RESUMEN

The worldwide increase of antimicrobial resistance (AMR) is a serious threat to human health. To avert the spread of AMR, fast reliable diagnostics tools that facilitate optimal antibiotic stewardship are an unmet need. In this regard, Raman spectroscopy promises rapid label- and culture-free identification and antimicrobial susceptibility testing (AST) in a single step. However, even though many Raman-based bacteria-identification and AST studies have demonstrated impressive results, some shortcomings must be addressed. To bridge the gap between proof-of-concept studies and clinical application, we have developed machine learning techniques in combination with a novel data-augmentation algorithm, for fast identification of minimally prepared bacteria phenotypes and the distinctions of methicillin-resistant (MR) from methicillin-susceptible (MS) bacteria. For this we have implemented a spectral transformer model for hyper-spectral Raman images of bacteria. We show that our model outperforms the standard convolutional neural network models on a multitude of classification problems, both in terms of accuracy and in terms of training time. We attain more than 96% classification accuracy on a dataset consisting of 15 different classes and 95.6% classification accuracy for six MR-MS bacteria species. More importantly, our results are obtained using only fast and easy-to-produce training and test data.

Asunto(s)

Antiinfecciosos , Espectrometría Raman , Bacterias , Humanos , Aprendizaje Automático , Meticilina , Fenotipo , Espectrometría Raman/métodos

RESUMEN

The presence of pronounced chemical signatures of several important molecules and gases in the mid-infrared (mid-IR) spectral region has enabled numerous spectroscopic applications in diverse fields such as environmental monitoring and the life sciences. This has motivated the development of new, efficient light sources and detectors in this spectral region. In this work, we report on the development and characterization of a simple, low-cost, tunable mid-IR light source, covering the 1.45 µm to 3.5 µm spectral region, based on spontaneous parametric down conversion (SPDC) in MgO doped periodically poled lithium niobate (MgO:PPLN). The spectral coverage can be easily extended to 5 µm simply by choosing a crystal with an appropriate poling period. A low repetition rate, passively Q-switched laser provides high-energy pump pulses at 1.03 µm for the SPDC source, resulting in the generation of signal and idler fields. The source is characterized in terms of pulse-to-pulse spectral stability of the generated signal field. Interestingly, a high pulse-to-pulse energy stability does not necessarily lead to high spectral stability. Pulse-to-pulse spectral stability is measured at different pump pulse energies, showing improved spectral stability when operated well above the SPDC threshold.

RESUMEN

The use of fluorescence in microscopy is a well-known technology. Due to autofluorescence in the materials of optical components, the contrast of the image is degraded. The calculation of autofluorescence is usually performed by brute-force methods such as the Monte Carlo-based volume scattering. The efficiency of calculations in this case is extremely low, and a huge number of rays must be calculated. In stray light calculations, the concept of important sampling is used to reduce computational effort. The idea is to calculate only rays, which have the chance to reach the target surface. The fluorescence conversion can be considered to be a scatter process, and therefore a modification of this idea is used here. The reduction factor is calculated by comparing the size of the illuminated phase space domain with the corresponding acceptance domain in every z plane of the lenses. The boundaries of the domains are determined by tracing the limiting rays of the light cone of the source as well as the pixel area under consideration. The small overlap of both domains can be estimated by geometrical considerations. The correct photometric scaling and the discretization of the volumes must be performed. The errors resulting from necessary approximations can be corrected without greatly increasing computational effort. The run time is reduced by a factor of 104. It is shown with some practical examples of microscope lenses that the results are comparable with conventional methods. Additionally, a quasi-analytical model that describes the dependence of autofluorescence on various lens parameters is derived.

RESUMEN

We demonstrate a low-cost 343 nm solid-state laser delivering up to 20 µJ per pulse, with a pulse width of 2.3 ns at a repetition rate of 100 Hz. The 343 nm is obtained through a third harmonic generation of a passively Q-switched 1030 nm Yb:YAG laser with pulse energy of 190 µJ at 100 Hz and a pulse width of 5.4 ns. The IR-UV conversion efficiency is 10.4%, comparable to that achieved with mode-locked IR lasers. The light source is electronically controlled for easy synchronization with a detection circuit. The low repetition rate specifically targets applications exploiting the millisecond scale lifetime of lanthanides employed in fluoroimmunoassay measurements for time-resolved fluorescence spectroscopy. Low repetition rate and even pulse-on-demand operation is demonstrated.

RESUMEN

High-sensitivity cardiac troponin assay development enables determination of biological variation in healthy populations, more accurate interpretation of clinical results and points towards earlier diagnosis and rule-out of acute myocardial infarction. In this paper, we report on preliminary tests of an immunoassay analyzer employing an optimized LED excitation to measure on a standard troponin I and a novel research high-sensitivity troponin I assay. The limit of detection is improved by factor of 5 for standard troponin I and by factor of 3 for a research high-sensitivity troponin I assay, compared to the flash lamp excitation. The obtained limit of detection was 0.22 ng/L measured on plasma with the research high-sensitivity troponin I assay and 1.9 ng/L measured on tris-saline-azide buffer containing bovine serum albumin with the standard troponin I assay. We discuss the optimization of time-resolved detection of lanthanide fluorescence based on the time constants of the system and analyze the background and noise sources in a heterogeneous fluoroimmunoassay. We determine the limiting factors and their impact on the measurement performance. The suggested model can be generally applied to fluoroimmunoassays employing the dry-cup concept.

RESUMEN

We report on the design, development and investigation of an optical system based on UV light emitting diode (LED) excitation at 340 nm for time-resolved fluorescence detection of immunoassays. The system was tested to measure cardiac marker Troponin I with a concentration of 200 ng/L in immunoassay. The signal-to-noise ratio was comparable to state-of-the-art Xenon flash lamp based unit with equal excitation energy and without overdriving the LED. We performed a comparative study of the flash lamp and the LED based system and discussed temporal, spatial, and spectral features of the LED excitation for time-resolved fluorimetry. Optimization of the suggested key parameters of the LED promises significant increase of the signal-to-noise ratio and hence of the sensitivity of immunoassay systems.