondylitida * farmakoterapie MeSH
- bolest MeSH
- lidé MeSH
- psoriatická artritida * farmakoterapie epidemiologie MeSH
- registrace MeSH
- spondylartritida * farmakoterapie epidemiologie MeSH
- Check Tag
- lidé MeSH
- Publikační typ
- časopisecké články MeSH
Correct virtual reconstruction of a defective skull is a prerequisite for successful cranioplasty and its automatization has the potential for accelerating and standardizing the clinical workflow. This work provides a deep learning-based method for the reconstruction of a skull shape and cranial implant design on clinical data of patients indicated for cranioplasty. The method is based on a cascade of multi-branch volumetric CNNs that enables simultaneous training on two different types of cranioplasty ground-truth data: the skull patch, which represents the exact shape of the missing part of the original skull, and which can be easily created artificially from healthy skulls, and expert-designed cranial implant shapes that are much harder to acquire. The proposed method reaches an average surface distance of the reconstructed skull patches of 0.67 mm on a clinical test set of 75 defective skulls. It also achieves a 12% reduction of a newly proposed defect border Gaussian curvature error metric, compared to a baseline model trained on synthetic data only. Additionally, it produces directly 3D printable cranial implant shapes with a Dice coefficient 0.88 and a surface error of 0.65 mm. The outputs of the proposed skull reconstruction method reach good quality and can be considered for use in semi- or fully automatic clinical cranial implant design workflows.
- Klíčová slova
- 3D convolutional neural networks, Cranial implant design, Cranioplasty, Skull reconstruction,
- MeSH
- deep learning * MeSH
- lebka diagnostické zobrazování chirurgie MeSH
- lidé MeSH
- protézy a implantáty MeSH
- zákroky plastické chirurgie * MeSH
- Check Tag
- lidé MeSH
- Publikační typ
- časopisecké články MeSH
Interpretation of QSAR models is useful to understand the complex nature of biological or physicochemical processes, guide structural optimization or perform knowledge-based validation of QSAR models. Highly predictive models are usually complex and their interpretation is non-trivial. This is particularly true for modern neural networks. Various approaches to interpretation of these models exist. However, it is difficult to evaluate and compare performance and applicability of these ever-emerging methods. Herein, we developed several benchmark data sets with end-points determined by pre-defined patterns. These data sets are purposed for evaluation of the ability of interpretation approaches to retrieve these patterns. They represent tasks with different complexity levels: from simple atom-based additive properties to pharmacophore hypothesis. We proposed several quantitative metrics of interpretation performance. Applicability of benchmarks and metrics was demonstrated on a set of conventional models and end-to-end graph convolutional neural networks, interpreted by the previously suggested universal ML-agnostic approach for structural interpretation. We anticipate these benchmarks to be useful in evaluation of new interpretation approaches and investigation of decision making of complex "black box" models.
- Klíčová slova
- Atom contributions, Benchmark data set, Graph convolutional neural networks, Interpretability metrics, QSAR model interpretation, Synthetic data set,
- Publikační typ
- časopisecké články MeSH
OBJECTIVE: LDL-cholesterol (LDL-C) is determined by methods whose accuracy is significantly affected in various clinical or analytical situations. Two computational methods have recently been described, the Martin equation and the Sampson equation, validity of which we compare with the Friedewald equation. METHODS: LDL-C comparisons determined by the 3 equations were performed on 4 real sets of lipid data, generated in various previous studies, ranging from n = 140 to n = 7 393. We have created an artificial set of data on the extent of 900 members with equally distributed values of TC, HDL-C and TG troughout the commonly found range. Such a data set is independent of the phrase "we performed the calculations on our file". Comparisons were also made on this artificial file. RESULTS: The difference between the LDL-C values determined by the different equations gradually increases with decreasing LDL-C levels, both in the subgroup of low TG values and in the subgroups of medium and higher TG values. This applies to all 4 real files as well as to the artificial file. These differences are more visible the larger the file size. For the artificial set, the overall agreement between the LDL-C categories was lowest when comparing the Friedewald and Martin equations (83.1%), higher between the Sampson and Martin equations (88.9%) and highest when comparing the Friedewald and Sampson equations (90.9%). In all 4 real sets, the trends of overestimation and underestimation between the equations were exactly the same as in the artificial set. CONCLUSION: The results of clinical and epidemiological studies are significantly influenced by the method used to determine LDL-C. When comparing the calculation methods for determining LDL-C, it is possible to preferably use the described artificial set.
- Klíčová slova
- Friedewald equation, LDL cholesterol, Martin/Hopkins equation, Sampson equation, hypertension, method comparison,
- MeSH
- LDL-cholesterol * MeSH
- lidé MeSH
- triglyceridy MeSH
- Check Tag
- lidé MeSH
- Publikační typ
- časopisecké články MeSH
- Názvy látek
- LDL-cholesterol * MeSH
- triglyceridy MeSH
The toxicity of food additives is widely studied and concerns many consumers worldwide. Synthetic food colors are often considered an unnecessary risk to consumer health. Since the European Food Safety Authority's (EFSA) re-evaluation between 2009 and 2014, the body of scientific literature on food colors has grown, and new evaluations are being published by the Joint FAO/WHO Expert Committee on Food Additives (JECFA). Therefore, this narrative review aims to review the toxicological data that have become available since 2014. The reviewed colors are Quinoline Yellow, Sunset Yellow, Azorubine, Amaranth, Ponceau 4R, Erythrosine, Allura Red, Patent Blue, Indigo Carmine, Brilliant Blue FCF, Green S, Brilliant Black, Brown HT, and Lithol Rubine BK. Tartrazine was not included in this paper; the overwhelming amount of recent data on Tartrazine toxicity requires more space than this review can provide. The issues regarding the toxicity of synthetic food colors and real population exposures are being regularly examined and reviewed by relevant authorities, such as the EFSA and JECFA. The current ADI limits set by the authorities are mostly in agreement, and they seem safe. However, the EFSA and JECFA assessments of some of the colors are more than a decade old, and new evidence will soon be required.
Xanthene dyes can be appended to cyclodextrins via an ester or amide bridge in order to switch the fluorescence on or off. This is made possible through the formation of nonfluorescent lactones or lactams as the fluorophore can reversibly cyclize. In this context we report a green approach for the synthesis of switchable xanthene-dye-appended cyclodextrins based on the coupling agent 4-(4,6-dimethoxy-1,3,5-triazin-2-yl)-4-methylmorpholinium chloride (DMT-MM). By using 6-monoamino-β-cyclodextrin and commercially available inexpensive dyes, we prepared rhodamine- and fluorescein-appended cyclodextrins. The compounds were characterized by NMR and IR spectroscopy and MS spectrometry, their UV-vis spectra were recorded at various pH, and their purity was determined by capillary electrophoresis. Two potential models for the supramolecular assembly of the xanthene-dye-appended cyclodextrins were developed based on the set of data collected by the extensive NMR characterization.
- Klíčová slova
- DMT-MM, fluorescein, rhodamine, supramolecular assembly,
- Publikační typ
- časopisecké články MeSH
Sequential Injection Chromatography (SIC) evolved from fast and automated non-separation Sequential Injection Analysis (SIA) into chromatographic separation method for multi-element analysis. However, the speed of the measurement (sample throughput) is due to chromatography significantly reduced. In this paper, a sub-1min separation using medium polar cyano monolithic column (5mm×4.6mm) resulted in fast and green separation with sample throughput comparable with non-separation flow methods The separation of three synthetic water-soluble dyes (sunset yellow FCF, carmoisine and green S) was in a gradient elution mode (0.02% ammonium acetate, pH 6.7 - water) with flow rate of 3.0mLmin-1 corresponding with sample throughput of 30h-1. Spectrophotometric detection wavelengths were set to 480, 516 and 630nm and 10Hz data collection rate. The performance of the separation was described and discussed (peak capacities 3.48-7.67, peak symmetries 1.72-1.84 and resolutions 1.42-1.88). The method was represented by validation parameters: LODs of 0.15-0.35mgL-1, LOQs of 0.50-1.25mgL-1, calibration ranges 0.50-150.00mgL-1 (r>0.998) and repeatability at 10.0mgL-1 of RSD≤0.98% (n=6). The method was used for determination of the dyes in "forest berries" colored pharmaceutical cough-cold formulation. The sample matrix - pharmaceuticals and excipients were not interfering with vis determination because of no retention in the separation column and colorless nature. The results proved the concept of fast and green chromatography approach using very short medium polar monolithic column in SIC.
- Klíčová slova
- Cyano monolithic column, Fast separation, Green chromatography, No-solvent mobile phase, Sequential injection chromatography, Water-soluble food dyes,
- MeSH
- barvicí látky MeSH
- chromatografie MeSH
- farmaceutická chemie metody MeSH
- příprava léků MeSH
- voda MeSH
- Publikační typ
- časopisecké články MeSH
- Názvy látek
- barvicí látky MeSH
- voda MeSH
Cytochrome P4501A activity, oxidative stress and inhibition of gap junctional intercellular communication (GJIC) are involved in metabolic activation of promutagens and tumor-promoting activity of various xenobiotics, and their prevention is considered to be an important characteristic of chemoprotective compounds. In this study, a series of 31 chalcones and their corresponding dihydroderivatives, substituted in 2,2'-, 3,3'-, 4- or 4'-position by hydroxyl or methoxy group, were tested for their ability to inhibit Fe(II)/NADPH-enhanced lipid peroxidation and cytochrome P4501A-dependent 7-cethoxyresorufin-O-deethylase (EROD) activity in rat hepatic microsomes. Effects of the compounds on GJIC were determined in rat liver epithelial WB-F344 cells. Most of the chalcones and dihydrochalcones inhibited EROD activity in a dose-dependent manner at the range 0.25-25 microM, which was comparable to model flavonoid inhibitors alpha-naphthoflavone and quercetin. The chalcones exhibited higher inhibition activity than the corresponding dihydroderivatives. Mono and dihydroxylated chalcones, and dihydrochalcones showed none or only a weak antioxidant activity; trihydroxyderivatives inhibited in vitro lipid peroxidation significantly only at 50 microM concentration. Potential adverse effects, namely inhibition of GJIC and/or cytotoxicity were detected after treatment of WB-F344 cells with a number of chalcone and dihydrochalcone derivatives, suggesting that they should be excluded from additional screening as chemoprotective compounds. Chalcones and dihydrochalcones substituted at 4- and/or 4'-position, which elicited no inhibition of GJIC, were further tested for the potential enhancing effects on GJIC. The present data seem to suggest that 4-hydroxy, 2',4'-dihydroxy-3-methoxy, 2,4,4'-trihydroxy, and 2',4,4'-trihydroxychalcone, 2',4-dihydroxy and 2'-hydroxy-3,4-dimethoxydihydrochalcone might be promising chemoprotective compounds against CYP1A activity, and partly also against oxidative damage without inducing adverse effects, such as GJIC inhibition. In general, determination of potencies of tested compounds to inhibit GJIC should be involved in any set of methods for the in vitro screening of chemoprotective characteristics of potential drugs, in order to reveal their potential adverse effects associated with tumor promotion.
- MeSH
- buněčné linie MeSH
- chalkonoidy farmakologie toxicita MeSH
- epitelové buňky účinky léků metabolismus MeSH
- inhibitory cytochromu P450 MeSH
- jaterní mikrozomy účinky léků enzymologie MeSH
- játra účinky léků ultrastruktura MeSH
- karcinogeny metabolismus toxicita MeSH
- krysa rodu Rattus MeSH
- mezerový spoj účinky léků metabolismus fyziologie MeSH
- mezibuněčná komunikace účinky léků fyziologie MeSH
- peroxidace lipidů účinky léků MeSH
- potkani Wistar MeSH
- systém (enzymů) cytochromů P-450 metabolismus MeSH
- techniky in vitro MeSH
- vztah mezi dávkou a účinkem léčiva MeSH
- vztahy mezi strukturou a aktivitou MeSH
- zvířata MeSH
- Check Tag
- krysa rodu Rattus MeSH
- mužské pohlaví MeSH
- zvířata MeSH
- Publikační typ
- časopisecké články MeSH
- práce podpořená grantem MeSH
- srovnávací studie MeSH
- Názvy látek
- chalkonoidy MeSH
- inhibitory cytochromu P450 MeSH
- karcinogeny MeSH
- systém (enzymů) cytochromů P-450 MeSH
