Optical microscope and method for imaging with an optical microscope - Patents.com

Abstract

A method for imaging by an optical microscope includes the steps of: sending illumination light (12) to a sample (35), sending detection light (15) from the sample (35) to a plurality of photon-counting sensor elements (61), each capturing a plurality of count values ​​(x) in succession, forming (S3) a plurality of analyzed count value frequency distributions (81-83) and at least one reference count value frequency distribution (80) from the count values ​​(x), calculating a similarity between each one of the analyzed count value frequency distributions (81-83) and the reference count value frequency distribution (80), and identifying an overloaded sensor element (61) depending on the calculated similarity of the associated analyzed count value frequency distributions (81-83). Further, a corresponding optical microscope is disclosed.

Description

[Technical field]

[0001] The present disclosure relates to a method for image capture with an optical microscope as claimed in claim 1 and to an optical microscope as claimed in claim 19. [Background technology]

[0002] Especially for biological samples, strong fluorescent markers are used. High illumination intensity is useful for fast image data acquisition. However, the illumination intensity should not be too high, as the sample may be damaged by the illumination.

[0003] For this reason among others, particularly sensitive sensor elements are used. Photomultiplier tubes (PMTs) with multi-alkali cathodes have often been substituted by PMTs with GaAsP cathodes, which allowed to achieve improved quantum efficiencies in the visible spectral range. However, the multiplicative nature of the enhancement process used in this case introduces noise into the measured signal (so-called amplification noise or excess noise). This results in a lower net sensitivity for a given quantum efficiency.

[0004] To eliminate the multiplicative noise, the amplification can be increased until the individual pulses caused by individual photons can be counted. For example, electronics capable of time-correlated single-photon counting (TCSPC) are used for this purpose. The measured signal is a purely digital signal, and the measured intensity value can correspond to the number of photons impinging on the photocathode, reduced by the quantum efficiency. However, the electronics have a dead time, for example, tens of nanoseconds, during which the impinging photons cannot be recorded.

[0005] As a result of the dead time, the signal can be corrupted if the illumination density of the sensor is too high. In particular, if the illumination intensity on the sensor elements is high, the measured values ​​can become too small (so-called pile-up effect). Photon-counting measurements are therefore very sensitive, but they take a relatively long time to achieve a certain signal-to-noise ratio (SNR) in the image data. This problem can be reduced by parallel photon-counting methods, where the detection PSF (point spread function) of the signal is distributed over an array or structure of several photon-counting sensor elements. The photons thus impinge on a large number of sensor elements in a statistically distributed manner. There is thus a high probability that most sensor elements are active when a photon impinges, and only a few sensor elements are in the dead time after recording the previous photon.

[0006] A method according to the present invention for imaging with an optical microscope includes at least the following steps: sending illumination light to a sample; and sending detection light from the sample to a number of photon-counting sensor elements, each of which captures a number of counts in succession.

[0007] In a corresponding manner, an optical microscope according to the present invention includes a light source for emitting illumination light to a sample and a number of photon-counting sensor elements for sensing detection light from the sample, the photon-counting sensor elements being each equipped to capture a number of counts in succession.

[0008] In this case, even if the sensitivity of the individual sensor elements can in principle be set individually, as described, for example, in the patent document EP 1 239 633 A1, excessive loading of the sensor elements is problematic. In particular, if the user increases the illumination power in order to achieve the best possible image quality, even an array of sensor elements leads to undesirable saturation with the falsification of the image information. In the case of low illumination intensity, the image quality can be improved by increasing the exposure time per sample point, but this leads to longer measurement times, which are undesirable for observing the progress of time. [Prior art documents] [Patent documents]

[0009] [Patent Document 1] U.S. Patent No. 9,997,551 Summary of the Invention [Problem to be solved by the invention]

[0010] It can be an object of the present invention to provide an optical microscope and a method for image acquisition by means of an optical microscope using photon-counting sensor elements, which recognizes and possibly corrects image acquisition errors. [Means for solving the problem]

[0011] This problem is solved by a method having the features of claim 1 and by an optical microscope having the features of claim 19.

[0012] Advantageous variants of the optical microscope according to the invention and the method according to the invention are the subject of the dependent claims and are explained in the following description.

[0013] The method as described above comprises, according to the invention, at least the following steps: forming a plurality of analyzed photon count distributions and at least one reference photon count distribution from the photon counts; - calculating the similarity between each one of the count value frequency distributions to be analyzed and a reference count value frequency distribution; - identifying an overloaded sensor element depending on the calculated similarity of the associated analyzed count frequency distributions; is executed.

[0014] According to the present invention, the optical microscope as described above comprises: - forming a plurality of analyzed count frequency distributions and at least one reference count frequency distribution from the counts; - to calculate the similarity between each one of the count value frequency distributions to be analyzed and a reference count value frequency distribution; - to identify excessive loading of the sensor element depending on the calculated similarity of the associated analyzed count frequency distributions; The control unit is provided in the

[0015] In various embodiments of the invention, the sensor elements measure a very similar image content, but receive different frequencies of light intensity. A relatively easy-to-understand example is the so-called Airyscan (also called image scanning microscopy), where a sample point is illuminated and this sample point is mapped onto an array of sensor elements. The detection PSF is spread across the sensor elements, so that the measurements result from approximately the same sample point. However, the detection PSF does not have a constant value across the sensor elements, but may take a Gaussian curve, for example, laterally across the sensor elements. The central sensor elements therefore receive more light than the outer sensor elements and measure higher counts than the outer sensor elements. The sample can be scanned with an illumination beam, and the sensor elements measure one count for each illuminated sample point. The counts of the sensor elements can be integrated into one count frequency distribution, so that, for example, one count frequency distribution per sensor element is formed. The count frequency distributions are due to approximately the same image content, but the magnitudes of the counts differ, mainly due to the course of the detection PSF. In particular, after a rescaling of the count frequency distributions, which corrects / standardizes the magnitudes of the different counts, the count frequency distributions should therefore approximately coincide with each other. If a sensor element is overloaded (saturated) during the measurement course, its highest count will be too small. It can therefore be recognized by comparing the count frequency distribution whether a sensor element is overloaded during the capture of the counts and thus corresponds to corrupted image information. If a sensor element is identified as being overloaded, then optionally appropriate measures can be automatically performed, such as a mathematical correction of the overloaded counts or a new image capture with appropriately changed microscope settings, which will be described in more detail later.

[0016] Formation of count frequency distribution For the comparison of the count frequency distributions, at least one count frequency distribution is used as a reference count frequency distribution, which is known or assumed to be free of saturation during the acquisition of the counts, as will be explained in more detail later. The other count frequency distributions are referred to as the analyzed count frequency distributions.

[0017] The count value frequency distribution may be formed from the count value of just one sensor element or from the count values ​​of several sensor elements. The number of count value frequency distributions, i.e. the number of analyzed count value frequency distributions and reference count value frequency distributions, may therefore be less than, the same as or in principle more than the number of sensor elements. In particular, groups of sensor elements may be formed ("binning"), and the count values ​​of the sensor elements of a group are combined and jointly form a count value frequency distribution. At least one reference count value frequency distribution may likewise be formed in this manner. Alternatively, the reference count value frequency distribution may also be calculated by the aggregation of several count value frequency distributions.

[0018] Calculation of reference count frequency distribution The reference count frequency distribution will be formed from the counts of one or more sensor elements that are not saturated / overstressed during the measurement. This selection may be done in various ways.

[0019] Thus, at least one count value frequency distribution may be selected for use as the reference count value frequency distribution, the selection being made depending on the magnitude of the count value of the corresponding count value frequency distribution. By taking into account the magnitude of the count value, it is possible to ensure that the reference count value frequency distribution is formed from the count values ​​of one or more sensor elements that are not overloaded. Instead of using the selected count value frequency distribution as the reference count value frequency distribution, individual reference count value frequency distributions may also be calculated from the selected count value frequency distributions by aggregation (e.g. by addition, averaging or normalization with subsequent addition or averaging).

[0020] A prerequisite for the count value of a sensor element to be used in the reference count value frequency distribution may be that the maximum value of the count values ​​measured by this sensor element is smaller than a given upper limit value. The prerequisite may be changed to that the average value formed from the maximum count values ​​of this count value frequency distribution (for example the average value of the maximum percentage of the count values) must be smaller than a given upper limit value. The upper limit value may be determined depending on the maximum count rate of the sensor element and may in particular be a number between 1% and 30% of the maximum count rate multiplied by the exposure time. In this case, the exposure time indicates the measurement time for capturing the count values. Such an upper limit value ensures that the count values ​​of the reference count value frequency distribution are not captured at saturation. The maximum count rate may indicate the maximum count rate that can be achieved at saturation of a sensor element. Alternatively, the maximum count rate may be indicated as the inverse of the dead time of the sensor element, for example approximately 0.1 / dead time. The dead time of the sensor element is often known and is a value between 10 and 100 nanoseconds. So, for a sensor element with a dead time of 50 nanoseconds, an upper limit of 2 MHz can be established for the count rate.

[0021] It may also be intended to calculate a signal strength criterion from the counts of the sensor elements, for example the average value of the counts or the maximum value of the counts may be used as the signal strength criterion. A prerequisite for the counts of the sensor elements to be used in the reference count frequency distribution may be that the associated signal strength criterion is above a given minimum value and / or below the aforementioned upper limit value. Exceeding the minimum value ensures that a sufficiently large signal value is used in the reference count frequency distribution.

[0022] Alternatively, one count value frequency distribution may be selected as a reference count value frequency distribution from a number of count value frequency distributions by calculating the correlation between the count value frequency distributions. In particular, one count value frequency distribution that has a given high correlation with a predetermined number of other count value frequency distributions, as large as possible, may be selected as a reference count value frequency distribution. This is based on the idea that often only some sensor elements in the center of the light spot are oversaturated, while most sensor elements operate in the linear region. If one count value frequency distribution has a high correlation with many other count value frequency distributions, the associated sensor element or elements probably operated in the linear (non-saturated) region. Furthermore, depending on the degree of saturation, the shape of the count value frequency distributions changes, so that even if many sensor elements are overloaded, the count value frequency distributions do not always have a high correlation with each other. Optionally, a further prerequisite for use as a reference count value frequency distribution may be that the signal strength criterion of the count value frequency distribution is sufficiently high. For example, a count value frequency distribution in which the signal strength criterion defined in more detail above is maximum or at least the signal strength criterion value is not within the lower half may be selected for the reference count value frequency distribution from the count value frequency distributions that are highly correlated with each other, thereby avoiding the selection of a count value frequency distribution with an unfavorably low SNR of the count value as the reference count value frequency distribution.

[0023] The selection of at least one count value frequency distribution for use as the reference count value frequency distribution may additionally or alternatively be made as a function of the position of one or more associated sensor elements. For example, a count value frequency distribution may be formed from the count values ​​of one or more sensor elements arranged in an outer region and used as the reference count value frequency distribution. The outer region may be taken as, for example, the outermost 30% of all sensor elements. Depending on the measurement method, a binning may be used in which the count value frequency distribution is formed respectively from the count values ​​of sensor elements arranged, for example, in a ring shape. The count value frequency distribution on the outside (or the outermost) side of the ring shape of the sensor elements may be used as the reference count value frequency distribution. The selection of at least one count value frequency distribution for use as the reference count value frequency distribution may also be made as a function of the position of one or more associated sensor elements with respect to the center point of the incident light intensity distribution. In particular, one or more sensor elements may be selected whose position is at a distance with respect to the center point of the light intensity distribution in order to keep the risk of oversaturation of the selected sensor element small.

[0024] Rescaling and similarity determination Optionally, a calculation of a rescaling between each one of the analyzed count value frequency distributions and the reference count value frequency distribution may be performed. A calculation of the similarity between each one of the analyzed count value frequency distributions and the reference count value frequency distribution is then performed taking into account the respective rescaling.

[0025] The purpose of the calculation of the rescaling is to make it possible to better compare the count value frequency distributions of different signal strengths with one another. Rescaling can be understood as changing the area in which the count values ​​of the count value frequency distributions appear and / or the magnitude of the count values. Rescaling can also be called scaling or scaling of the count values ​​of the frequency distributions. If the image contents are similar, the count value frequency distributions should be similar apart from their scaling, i.e. should have similar course shapes. If this is not the case, the cause may lie in saturation of the sensor elements. For the rescaling, the analyzed count value frequency distribution and / or the reference count value frequency distribution can be rescaled. A size standard for the analyzed count value frequency distribution and the reference count value frequency distribution can be determined and one of the count value frequency distributions can be stretched or compressed depending on the size standard so that the size standards subsequently coincide with one another.

[0026] Rescaling can be seen as a (linear) mapping of the analyzed count value frequency distribution onto the reference count value frequency distribution (or vice versa). Therefore, for rescaling, one or more parameters of a mapping function that is applied to the analyzed count value frequency distribution or to the reference count value frequency distribution may be determined.

[0027] Rescaling may, for example, involve fitting a fitting function to the reference count value frequency distribution and also fitting the fitting function to one of the count value frequency distributions to be analyzed, respectively. That is, for each count value frequency distribution, one or more fitting parameters are calculated, which describe the course of the respective count value frequency distribution. An expansion or compression may be performed of one of the analyzed count value frequency distributions or of the reference count value frequency distribution, the expansion / compression being dependent on the calculated fitting parameters. In particular, the expansion / compression may be performed in such a way that fitting the fitting function to the expanded / compressed count value frequency distribution produces the same fitting parameters as the fitting parameters of the count value frequency distribution to be compared.

[0028] The fit function may, for example, exhibit an exponential decay, and in particular the fit function f(x) may be defined as f(x)=f0*exp(-x / c), or may include at least the expression f0*exp(-x / c) in combination with further expressions by multiplication, addition or other means, where x is the count of the count frequency distribution, f is the frequency of the count, exp is the exponential function, and f0 and c are fit parameters. Other fit functions may also be used, and it may be advantageous if the fit function includes only two fit parameters or at most three fit parameters.

[0029] It may be intended that the fitting function is adapted only to a portion of the counts of the count frequency distribution, for example to the half with the larger counts, while the half with the smaller counts is ignored. This has the advantage that the fitting and subsequent rescaling make the frequencies of the maximum counts of the reference count frequency distribution and the analyzed count frequency distribution particularly well comparable to one another. A difference in the frequency of the maximum count x is a particularly clear sign of saturation in the analyzed count frequency distribution.

[0030] In the above example of an exponential function as a fitting function, the calculated values ​​of f0 and c generally differ between the reference count frequency distribution and the count frequency distribution to be analyzed. From the difference between the two calculated values ​​of c (and optionally f0 as well), a factor is calculated that stretches or compresses the gray-value axis (count frequency axis). A subsequent fitting stretches or compresses the gray-value axis to produce the same values ​​of c and f0 for the two count frequency distributions being compared.

[0031] The calculation of the similarity may include at least one calculation of a correlation coefficient R (or a value corresponding thereto) between one of the analyzed count value frequency distributions Hn and a reference count value frequency distribution Hr, in particular after rescaling has been performed between the one of the analyzed count value frequency distributions Hn and the reference count value frequency distribution Hr, for example according to Equation 1. R = covar(Hn, Hr,) / (stdev(Hn)*stdev(Hr)) In this case, covar(Hn, Hr,) is the covariance between Hn and Hr, and stdev(Hn) and stdev(Hr) are the standard deviations of Hn and Hr.

[0032] If the calculated correlation coefficient R is below a determined minimum value, the associated analyzed count value frequency distribution is identified as being overloaded or the sensor element or elements from which the analyzed count value frequency distribution is formed are identified as being overloaded. For example, it may be intended that the similarity of two count value frequency distributions is affirmed if the associated correlation coefficient R or its value is greater than a limit value, in particular between 0.80 and 0.98, in particular between 0.92 and 0.97.

[0033] The rescaling and similarity determination may be performed in a common computational process.

[0034] In a further embodiment, a deviation criterion of the frequency value of the maximum count value is calculated first for a fitting function fitted to these count value frequency distributions. A rescaling can then be applied so that the deviation criterion of the analyzed count value frequency distribution and the deviation criterion of the reference count value frequency distribution can be compared with each other in order to perform a similarity determination. In particular when so calculated, the calculation procedures for rescaling and the similarity determination can be interchanged or performed in a different order.

[0035] Depending on the method of similarity determination, rescaling alone is not necessary. This is especially the case in the following example, where the similarity is determined by calculating a modified cross-correlation between the reference count value frequency distribution and the analyzed count value frequency distribution. The cross-correlation coefficient corr{k} can then be calculated by the mathematical formula 2: corr{k}:=sum [Hn(x) ·Hr(x,k)] / [sum Hn(x) 2 sum Hr(x,k) 2 ] 1 / 2 The sum then exceeds the count value x, and the count value frequency distributions Hn(x) and Hr(x) indicate the frequency of the various count values ​​x. In the usual calculation of the cross-correlation, k represents a displacement in the form Hr(x+k). In contrast, in the present application, the parameter k can describe a stretching or compression of the count value x. In particular, the count value x of at least one of the count value frequency distributions Hn(x) and Hr(x) can be stretched / compressed depending on the parameter k, and optionally or additionally, a stretching / compression of the frequency at which the count value x appears can be performed. The stretching / compression can be described by various formulas, for example k can be a factor (multiplied by the count value x) and k=k(x) can be dependent on x. More complex formulas, including a displacement and a multiplication of the x values, can be used to describe the stretching / compression in the calculation of the cross-correlation coefficient corr{k}.

[0036] Image capture mode The procedure according to the invention for identifying overloaded sensor elements is suitable for different image acquisition modes, in particular for various sample scanning methods and for wide-field acquisition.

[0037] In some embodiments, the sample is scanned with illumination light while the photon-counting sensor elements continuously capture counts. In this case, at least one illumination point (illumination spot) is generated on the sample, which is successively rotated over the sample area to be examined. In image scanning microscopy, often also called airy scanning, the distance between adjacent sensor elements (measured from the center of one sensor element to the center of the adjacent sensor element) is smaller than one airy (1 AU, airy unit). One airy is the size of a diffraction disk in the image plane onto which a point of the sample plane is mapped. In this case, the sensor elements are arranged in the image plane. The size of one airy is determined by the detection PSF, which may have, for example, the shape of a Gaussian function or a well-known diffraction disk function-like shape. The illuminated sample point is then mapped to a number of photon-counting sensor elements according to the detection PSF, with the central sensor elements receiving more light than the outer sensor elements. Therefore, a non-overloaded reference count frequency distribution can be formed, for example, from the counts of one or more outer sensor elements, and saturated sensor elements in the image center can then be recognized by comparing the count frequency distribution.

[0038] A similar image acquisition mode is confocal (laser) scanning microscopy, where the counts of several or all simultaneously measured sensor elements are integrated to generate a sample point signal attributable to the currently illuminated sample point.

[0039] Alternatively, wide-field illumination may be provided by directing illumination light to the sample. As in the above-mentioned Airiscan, the sensor elements may be closely grouped together such that a sample point is mapped onto a number of sensor elements. In particular, the sensor elements may be closely grouped together such that the resulting resolution of the sensor elements is at least as high as the Nyquist criterion, in particular at least twice as high. This allows a comparison of count value frequency distributions even in the case of wide-field illumination, since adjacent sensor elements measure very similar image contents and differences are mainly due to oversaturation. In this case, it may be intended to use not only one reference count value frequency distribution, but each reference count value frequency distribution for a group of adjacent sensor elements receiving similar image information.

[0040] If wide-field illumination is provided by directing illumination light to the sample, then time-series measurements can also be performed in addition to or as an alternative to the above-mentioned features. In time-series measurements, different illumination and / or detection settings or always the same illumination and / or detection settings can be performed in succession while the same sample area is illuminated, and counts for one or several count frequency distributions can be acquired in succession. Each count frequency distribution can be formed from counts of adjacent sensor elements and / or from counts acquired in succession by the same sensor element. The illumination and / or detection settings can differ, for example, in the illumination light intensity or illumination light power reaching the sample, in the filter settings that affect the light power impinging on the sensor element and / or in the sensitivity settings of the sensor element (thereby changing the probability that an impinging photon is registered). The sensor elements thereby acquire counts of different magnitudes in the time-series measurements, which are due to the same sample point but which differ due to different illumination and / or detection settings.

[0041] Tracking behavior when identifying overloaded sensor elements If one or more sensor elements are identified as being overloaded, an appropriate output may be provided, on the one hand by a warning to the user, for example that the illumination light output is too high. Additionally or alternatively, further actions may be automatically performed.

[0042] Where it is described below that an action is performed when a sensor element is identified as being overloaded, this will also include the variation that the action is implemented only if a given minimum number of sensor elements or associated count value frequency distributions are identified as being overloaded.

[0043] In particular, a result image may be calculated from the counts of the captured sensor elements, i.e. from at least a portion of the captured counts. The counts of sensor elements identified as overloaded may not be added here or may at least be mathematically corrected before being added to the calculation of the result image. In order to avoid falsification of the image grey values, it may be intended to reject all the counts captured by the sensor elements identified as overloaded and not add them to the calculation of the result image. In the case of the mathematical correction, all the counts of the sensor elements identified as overloaded may be relevant or only a portion of the counts exceeding the upper limit value (since the smaller counts of the sensor elements identified as overloaded are probably not attacked by saturation). For the mathematical correction, for example, the counts of the overloaded sensor elements may be replaced by extrapolated counts of neighboring (not overloaded) sensor elements. The extrapolation may be performed taking into account the detection PSF and in particular depending on the average count of the respective count frequency distribution. In the case where a count value is derived from the count values ​​of adjacent sensor elements, an interpolation method may be used instead of an extrapolation method.

[0044] Alternatively, if a sensor element is identified as being overloaded, the sample examination ( / image acquisition) can be repeated with modified microscope settings. Depending on the number of sensor elements identified as being overloaded, it may be intended that a mathematical correction is performed or that the sample examination is repeated with modified microscope settings without a mathematical correction. In this case, the microscope settings are changed such that the impinging light power per sensor element is reduced. For example, the light intensity of the illumination light source may be reduced and / or the zoom optics in the detection light path may be adjusted so that the beam cross section of the detection light is increased when it impinges on the sensor element.

[0045] Based on the modified microscope settings and optionally based on a given target image quality, the control unit can calculate how much to extend the exposure time if the sample inspection is repeated. The exposure time may be the pixel dwell time during the scan during which approximately the same sample point is illuminated. This allows a sufficient image quality to be achieved even if the light power falling on the sensor element is reduced.

[0046] If sensor elements are still identified as being overly loaded when the sample test is repeated, the above process of reducing the impinging light output and optionally increasing the light exposure time may be repeated until no overly loaded sensor elements are identified or until no more than a given maximum number of overly loaded sensor elements are identified.

[0047] The above-mentioned scanning method may also be implemented with multi-spot illumination. In this case, the illumination light is scanned over the sample in the form of a plurality of illumination points. The illumination points may optionally be scanned successively over the same area of ​​the sample. The detected light resulting from the various illumination points is detected solely, for example by a plurality of arrays of sensor elements as described above. These arrays may be adjacent parts of one larger array of sensor elements. The count value frequency distribution is due to the first illumination point. If an excessive load as described above is identified (i.e. for the first illumination point), it may be intended to change the microscope settings for the further illumination points so that the impinging light power is reduced for each sensor element belonging to the respective illumination point. In this way, excessive load for further illumination points can be avoided. Advantageously in this case, the identification of excessive load is already performed while the scanning of the sample with the first illumination point is still taking place.

[0048] In some embodiments, if no overloaded sensor elements are identified or if less than a given maximum number of overloaded sensor elements are identified, the image quality calculated from the counting values ​​is checked whether it reaches a given target image quality. The image quality may be calculated directly from the counting values ​​or from a result image formed from the counting values. For example, a signal-to-noise ratio (SNR) may be calculated as the image quality, the given target image quality being the target SNR. If the given target image quality is not reached, a new image acquisition is performed with modified microscope settings. The modified microscope settings result in a higher detected light intensity per sensor element or a longer light exposure time, in particular a longer pixel dwell time when scanning a sample. For example, the power of the illumination light source can be increased, filters in the light path can be adjusted accordingly, or the zoom optics can change the size of the detected PSF. The pixel dwell time can be increased by moving the illumination light more slowly over the sample. This has the same effect, allowing the illumination light to also scan the same sample area many times in succession. For the count values ​​measured in the new image capture, it is again calculated via the count value frequency distribution whether the sensor element was overloaded and whether the target image quality was achieved. If the target image quality is still not achieved, the microscope settings are changed again as described above.

[0049] In addition to the desired target image quality, the measurement should be achieved as quickly as possible. For this purpose, it may be intended to first perform a new image capture with increased illumination light intensity but without increasing the pixel dwell time. Only if it is determined during this new image capture that the sensor element is overloaded, a further image capture is performed with a reduced illumination light intensity again and an increased pixel dwell time instead. If the target image quality is not achieved in this case, a new image capture is again performed with a further increase in pixel dwell time and no further increase in illumination light intensity.

[0050] The photon counting sensor element may be formed, for example, by a PMT (photomultiplier tube) or a multi-anode PMT. Alternatively, the photon counting sensor element may be a SPAD element (single photon avalanche diode). The sample plane may be imaged and equipped with a fiber bundle that transfers the detected light to the photon counting sensor element. The SPAD element may comprise a semiconductor compound in which there is a depletion region with no free charges. A voltage above the breakdown voltage is applied to the compound. When a photon is absorbed, it may generate free charges in the depletion region, which may result in a charge avalanche that spreads across the diode and is detected as a photon count value. The photon counting sensor element therefore outputs a discrete count value that corresponds to the number of recorded photons. Several sensor elements may be formed together as a SPAD array, which may be manufactured as a photon counting array, for example in CMOS technology or a CMOS-compatible process. The sensor elements mentioned in this disclosure may be formed by one or more such arrays.

[0051] As long as saturation of the sensor element does not occur, the number of recorded photons (count value) is proportional to the number of impinging photons. The photon transfer curve then has a linear course. However, in the case of a high number of impinging photons, some photons reach the sensor element, while on the other hand the sensor element is in a dead time after the recording of the previous photon. As a result, the photon transfer curve flattens out with an increasing number of impinging photons, since the impinging photons are not recorded. Saturation may be defined by the deviation of the recorded photon number from the linear photon transfer curve by a minimum value. Thus, if the minimum value is 10%, a saturation state is called if the count value or count rate differs by more than 10% from the value predicted by the linear photon transfer curve at the currently impinging photon rate.

[0052] In the case of active suppression, i.e. active restoration of the operating voltage of the SPAD element, the photon transfer curve becomes asymptotically flat. In contrast, in the case of passive suppression, impinging photons have a dead time during which they need a longer time to restore the operating voltage. The longer dead time results in fewer photons being recorded if it saturates with an increasing number of impinging photons.

[0053] The above-mentioned sensor elements may be pixels of a common sensor. This sensor may include further sensor elements, the measured values ​​of which do not necessarily have to be processed as described for the other sensor elements. Moreover, the above-mentioned sensor elements may be part of a plurality of sensors. In particular in the case of multi-spot illumination, for example, there may be a plurality of sensors, each of which includes a respective sensor element whose measured value is determined in the described manner.

[0054] The control unit may in principle be any computing device, which may be distributed or alternatively formed by a single device or assembly. For example, the control unit may be formed by a computer, a processor, an FPGA (Field Programmable Gate Array) or another programmable assembly or may include one or more of the above-mentioned components. The one or more FPGAs or other programmable assemblies may in particular be connected to the sensor or may be part of the sensor including the sensor element.

[0055] The embodiments of the optical microscope according to the invention may be equipped to carry out variants of the described methods. Calculation and control processes, for example steps relating to the count value frequency distribution and the corresponding control of the microscope settings, may be carried out by a control unit of the optical microscope. In particular, the control unit may comprise an FPGA or another programmable assembly in which the described methods or steps of the methods are implemented. This allows a fully automated process or, alternatively, at least the method steps relating to the control of the sensor elements and the processing of the count values, in particular the features of the characterizing parts of the independent method claims, to be carried out in an automated manner. In the opposite way, the described embodiments of the optical microscope according to the invention may also be used as specified to carry out variants of the methods according to the invention. Numerical values ​​and the expressions "approximately" or "about" used in connection with numbers may be understood to include the exact relation or value and / or a deviation up to 10%.

[0056] Further advantages and features of the present invention are described below in conjunction with the accompanying schematic drawings. [Brief description of the drawings]

[0057] [Figure 1] 1 is a schematic diagram of an embodiment of an optical microscope disclosed herein; [Diagram 2] FIG. 2 is a schematic diagram of the sensor of FIG. 1. [Diagram 3] 2 is a bar graph showing the frequency distribution of count values ​​of a sensor element of the optical microscope of FIG. 1. [Figure 4] This is a bar graph calculated from the count frequency distribution of the bar graphs in Figures 3, 5, and 7. [Diagram 5] 2 is a bar graph showing the frequency distribution of count values ​​of a sensor element of the optical microscope of FIG. 1. [Figure 6] This is a bar graph calculated from the count frequency distribution of the bar graphs in Figures 3, 5, and 7. [Figure 7] 2 is a bar graph showing the frequency distribution of count values ​​of a sensor element of the optical microscope of FIG. 1. [Figure 8]This is a bar graph calculated from the count frequency distribution of the bar graphs in Figures 3, 5, and 7. [Figure 9] 1 is a flow chart of one embodiment of a method according to the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0058] Identical elements or elements having the same function are generally provided with the same reference numbers in the figures.

[0059] 1 shows an embodiment of an optical microscope 100. The optical microscope comprises a light source 10, e.g. one or more lasers, which emits illumination light 12 that is sent via a scanner 25, optional optical elements 23, 24 and an objective lens 30 to a sample plane where a sample 35 to be inspected may be located. Due to the scanning movement of the scanner 25, the illumination light 12 is sent successively along various optical paths 12A, 12B to scan the sample 35. The detection light 15, e.g. fluorescent light, emerges from the sample 35 and is sent via the objective lens 30, the optical elements 24, 23 and the scanner 25. The optical microscope 100 further comprises a beam splitter 22, which reflects the illumination light 12 and transmits the detection light 15 or vice versa, e.g. depending on the wavelength. After the beam splitter 22, the detection light 15 reaches a sensor 60 via further optional optical elements 27, 28, 29. A control unit 70 , shown diagrammatically, controls the light source 10 , the scanner 25 and the sensor 60 .

[0060] The sensor 60 is shown enlarged in FIG. 2 and includes a number of photon-counting sensor elements 61, here SPAD elements as an example. The detection light coming from the sample point illuminated according to the current scanner position is mapped to the detection / image plane as a detection light spot 16 corresponding to the detection PSF. In the example of FIG. 2, the detection light spot 16 is spread over all sensor elements 61. In principle, each sensor element 61 can continuously capture a number of counts resulting from different illuminated sample points during the scanning process. Alternatively, the sensor elements 61 can be combined into groups 65A, 65B, 65C, 65D ("binning"), and the counts of the sensor elements of the same group are combined.

[0061] The detection PSF or the detection light spot 16 is spread over several sensor elements 61, so that the sensor elements measure very similar image information that is only slightly displaced from one another. However, the counting values ​​of the sensor elements 61 are very different based on the course of the detection PSF. For example, the detection PSF may have a Gaussian curve shape, in which the central sensor elements 61 (e.g. the sensor elements of groups 65C and 65D) receive more detection light than the outer sensor elements 61 (e.g. the sensor elements of group 65A).

[0062] When a photon-counting sensor element 61 registers an impinging photon, a short dead time occurs during which this sensor element 61 cannot measure further impinging photons. High detected light intensities therefore threaten saturation / overloading of the sensor element 61. The measured count value is then no longer proportional to the number of impinging photons. If an image of the sample (=result image) / sample image is calculated from the count values, saturation of the individual sensor elements will therefore lead to erroneous image information.

[0063] According to the invention, such saturation is recognized. For this purpose, the fact that the sensor elements 61 measure substantially the same image content, but are exposed to different heights of detected light intensity, substantially based on the course of the detected PSF, may be utilized. If the count rates of the sensor elements 61 are compared and a significant difference appears, this may indicate saturation. The measures according to the invention for identifying saturated sensor elements are described in conjunction with Figs. 3 to 8.

[0064] 3 shows a bar graph of a count frequency distribution 81 showing the frequency of grey values ​​or counts x of a sensor element 61 or group of sensor elements 65A. For ease of viewing, only the larger count half of the count frequency distribution 81 is shown. The counts may be acquired at various sample points during a sample scan. The abscissa characterizes the counts x and the ordinate indicates the frequency at which each count was measured.

[0065] 3 further shows a further bar graph (reference bar graph) of a reference count frequency distribution 80, which bar graph shows the frequency of counts acquired during a sample scan by another sensor element 61 or another group of sensor elements 65B. As reference count frequency distribution 80, one count frequency distribution is used which is sure not to have saturated during the measurement or at least to have been very rarely saturated, as will be explained in more detail elsewhere.

[0066] In order to make it possible to better compare the shapes of the two count value frequency distributions 80, 81 with each other, first the analyzed count value frequency distribution 81 is rescaled. For this purpose, a fitting function 91 is fitted to the count value frequency distribution 81 and a fitting function 90 is fitted to the reference count value frequency distribution 80. The fitting functions 91, 90 are the same and in the illustrated example are f(x)=f0*exp(-x / c), where x denotes the count value. In the exponential axis notation of FIG. 3, f(x) corresponds to a line with a slope of -1 / c and an ordinate axis intercept of f0. By means of the function fitting, values ​​of the fitting parameters f0 and c for the count value frequency distribution 80 and the reference count value frequency distribution 81 are calculated. From the calculated values ​​of the fitting parameters, it is calculated how the count value frequency distribution 80 should be deformed so that, in the new function fitting, approximately the same values ​​of the fitting parameters as the reference count value frequency distribution 81 are achieved. The result of this deformation is shown in FIG. 4.

[0067] Fig. 4 shows a reference bar graph of the reference count frequency distribution 80 accompanied by the fit function 90 of Fig. 3. The analyzed count frequency distribution 81 of Fig. 3 has been modified as described, and a bar graph of this modified count frequency distribution 81B is represented in Fig. 4. The fit function 90 is approximately the same as the fit function 91B that can be calculated when fitting to the count frequency distribution 81B (calculation of the fit function 91B is not necessary, it is merely given here for better understanding). It can be seen that the two count frequency distributions 80, 81B are approximately identical except for statistical fluctuations. Accordingly, the saturation of the sensor elements did not lead to any significant influence when capturing the counts of the count frequency distribution 81.

[0068] Another example is shown in Fig. 5. This again shows a reference bar graph of the reference count frequency distribution 80 with a fit function 90 fitted to the reference count frequency distribution 80. Moreover, a further bar graph of the analyzed count frequency distribution 82, which is formed by counts of another sensor element 61 or another group 65C, is shown. The fit function 92 is fitted to the count frequency distribution 82, which is then deformed as described above. The result is shown in Fig. 6.

[0069] 6, the deviation between the reference bar graph of the reference count frequency distribution 80 and the bar graph of the modified analyzed count frequency distribution 82B can be seen, especially when the counts are very large, exceeding a value of about 150. This difference is mostly due to saturation of the sensor elements when acquiring the counts of the count frequency distribution 82. Therefore, comparing the reference count frequency distribution 80 with the count frequency distribution 82B makes this saturation clear.

[0070] Quantitatively, the comparison may be performed, for example, by calculation of a correlation coefficient R, which may be defined as Equation 3: R = covar(Hn, Hr,) / (stdev(Hn)*stdev(Hr)) In this case, Hn denotes the analyzed count frequency distribution 82B after rescaling, and Hr denotes the reference count frequency distribution 80. In the case of perfect agreement, R=1. Therefore, the limit R G can be determined, and the limit R G may be, for example, 0.95 or, more generally, between 0.92 and 0.97. <R G If so, then count value frequency distribution 82 is identified as being saturated / overloaded.

[0071] A particularly clear overload of the sensor element is illustrated in connection with Fig. 7 and Fig. 8. Fig. 7 again shows a reference bar graph of the reference count value frequency distribution 80 with the fitted fit function 90 and a bar graph of the further analyzed count value frequency distribution 83 with the fitted fit function 93. For better visibility, only the part of the count value frequency distribution 83 with larger count values ​​is shown here, and the fit function 93 can be fitted to all count values ​​of the count value frequency distribution 83, or only to the represented part of the count values, or to count values ​​larger than the represented part. A rescaling of the count value frequency distribution 83 is then performed. The result is shown in Fig. 8. The fit function 90 of the reference count value frequency distribution 80 certainly coincides with the fit function 93B of the rescaled count value frequency distribution 83B, but the differences between the reference count value frequency distribution 80 and the rescaled count value frequency distribution 83B are clearly visible, in particular for count values ​​above about 100. In this case, the calculation of the correlation coefficient R clearly shows that the limit R G Since values ​​below 0.01 are obtained, saturation of the associated sensor element is assumed.

[0072] The progress of this procedure and possible actions upon saturation are explained below in relation to figure 9. Figure 9 shows a flow chart of an embodiment of the method according to the invention.

[0073] In step S1, the control unit of the optical microscope sets the microscope settings according to user defaults, which include, for example, the choice of the objective lens, at least one spectral band for the illumination light and / or the detection light, a target image quality, such as a target SNR or target SBR (SBR: signal-to-background ratio) value, and / or parameters for resolution and scanning and an indication of the fluorescent dyes to be excited. The microscope settings set by the control unit may include, for example, the illumination light power and / or, in the case of a sample scan, the pixel dwell time. From the information about the objective lens and the dyes, the control unit can calculate the size of the detection light spot on the sensor and set the spot size on the sensor via the system zoom.

[0074] The maximum count rate per sensor element is known to the control unit, which can calculate the number of illuminated sensor elements from the calculated spot size on the sensor. This allows the control unit to calculate the average count rate f that can be expected. Photons , and the pixel dwell time T to be set can be determined. pixel According to the target SNR, especially T pixel =SNR 2 / f Photons It can be determined as follows.

[0075] With these microscope settings, a sample image is taken in step S2. For example, the sample is scanned with illumination light and the photon-counting sensor elements each detect a number of counts, so that each sensor element outputs a number of counts that belong to successively scanned sample points.

[0076] Subsequently, in step S3, the count values ​​of each sensor element or each group of sensor elements are integrated as a count value frequency distribution, which may also be called bar graph data. In this case, the invention does not require a graphical representation of the count value frequency distribution in the form of a bar graph. Instead, the count value frequency distribution may be present, for example, as a table. The graphical representation of the bar graphs in Figures 3 to 8 is optional and is intended mostly for easier comprehension.

[0077] The image of the count frequency distribution also includes an image of at least one reference count frequency distribution. The determination of the reference count frequency distribution can be performed, for example, according to the magnitude of the counts of one count frequency distribution, so that the reference count frequency distribution does not include excessively large counts, which may cause saturation, but also does not only include small counts, which may result in a small measurement accuracy or SNR, corresponding to the small recorded detected light output. To determine the reference count frequency distribution, it is also possible to carry out a first image acquisition with a relatively low illumination light intensity, followed by a further image acquisition with a higher illumination light intensity, in which the counts that form the count frequency distribution to be analyzed are acquired. Accordingly, the reference count frequency distribution and the count frequency distribution to be analyzed may be acquired simultaneously or successively.

[0078] Then, in step S4, a rescaling of the count value frequency distributions and an evaluation of the similarity are performed, as also described in figures 3 to 8. If the result is that an insufficient similarity is identified for at least one of the count value frequency distributions (or more generally for at least one given number of the count value frequency distributions), the corresponding sensor element or elements are identified as being overloaded. In principle, the method may end with an output of information to this effect to the user.

[0079] However, if in the illustrated example the similarity of the count value frequency distributions is insufficient, step S5 follows, in which the microscope settings are changed in order to avoid overloading as occurred in the previous measurement. The microscope settings are illumination and / or detection settings, which are changed so that weaker illumination light falls on the sample and / or the number of impinging photons per sensor element is reduced. For example, the illumination light power falling on the sample can be reduced by appropriately controlling the light source or the filter. The sensitivity of at least one sensor element identified as being overloaded may be reduced. Alternatively, the size of the detection light spot on the sensor can be increased by zoom optics, so that the detection light is distributed over more sensor elements. Steps S2 to S4 then follow again, i.e. a new image acquisition is performed and it is again checked in S4 whether the rescaled count value frequency distribution is similar to the reference count value frequency distribution.

[0080] If so, step S7 follows, in which it is checked whether the counts or the resulting image formed from the counts reach a given target image quality. The target image quality can be, for example, a given SNR value. The SNR value can be calculated from the photon count frequency or from the resulting image in a manner known in principle.

[0081] If the target image quality is not reached, the method proceeds to step S8, where the microscope settings are changed to provide a higher image quality for the newly captured image, for example increasing the illumination light output or pixel dwell time, followed again by steps S2 to S7.

[0082] Once it is determined in step S7 that the target image quality has been achieved, the result image formed from the last captured counts is output in step S9, with which the method ends.

[0083] In one optional embodiment, if step S5 has not been performed thus far with a reduction in illumination light output, then in step S8 the illumination output is increased, otherwise the similarity may be denied in the next execution of step S4. If in the previous execution the illumination light output was reduced in step S5, then in S8 the pixel dwell time is extended.

[0084] In an optional further embodiment, if insufficient similarity is identified in step S4, step S5 is not necessarily performed. Rather, depending on the criterion, either step S5 or step S6 is performed. The criterion for performing step S5 may be, for example, that a predetermined minimum number of count value frequency distributions are denied similarity in S4. In contrast, if fewer than a given minimum number of count value frequency distributions are denied similarity, step S6 follows. Another criterion may be that if the cycle of S2-S3-S4-S5 has been repeated a predetermined number of times, if similarity is denied in S4, S6 follows instead of S5 as the next step.

[0085] In optional step S6, the count frequency distributions for which similarity was denied in S4 are selected and replaced or modified, for example by replacing them with extrapolated counts of adjacent sensor elements, followed by step S7.

[0086] The method steps have been described diagrammatically for image acquisition by means of scanning the sample. The method steps can also be carried out for image acquisition with wide-field illumination without scanning. In the case of wide-field illumination, the scanner 25 in FIG. 1 can be omitted. Optical elements can be supplemented in FIG. 1 and the optical elements shown can in principle also be omitted. Moreover, the depicted descanned structure is merely exemplary. The illumination light and the detection light do not have to be transmitted through the same objective lens 30, in which case the beam splitter 22 can be omitted. A common scanner for the illumination light 12 and the detection light 15 can be replaced by a scanner for the illumination light 12 and a separate scanner for the detection light.

[0087] In the case of confocal measurements, a pinhole stop may be added, but this is not mandatory, since the sensor element may act as a digital pinhole stop or, optionally, an optical fiber arranged in front of the sensor element may act as a pinhole.

[0088] In the described manner, overloaded sensor elements can be efficiently identified and appropriate measures can be implemented in order to avoid overloading and to automatically perform a new image acquisition that at the same time achieves a sufficiently high image quality. [Explanation of symbols]

[0089] 10 light source 12 Illumination 12A, 12B Optical path corresponding to the current setting of the scanner 25 15 Detection light 16 Detection light spot on sensor 60 22 Beam splitter 23, 24, 27, 28, 29 Optical elements 25 Scanner 30 Objective Lens 35 Samples 60 Sensor including photon counting sensor element 61 61 Photon counting sensor element 65A-65D Group of 61 integrated sensor elements 70 Control Unit 80 Reference count frequency distribution 81-83 Frequency distribution of count values ​​to be analyzed 81B-83B Rescaled count frequency distribution of the analyzed item 90, 91-93, 91B-93B Fit functions for count frequency distribution 100 Optical Microscope S1-S9 Method Steps x is the count value of the photon-counting sensor element

Claims

1. 1. A method for imaging by optical microscopy, comprising: transmitting illumination light (12) to a sample (35); sending detection light (15) from the sample (35) to a plurality of photon counting sensor elements (61), whereby a plurality of count values ​​(x) are captured in succession by each photon counting sensor element (61); In a method comprising: forming (S3) a plurality of analyzed count frequency distributions (81-83) and at least one reference count frequency distribution (80) from said counts (x), where each analyzed count frequency distribution (81-83) indicates how frequently each of several counts occurs; calculating a similarity between each one of the analyzed count value frequency distributions (81-83) and the reference count value frequency distribution (80); - identifying an overloaded photon-counting sensor element (61) in response to the calculated similarity of the associated one or more of the analyzed count value frequency distributions (81-83); The method comprising:

2. Calculating a rescaling between each one of the analyzed count value frequency distributions (81-83) and the reference count value frequency distribution (80), 2. The method according to claim 1, characterized in that the calculation of the similarity between each one of the analyzed count value frequency distributions (81-83) and the reference count value frequency distribution (80) is performed taking into account a respective rescaling.

3. 3. The method according to claim 1 or 2, characterized by the step of forming a bar graph from the analyzed count value frequency distribution (81-83) and / or forming a reference bar graph from the reference count value frequency distribution (80), the reference count value frequency distribution (80) being formed from count values ​​of one or more photon counting sensor elements (61) that were non-saturated.

4. 4. The method according to claim 1, further comprising a step of selecting at least one count value frequency distribution as or for forming a reference count value frequency distribution (80), the selection being made depending on the magnitude of the count value (x) of the corresponding count value frequency distribution.

5. 5. The method according to claim 4, characterized in that a prerequisite for the step of selecting at least one count value frequency distribution as a reference count value frequency distribution (80) or for forming said reference count value frequency distribution (80) is that the maximum count value (x) of this count value frequency distribution or the average value formed from the maximum count values ​​(x) of this count value frequency distribution is smaller than a given upper limit value.

6. 6. The method according to claim 4 or 5, characterized in that a prerequisite for the step of selecting at least one count value frequency distribution as or for forming said reference count value frequency distribution (80) is that a signal strength criterion calculated from the count values ​​(x) of this count value frequency distribution is above a given minimum value.

7. 7. The method according to claim 1, wherein the selection of at least one count value frequency distribution for use as a reference count value frequency distribution (80) is made depending on the position of the associated photon counting sensor element or elements (61).

8. 3. The method according to claim 2, characterized in that the calculation of the rescaling between each one of the analyzed count value frequency distributions (81-83) and the reference count value frequency distribution (80) is performed by stretching or compressing the analyzed count value frequency distribution (81-83) or the reference count value frequency distribution (80).

9. The rescaling should be at least: - fitting a fit function (90-93) to the reference count frequency distribution (80) and to each of the analyzed count frequency distributions (81-83), whereby fit parameters are calculated; - stretching or compressing one of the analyzed count value frequency distributions (81-83) or the reference count value frequency distribution (80) depending on the calculated fit parameters; 3. The method of claim 2, comprising:

10. The calculation of the similarity comprises at least the following steps: calculating a correlation coefficient between one of the analyzed count value frequency distributions (81-83) and the reference count value frequency distribution (80) after rescaling between the one of the analyzed count value frequency distributions (81-83) and the reference count value frequency distribution (80), 3. The method of claim 2, wherein if the calculated correlation coefficient is below an established minimum value, the associated analyzed count value frequency distribution (81-83) is identified as being overloaded.

11. The sample (35) is scanned with the illumination light (12); The distance between adjacent photon counting sensor elements (61) is less than 1 airy; 11. The method according to claim 1, wherein the count values ​​(x) successively captured by one of the photon counting sensor elements (61) during scanning of the sample (35) are used to form the same count value frequency distribution (81-83).

12. wide-field illumination is provided by directing the illumination light (12) to the sample (35); 11. The method according to claim 1, wherein a sample plane is imaged onto the photon-counting sensor element (61) such that the resolution produced by the structure of the photon-counting sensor element (61) is at least as high as that defined by the Nyquist criterion.

13. wide-field illumination is provided by directing the illumination light (12) to the sample (35); 13. The method according to claim 1, wherein a time series measurement is performed in which count values ​​(x) for one or more count value frequency distributions (81-83) and the reference count value frequency distribution (80) are successively acquired by the photon counting sensor element (61).

14. A result image is calculated from the captured counts (x) of the photon counting sensor elements (61); 14. The method according to any one of claims 1 to 13, characterized in that the count values ​​(x) of the photon counting sensor elements (61) identified as being overloaded are not added to the calculation of the result image or are at least mathematically corrected before being used in the calculation of the result image.

15. If it is identified that the photon counting sensor element (61) is overloaded, A) image capture is repeated and the microscope settings are changed so that the output of the detected light impinging on the photon counting sensor element (61) is reduced; B) the control unit (70) calculates, based on at least the changed microscope settings, how much to extend the exposure time if the image capture is repeated; 15. The method according to claim 1, wherein if the photon counting sensor element (61) is identified as being overloaded even after repeated image capture, processes A) and B) are repeated.

16. the illumination light (12) is scanned across the sample (35) in the form of a plurality of illumination points, the illumination points successively scanning the same area of ​​the sample (35); 16. The method according to claim 1, further comprising: during scanning of the sample (35), identification of excessive loading is performed, and if excessive loading is confirmed for a first illumination point, the microscope settings for the remaining illumination points are changed so that the power of the detected light impinging on the associated photon counting sensor element (61) is reduced.

17. If it is identified that there are no overloaded photon counting sensor elements (61), or if there are at most a given maximum number of overloaded photon counting sensor elements (61), it is checked (S7) whether the image quality calculated from the count values ​​(x) reaches a given target image quality; 17. The method according to claim 1, further comprising the step of: if a given target image quality is not reached, a new image acquisition (S2) is performed with modified microscope settings, the modified microscope settings resulting in a higher detected light intensity per photon counting sensor element (61) or a longer light exposure time.

18. First, the illumination light intensity is increased and a new image is taken (S2), 18. The method according to claim 17, characterized in that if during this new image capture (S2) it is identified that the photon counting sensor element (61) is under excessive load, then instead of increasing the illumination light intensity, the exposure time is extended and a new image capture (S2) is performed again.

19. a light source (10) for emitting illumination light (12) onto a sample (35); a plurality of photon counting sensor elements (61) for sensing detection light (15) from said sample (35), each photon counting sensor element (61) being equipped to capture a plurality of count values ​​(x) in succession; In an optical microscope having forming (S3) a plurality of analyzed count frequency distributions (81-83) and at least one reference count frequency distribution (80) from said counts (x), where each analyzed count frequency distribution (81-83) indicates how frequently each of several counts occurs; calculating a similarity between each one of the analyzed count value frequency distributions (81-83) and the reference count value frequency distribution (80); - identifying an excessive load on the photon-counting sensor element (61) in response to a calculated similarity of the associated one or more of the analyzed count value frequency distributions (81-83); 2. An optical microscope comprising: a control unit (70) mounted on said optical microscope;

20. 20. The optical microscope of claim 19, wherein the control unit (70) comprises an FPGA or another programmable assembly equipped to carry out the method according to any one of claims 1 to 18.

Images