Automated Data Interpretation Research Articles

MALDI-MS in first-line screening of newborns for sickle cell disease: results from a prospective study in comparison to HPLC.

Newborn screening (NBS) for sickle cell disease (SCD) requires a robust, high-throughput method to detect hemoglobin S (HbS). Screening for SCD is performed by qualitative methods, such as isoelectric focusing (IEF), and both qualitative and quantitative methods such as high performance liquid chromatography (HPLC), capillary electrophoresis (CE), and tandem mass spectrometry (MS/MS). All these methods detect HbS, as well as low-level or absent HbA, and also other variants of hemoglobin. HPLC is considered as a reference method for NBS, because of its high sensitivity and specificity in detecting HbS. NeoSickle®, a fully automated matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) platform, combined with automated sample processing, a laboratory information management system and NeoSickle® software for automatic data interpretation, has increased the throughput of SCD testing. The purpose of this study was to compare the performances of NeoSickle® and HPLC. A prospective study was conducted including 9,571 samples from the NBS program to compare MALDI-MS using NeoSickle® with an HPLC method. Correlation between the two methods was studied. For the MALDI-MS method, sensitivity, specificity, NPV, and PPV were calculated. We found over 99.4 % correlation between the HPLC and MALDI-MS results. NeoSickle® showed 100 % of sensitivity and specificity in detecting SCD syndrome, leading to positive and negative predictive values of 100 %. NeoSickle® is adapted to NBS for SCD, and can be used in first-line high-throughput screening to detect HbS, and beta-thalassemia major warning. When HbS is detected, second-line use of another specific method as HPLC is necessary.

Lithofacies Identification from Wire-Line Logs Using an Unsupervised Data Clustering Algorithm

Stratigraphic identification from wire-line logs and core samples is a common method for lithology classification. This traditional approach is considered superior, despite its significant financial cost. Artificial neural networks and machine learning offer alternative, cost-effective means for automated data interpretation, allowing geoscientists to extract insights from data. At the same time, supervised and semi-supervised learning techniques are commonly employed, requiring a sufficient amount of labeled data to be generated through manual interpretation. Typically, there are abundant unlabeled geophysical data while labeled data are scarcer. Supervised and semi-supervised techniques partially address the cost issue. An underutilized class of machine-learning-based methods, unsupervised data clustering, can perform consonant classification by grouping similar data without requiring known results, presenting an even more cost-effective solution. In this study, we examine a state-of-the-art unsupervised data clustering algorithm called piecemeal clustering to identify lithofacies from wire-line logs, effectively addressing these challenges. The piecemeal clustering algorithm groups similar wire-log signatures into clusters, determines the number of clusters present in the data, and assigns each signature to one of the clusters, each of which represents a lithofacies. To evaluate the performance, we tested the algorithm on publicly released data from ten wells drilled in the Hugoton and Panoma fields of southwest Kansas and northwest Oklahoma, respectively. The data consist of two major groups: marine and non-marine facies. The study herein is centered around addressing two fundamental research questions regarding the accuracy and practicality of the piecemeal clustering algorithm. The algorithm successfully identified nine distinct clusters in our dataset, aligning with the cluster count observed in previously published works employing the same data. Regarding mapping accuracy, the results were notable, with success rates of 81.90% and 45.20% with and without considering adjacent facies, respectively. Further detailed analysis of the results was conducted for individual types of facies and independently for each well. These findings suggest the algorithm’s precision in characterizing the geological formations. To assess its performance, a comprehensive comparative analysis was conducted, encompassing other data clustering algorithms, as well as supervised and semi-supervised machine learning techniques. Notably, the piecemeal clustering algorithm outperformed alternative data clustering methods. Furthermore, despite its unsupervised nature, the algorithm demonstrated competitiveness by yielding results comparable to, or even surpassing, those obtained through supervised and semi-supervised techniques.

Window into the mind: Advanced handheld spectroscopic eye-safe technology for point-of-care neurodiagnostic.

Traumatic brain injury (TBI), a major cause of morbidity and mortality worldwide, is hard to diagnose at the point of care with patients often exhibiting no clinical symptoms. There is an urgent need for rapid point-of-care diagnostics to enable timely intervention. We have developed a technology for rapid acquisition of molecular fingerprints of TBI biochemistry to safely measure proxies for cerebral injury through the eye, providing a path toward noninvasive point-of-care neurodiagnostics using simultaneous Raman spectroscopy and fundus imaging of the neuroretina. Detection of endogenous neuromarkers in porcine eyes' posterior revealed enhancement of high-wave number bands, clearly distinguishing TBI and healthy cohorts, classified via artificial neural network algorithm for automated data interpretation. Clinically, translating into reduced specialist support, this markedly improves the speed of diagnosis. Designed as a hand-held cost-effective technology, it can allow clinicians to rapidly assess TBI at the point of care and identify long-term changes in brain biochemistry in acute or chronic neurodiseases.

Organic contaminants in bio-based fertilizer treated soil: Target and suspect screening approaches

Using bio-based fertilizer (BBF) in agricultural soil can reduce the dependency on chemical fertilizer and increase sustainability by recycling nutrient-rich side-streams. However, organic contaminants in BBFs may lead to residues in the treated soil. This study assessed the presence of organic contaminants in BBF treated soils, which is essential for evaluating sustainability/risks of BBF use. Soil samples from two field studies amended with 15 BBFs from various sources (agricultural, poultry, veterinary, and sludge) were analyzed. A combination of QuEChERS-based extraction, liquid chromatography quadrupole time of flight mass spectrometry-based (LC-QTOF-MS) quantitative analysis, and an advanced, automated data interpretation workflow was optimized to extract and analyze organic contaminants in BBF-treated agricultural soil. The comprehensive screening of organic contaminants was performed using target analysis and suspect screening. Of the 35 target contaminants, only three contaminants were detected in the BBF-treated soil with concentrations ranging from 0.4 ng g−1 to 28.7 ng g−1; out of these three detected contaminants, two were also present in the control soil sample. Suspect screening using patRoon (an R-based open-source software platform) workflows and the NORMAN Priority List resulted in tentative identification of 20 compounds (at level 2 and level 3 confidence level), primarily pharmaceuticals and industrial chemicals, with only one overlapping compound in two experimental sites. The contamination profiles of the soil treated with BBFs sourced from veterinary and sludge were similar, with common pharmaceutical features identified. The suspect screening results suggest that the contaminants found in BBF-treated soil might come from alternative sources other than BBFs.

Online Service with Automated Interpretation of Sequencing Data and Prediction of Pyrazinamide Resistance in Mycobacterium tuberculosis.

Pyrazinamide plays an important role in the treatment of tuberculosis. However, the microbiological test for pyrazinamide resistance is more complex and less reliable than testing of susceptibility to other anti-tuberculosis drugs due to the need to grow the pathogen at pH 5.5. Identification of mutations that cause resistance to anti-tuberculosis drugs can replace microbiological methods. Mutations in the pncA gene are responsible for the main mechanism of the resistance to pyrazinamide and are found in more than 90% of resistant strains. However, the genetic method for determining drug susceptibility is very complex, because mutations leading to pyrazinamide resistance are diverse and scattered throughout the gene. We have developed a software package for automatic data interpretation and prediction of the resistance to pyrazinamide based on Sanger sequencing results. The effectiveness of detection of pyrazinamide resistance in 16 clinical samples was compared using the BACTEC MGIT 960 automated system and pncA gene Sanger sequencing with automated analysis of the results. A significant advantage of the developed method over a single microbiological study was shown, due to greater reliability of the results irrespective of the purity of isolates.

Synthesis and Characterization of High‐Entropy CrMoNbTaVW Thin Films Using High‐Throughput Methods

High‐entropy alloys (HEAs) or complex concentrated alloys (CCAs) offer a huge research area for new material compositions and potential applications. Since the combination of several elements sometimes leads to unexpected and unpredictable material properties. In addition to the element combinations, the optimization of the element proportions in CCAs and HEAs is also a decisive factor in tailoring desired material properties. However, it is almost impossible to achieve the composition and characterization of CCAs and HEAs with a sufficient number of compositions by conventional experiments. Therefore, an optimized high‐throughput magnetron sputtering synthesis to fabricate a new HEA gradient layer of six elements is presented. With this approach, the compositional space of the HEA system CrMoNbTaVW can be studied in different subsections to determine the influence of the individual elements and their combinations on the structure, morphology, and physical properties (hardness and resistivity). It is found that the Cr‐, Ta‐, and W‐rich phases, which have a grain size of 10–11 nm, exhibit the hardest mechanical properties, whereas V‐, Ta‐, and Cr‐rich compounds exhibit the highest electrical resistivity. The combination of high‐throughput synthesis, automated analysis tools, and automated data interpretation enables rapid and time‐efficient characterization of the novel CrMoNbTaVW gradient film.

SPIN enables high throughput species identification of archaeological bone by proteomics.

Species determination based on genetic evidence is an indispensable tool in archaeology, forensics, ecology, and food authentication. Most available analytical approaches involve compromises with regard to the number of detectable species, high cost due to low throughput, or a labor-intensive manual process. Here, we introduce “Species by Proteome INvestigation” (SPIN), a shotgun proteomics workflow for analyzing archaeological bone capable of querying over 150 mammalian species by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Rapid peptide chromatography and data-independent acquisition (DIA) with throughput of 200 samples per day reduce expensive MS time, whereas streamlined sample preparation and automated data interpretation save labor costs. We confirm the successful classification of known reference bones, including domestic species and great apes, beyond the taxonomic resolution of the conventional peptide mass fingerprinting (PMF)-based Zooarchaeology by Mass Spectrometry (ZooMS) method. In a blinded study of degraded Iron-Age material from Scandinavia, SPIN produces reproducible results between replicates, which are consistent with morphological analysis. Finally, we demonstrate the high throughput capabilities of the method in a high-degradation context by analyzing more than two hundred Middle and Upper Palaeolithic bones from Southern European sites with late Neanderthal occupation. While this initial study is focused on modern and archaeological mammalian bone, SPIN will be open and expandable to other biological tissues and taxa.

Machine learning-based image processing in materials science and engineering: A review

Machine learning (ML) is playing a great role in every sphere of life including the area of materials science and engineering. At times, the experimental and computational data together create a complex scenario for interpretation giving rise to the requirement of ML. The incorporation of ML in the development of material science has a promised future to accelerate the research through automatic data interpretation with the help of different models like classification, Regression, Clustering, etc. Therefore, it becomes quick and easy to analyze the data sets with greater accuracy as compared to the analysis done by human researchers. This paper gives an overview of the applications and applicability of Machine Learning being used in various areas of materials development. The various image processing techniques which can be applied in material science are discussed in detail. In addition to that, the paper gives detail as to how a model can be trained with a smaller number of datasets. Subsequently, the different data ecosystems which can be useful for data collection and preparation before selecting any Machine Learning model are highlighted. Furthermore, the various opportunities and challenges that material scientists have been facing in the development of new materials are also outlined for a better understanding of ML-based image processing in material science and engineering.

In vivo detection of head and neck tumors by hyperspectral imaging combined with deep learning methods.

Currently, there are no fast and accurate screening methods available for head and neck cancer, the eighth most common tumor entity. For this study, we used hyperspectral imaging, an imaging technique for quantitative and objective surface analysis, combined with deep learning methods for automated tissue classification. As part of a prospective clinical observational study, hyperspectral datasets of laryngeal, hypopharyngeal and oropharyngeal mucosa were recorded in 98 patients before surgery in vivo. We established an automated data interpretation pathway that can classify the tissue into healthy and tumorous using convolutional neural networks with 2D spatial or 3D spatio-spectral convolutions combined with a state-of-the-art Densenet architecture. Using 24 patients for testing, our 3D spatio-spectral Densenet classification method achieves an average accuracy of 81%, a sensitivity of 83% and a specificity of 79%.

Light Intensity Modulated Photo-Electrochemical Methods As a Tool in Investigation of Perovskites for Photovoltaics

Electrical Impedance Spectroscopy, abbreviated is EIS and often thus explained as electrochemical impedance spectroscopy, found its place in material studies as a routine experiment. ISE proliferation became possible through advances in electronics. Current-voltage characteristics were standard features of potentiostats. Impedance techniques require additional components embedded into a potentiostat, namely signal generator and signal analyzer. Relatively simple devices are available, although dedicated instrumentation with little noise, broad frequency control and perhaps sample temperature control or automated sample handling are particularly desired and command premium in cost.Along with impedance over frequency data collection comes data interpretation, which is done with the assistance of software. Entirely automated data interpretation is also possible, though its utility in inexperienced hands is debatable. EIS as an electrical technique uses as the perturbing signal alternating electrical signal, either potential of current. In general, impedance studies are intended to study a transfer function and the input and output do not have to be only voltage-current or current-voltage as it would be in EIS. In mechanical engineering are for example useful studies focused on mechanical perturbation.In the field of material studies on system responding to illumination is finding its way impedance study in which the perturbing signal is light of certain wavelength (or white light) with variable intensity.Here we will look at the link between EIS and electrical response spectroscopy induced by fluctuating light. This is sometimes for short dubbed as optical spectroscopy although it does make a use of the of instruments for optical spectroscopy (such as UV-VIS, IR etc.) as they are known from analytical chemistry literature. Rather, an electric signal in a photoactive material is generated by intensity modulated light at different frequencies (similar range to electrical impedance spectroscopy) and a response (the magnitude and the phase shift) is analyzed. In this respect the magnitude and phase shift correspond to the magnitude and a vector of resistance on the polar plot of electrical impedance measurement. However, the data obtained serve a different purpose. For example, in a photovoltaic shorted system, the time constant associated with the response corresponds to mobility of the photogenerated charge in the material. For an open-circuit potential of such a photovoltaic system the measured time constant response corresponds to the rate of recombination of the generated hole-electron species.As the time constants are important in development of a photovoltaic system for harvesting energy from the Sun, such studies are complementary to obtaining such practical data as the fill factor or maximum power point. We will demonstrate some of these results on perovskite-type materials designed by our collaborative partners.Such materials and related constructions of particular solar cells could be possible replacements for the current silicon-based photovoltaic cells. Solar cells based on the general perovskite structure (ABX3) attracted in recent years much attention. These materials, due to the possible variability of both cations and anions in their composition, offer almost endless possibilities for synthetic preparation. For the active layer of a solar cell is often used the CH3NH3PbI3-2Cl2 perovskite structure, i.e., a structure containing an organic component. We have built and then tested such photovoltaic cells by static and dynamic methods using a unique photoelectric research instrument using the method CIMPS (Controlled Intensity Modulated Photocurrent Spectroscopy).

Untargeted Metabolomics for Metabolic Diagnostic Screening with Automated Data Interpretation Using a Knowledge-Based Algorithm.

Untargeted metabolomics may become a standard approach to address diagnostic requests, but, at present, data interpretation is very labor-intensive. To facilitate its implementation in metabolic diagnostic screening, we developed a method for automated data interpretation that preselects the most likely inborn errors of metabolism (IEM). The input parameters of the knowledge-based algorithm were (1) weight scores assigned to 268 unique metabolites for 119 different IEM based on literature and expert opinion, and (2) metabolite Z-scores and ranks based on direct-infusion high resolution mass spectrometry. The output was a ranked list of differential diagnoses (DD) per sample. The algorithm was first optimized using a training set of 110 dried blood spots (DBS) comprising 23 different IEM and 86 plasma samples comprising 21 different IEM. Further optimization was performed using a set of 96 DBS consisting of 53 different IEM. The diagnostic value was validated in a set of 115 plasma samples, which included 58 different IEM and resulted in the correct diagnosis being included in the DD of 72% of the samples, comprising 44 different IEM. The median length of the DD was 10 IEM, and the correct diagnosis ranked first in 37% of the samples. Here, we demonstrate the accuracy of the diagnostic algorithm in preselecting the most likely IEM, based on the untargeted metabolomics of a single sample. We show, as a proof of principle, that automated data interpretation has the potential to facilitate the implementation of untargeted metabolomics for metabolic diagnostic screening, and we provide suggestions for further optimization of the algorithm to improve diagnostic accuracy.

Model-based analysis of biocatalytic processes and performance of microbioreactors with integrated optical sensors

Design and development of scale-down approaches, such as microbioreactor (μBR) technologies with integrated sensors, are an adequate solution for rapid, high-throughput and cost-effective screening of valuable reactions and/or production strains, with considerably reduced use of reagents and generation of waste. A significant challenge in the successful and widespread application of μBRs in biotechnology remains the lack of appropriate software and automated data interpretation of μBR experiments. Here, it is demonstrated how mathematical models can be usedas helpful tools, not only to exploit the capabilities of microfluidic platforms, but also to reveal the critical experimental conditions when monitoring cascade enzymatic reactions. A simplified mechanistic model was developed to describe the enzymatic reaction of glucose oxidase and glucose in the presence of catalase inside a commercial microfluidic platform with integrated oxygen sensor spots. The proposed model allowed an easy and rapid identification of the reaction mechanism, kinetics and limiting factors. The effect of fluid flow and enzyme adsorption inside the microfluidic chip on the optical sensor response and overall monitoring capabilities of the presented platform was evaluated via computational fluid dynamics (CFD) simulations. Remarkably, the model predictions were independently confirmed for μL- and mL- scale experiments. It is expected that the mechanistic models will significantly contribute to the further promotion of μBRs in biocatalysis research and that the overall study will create a framework for screening and evaluation of critical system parameters, including sensor response, operating conditions, experimental and microbioreactor designs.

MTrack: Automated Detection, Tracking, and Analysis of Dynamic Microtubules

Microtubules are polar, dynamic filaments fundamental to many cellular processes. In vitro reconstitution approaches with purified tubulin are essential to elucidate different aspects of microtubule behavior. To date, deriving data from fluorescence microscopy images by manually creating and analyzing kymographs is still commonplace. Here, we present MTrack, implemented as a plug-in for the open-source platform Fiji, which automatically identifies and tracks dynamic microtubules with sub-pixel resolution using advanced objection recognition. MTrack provides automatic data interpretation yielding relevant parameters of microtubule dynamic instability together with population statistics. The application of our software produces unbiased and comparable quantitative datasets in a fully automated fashion. This helps the experimentalist to achieve higher reproducibility at higher throughput on a user-friendly platform. We use simulated data and real data to benchmark our algorithm and show that it reliably detects, tracks, and analyzes dynamic microtubules and achieves sub-pixel precision even at low signal-to-noise ratios.

A Multicentre Pilot Study of a Two-Tier Newborn Sickle Cell Disease Screening Procedure with a First Tier Based on a Fully Automated MALDI-TOF MS Platform.

The reference methods used for sickle cell disease (SCD) screening usually include two analytical steps: a first tier for differentiating haemoglobin S (HbS) heterozygotes, HbS homozygotes and β-thalassemia from other samples, and a confirmatory second tier. Here, we evaluated a first-tier approach based on a fully automated matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) platform with automated sample processing, a laboratory information management system and NeoSickle® software for automatic data interpretation. A total of 6701 samples (with high proportions of phenotypes homozygous (FS) or heterozygous (FAS) for the inherited genes for sickle haemoglobin and samples from premature newborns) were screened. The NeoSickle® software correctly classified 98.8% of the samples. This specific blood sample collection was enriched in qualified difficult samples (premature newborns, FAS samples, late and very late samples, etc.). In this study, the sensitivity of FS sample detection was found to be 100% on the Lille MS facility and 99% on the Dijon MS facility, and the specificity of FS sample detection was found to be 100% on both MS facilities. The MALDI-MS platform appears to be a robust solution for first-tier use to detect the HbS variant: it is reproducible and sensitive, it has the power to analyze 600–1000 samples per day and it can reduce the unit cost of testing thanks to maximal automation, minimal intervention by the medical team and good overall practicability. The MALDI-MS approach meets today’s criteria for the large-scale, cost-effective screening of newborns, children and adults.

Preserved Microarrays for Simultaneous Detection and Identification of Six Fungal Potato Pathogens with the Use of Real-Time PCR in Matrix Format.

Fungal diseases of plants are of great economic importance causing 70–80% of crop losses associated with microbial plant pathogens. Advanced on-site disease diagnostics is very important to maximize crop productivity. In this study, diagnostic systems have been developed for simultaneous detection and identification of six fungal pathogens using 48-well microarrays (micromatrices) for qPCR. All oligonucleotide sets were tested for their specificity using 59 strains of target and non-target species. Detection limit of the developed test systems varied from 0.6 to 43.5 pg of DNA depending on target species with reproducibility within 0.3−0.7% (standard deviation). Diagnostic efficiency of test systems with stabilized and freeze-dried PCR master-mixes did not significantly differ from that of freshly prepared microarrays, though detection limit increased. Validation of test systems on 30 field samples of potato plants showed perfect correspondence with the results of morphological identification of pathogens. Due to the simplicity of the analysis and the automated data interpretation, the developed microarrays have good potential for on-site use by technician-level personnel, as well as for high-throughput monitoring of fungal potato pathogens.

Routine detection of carbapenem-resistant gram-negative bacilli in clinical laboratories. A review of current challenge.

The detection of carbapenem-resistant organisms (CROs) represents a substantial challenge for many clinical laboratories. In this review, several phenotypic and non-phenotypic methods for detecting CROs are discussed. However, no consensus has yet been reached with regards to the single most optimal method. Due to differences in carbapenem-resistant activity between carbapenemases, the simultaneous use of 2 or more phenotypic detection methods can improve the detection of CROs compared with a single technique. Molecular methods are currently favored because the majority can be performed rapidly with a high level of accuracy. Whole-genome sequencing (WGS) yields unambiguous data pertaining to complete analysis of the entire genome and may ultimately become a highly powerful tool in routine clinical settings. However, WGS is still relatively expensive and requires an automated data interpretation system. The routine implementation of this technique in clinical laboratories may not occur for several years, particularly in developing countries.

Status Report on the High-Throughput Characterization of Complex Intact O-Glycopeptide Mixtures.

A very complex mixture of intact, human N- and O-glycopeptides, enriched from the tryptic digest of urinary proteins of three healthy donors using a two-step lectin affinity enrichment, was analyzed by LC-MS/MS, leading to approximately 45,000 glycopeptide EThcD spectra. Two search engines, Byonic and Protein Prospector, were used for the interpretation of the data, and N- and O-linked glycopeptides were assigned from separate searches. The identification rate was very low in all searches, even when results were combined. Thus, we investigated the reasons why was it so, to help to improve the identification success rate. Focusing on O-linked glycopeptides, we noticed that in EThcD, larger glycan oxonium ions better survive the activation than those in HCD. These fragments, combined with reducing terminal Y ions, provide important information about the glycan(s) present, so we investigated whether filtering the peaklists for glycan oxonium ions indicating the presence of a tetra- or hexasaccharide structure would help to reveal all molecules containing such glycans. Our study showed that intact glycans frequently do not survive even mild supplemental activation, meaning one cannot rely on these oxonium ions exclusively. We found that ETD efficiency is still a limiting factor, and for highly glycosylated peptides, the only information revealed in EThcD was related to the glycan structures. The limited overlap of results delivered by the two search engines draws attention to the fact that automated data interpretation of O-linked glycopeptides is not even close to being solved. Graphical abstract ᅟ.

Software system for automated data interpretation of potential fields (GMT-Auto)

Software system for automated data interpretation of potential fields (GMT-Auto)

Development and validation of a noncontact spectroscopic device for hemoglobin estimation at point-of-care.

Anemia severely and adversely affects human health and socioeconomic development. Measuring hemoglobin with the minimal involvement of human and financial resources has always been challenging. We describe a translational spectroscopic technique for noncontact hemoglobin measurement at low-resource point-of-care settings in human subjects, independent of their skin color, age, and sex, by measuring the optical spectrum of the blood flowing in the vascular bed of the bulbar conjunctiva. We developed software on the LabVIEW platform for automatic data acquisition and interpretation by nonexperts. The device is calibrated by comparing the differential absorbance of light of wavelength 576 and 600 nm with the clinical hemoglobin level of the subject. Our proposed method is consistent with the results obtained using the current gold standard, the automated hematology analyzer. The proposed noncontact optical device for hemoglobin estimation is highly efficient, inexpensive, feasible, and extremely useful in low-resource point-of-care settings. The device output correlates with the different degrees of anemia with absolute and trending accuracy similar to those of widely used invasive methods. Moreover, the device can instantaneously transmit the generated report to a medical expert through e-mail, text messaging, or mobile apps.

Automatic classification of glycaemia measurements to enhance data interpretation in an expert system for gestational diabetes

Expert systems for diabetes care need to automatically evaluate glycaemia measurements in relationship to meals to correctly determine patients’ metabolic condition and generate recommendations about therapy adjustments. Most glucose meters allow patients to manually label each measurement with a meal tag, but as this utility is not always used, a completion procedure is needed. Classification methods are usually based on predefined mealtimes and present insufficient accuracy that might affect the automatic data analysis. Expert systems in diabetes require a reliable method to manage incomplete glycaemia data so that they can determine if patients’ metabolic condition is altered due to a specific meal or due to an extended fasting period.This paper presents the design and application of a classification module to automatically assign the appropriate meal and ‘moment of measurement’ to incomplete glycaemia data. Different machine learning techniques were studied in order to design the best classification algorithm in terms of accuracy. The selected classifier was implemented with a C4.5 decision tree with 7 input features selected with a wrapper evaluator and the genetic search algorithm, which achieved 95.45% of accuracy with the training set on cross-validation. The classification module was integrated in the Sinedie expert system for gestational diabetes care and was evaluated in a clinical environment for 8 months with 42 patients. A total of 7,113 glycaemia measurements were uploaded by patients into the Sinedie system and were completed by the “classification module”. The 98.79% of the measurements were correctly classified, while patients modified the automatic classification of 1.21% of them. Classification results were improved by 21.04% compared to a classification based on predefined mealtimes. The automatic classification of glycaemia measurements minimizes the patient's intervention, allows structuring measurements in relationship to meals and makes automatic data interpretation by expert systems more reliable.