Fluorescence signal measurement method using deep learning, fluorescence lifetime measurement method and fluorescence lifetime imaging method using same, and fluorescence signal measurement device, fluorescence lifetime measurement device, and fluorescence lifetime imaging device performing same

Abstract

This fluorescence signal measurement method generates, by using a convolutional neural network (CNN) model on which adversarial training has been performed, a corrected fluorescence signal on the basis of a saturated fluorescence signal in which a saturation phenomenon has occurred. The fluorescence intensity of the saturated fluorescence signal in a first region is a threshold value, and the fluorescence intensity of the corrected fluorescence signal in a second region corresponding to the first region of the saturated fluorescence signal is greater than the threshold value. Fluorescence lifetime information is generated on the basis of the corrected fluorescence signal. A fluorescence lifetime image is generated on the basis of fluorescence lifetime information.

Description

A method for measuring fluorescence signals using deep learning, a method for measuring fluorescence lifetime and a method for imaging fluorescence lifetime using the same, and a fluorescence signal measuring device, a fluorescence lifetime measuring device, and a fluorescence lifetime imaging device for performing the same.

[0001] The present invention relates to a fluorescence signal and a method for measuring fluorescence lifetime, and more specifically, to a method for measuring a fluorescence signal using deep learning, a method for measuring fluorescence lifetime and a method for imaging fluorescence lifetime using the fluorescence signal measuring method, and a fluorescence signal measuring device, a fluorescence lifetime measuring device, and a fluorescence lifetime imaging device for performing the fluorescence signal measuring method, the fluorescence lifetime measuring method, and the fluorescence lifetime imaging method.

[0002] Fluorescence lifetime is the rate of time decay of fluorescence emitted by a fluorescence, and each fluorescence has a unique value. Fluorescence lifetime imaging (FLIm) is an imaging technique that can identify the composition and physiological state of a biological specimen with high sensitivity and specificity by measuring and imaging the fluorescence lifetime of the specimen.

[0003] Multichannel FLIm can acquire information about various components present in tissues by simultaneously measuring fluorescence lifetimes across multiple wavelength bands. To implement multichannel FLIm, the measured fluorescence is spectrally separated into channels for each wavelength band using a dichroic mirror and an optical filter. Subsequently, these spectrally separated fluorescence signals are passed through optical delay lines composed of multimode optical fibers of different lengths for each channel, thereby separating them temporally.

[0004] The time-resolved spectroscopic fluorescence signal can be detected using a single photoamplifier. While this method is preferred due to its simplicity compared to other approaches, it presents a challenge in achieving a high signal-to-noise ratio (SNR) for all spectroscopic channels due to issues such as highly variable fluorescence efficiencies among various phosphors and variations in the sensitivity of the photoamplifier and photodetector modules across the wavelength spectrum.

[0005] In other words, if the system is tuned to a channel with high fluorescence efficiency, image information with a slightly reduced SNR is inevitably acquired in channels with relatively lower efficiency. Conversely, if the system is tuned to a channel with low efficiency, fluorescence intensities exceeding the threshold are not properly measured due to saturation in channels with relatively high efficiency, leading to the problem of being recorded at the threshold value. While it is possible to obtain optimized images across all channels by appropriately adjusting fluorescence intensity through repetitive imaging, this approach may be limited in biological tissues where repetitive imaging is frequently impossible. Various studies have been proposed to measure fluorescence signals with high accuracy, such as configuring hardware-based wavelength-band multiple detectors, obtaining fluorescence images with a wide dynamic range by controlling multiple exposures, and restoring fluorescence signals lost due to saturation through mathematical modeling. However, these methods have limitations, such as the complexity of the measurement device and the fact that they can only be operated on specific systems.

[0006] A relevant prior art document is Korean Registered Patent No. 10-2527241.

[0007] One objective of the present invention is to provide a fluorescence signal measurement method that corrects a fluorescence signal saturated to a threshold value using deep learning.

[0008] One objective of the present invention is to provide a fluorescence lifetime measurement method and a fluorescence lifetime imaging method using the above-described fluorescence signal measurement method.

[0009] One objective of the present invention is to provide a fluorescence signal measuring device for performing the fluorescence signal measuring method, a fluorescence lifetime measuring device for performing the fluorescence lifetime measuring method, and a fluorescence lifetime imaging device for performing the fluorescence lifetime imaging method.

[0010] To achieve the above objective, in a method for measuring a fluorescence signal according to embodiments of the present invention, a signal generation unit generates a saturated fluorescence signal in which the fluorescence intensity in a first region is a threshold value, based on fluorescence photons generated by irradiating a sample with excitation light. A signal correction unit comprising a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator generates a corrected fluorescence signal based on the saturated fluorescence signal. The fluorescence intensity in a second region of the corrected fluorescence signal, corresponding to the first region of the saturated fluorescence signal, is greater than the threshold value, and the threshold value is changed according to the detection characteristics for the fluorescence photons.

[0011] In one embodiment, the signal correction unit may further include an input image processing unit. In generating the correction fluorescence signal, the input image processing unit may generate a saturated fluorescence signal image based on the saturated fluorescence signal, and the convolutional neural network model may generate a correction fluorescence signal image based on the saturated fluorescence signal image. The correction fluorescence signal image may be output as the correction fluorescence signal.

[0012] In one embodiment, in generating the correction fluorescence signal, the input image processing unit may further include the step of generating a saturated location indicator image based on the saturated fluorescence signal. The convolutional neural network model may generate the correction fluorescence signal image based on the saturated fluorescence signal image and the saturated location indicator image. The saturated location indicator image may indicate a location where the fluorescence intensity of the saturated fluorescence signal image is the threshold value.

[0013] In one embodiment, the signal generating unit may include a dichroic mirror. In generating the saturated fluorescence signal, the step of the fluorescence photon passing through the dichroic mirror may be included.

[0014] In one embodiment, the signal generating unit may include a plurality of dichroic mirrors. In generating the saturated fluorescence signal, the step of the fluorescence photon passing through the plurality of dichroic mirrors may be included.

[0015] In one embodiment, the convolutional neural network model may include a residual dense network model. The residual dense network model may include a plurality of computational layers. The plurality of computational layers may include a plurality of concatenation layers, a plurality of convolution layers, and a plurality of leaky RELU (rectified linear unit) layers.

[0016] In one embodiment, the discriminator may be a Markovian discriminator. The Markovian discriminator may include a plurality of convolution layers, a plurality of spectral normalization layers, and a plurality of leaky RELU (rectified linear unit) layers.

[0017] In one embodiment, the method may further include the step of performing training on the convolutional neural network based on saturated fluorescence signal data. The saturated fluorescence signal data may be obtained by performing random rescaling on the fluorescence signal data.

[0018] To achieve the above objective, in a method for measuring fluorescence lifetime according to embodiments of the present invention, a saturated fluorescence signal is generated in which the fluorescence intensity in a first region is a threshold value, based on fluorescence photons generated by irradiating a sample with excitation light. A signal correction unit comprising a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator generates a corrected fluorescence signal based on the saturated fluorescence signal. A signal processing unit generates fluorescence lifetime information based on the corrected fluorescence signal. The fluorescence intensity in a second region of the corrected fluorescence signal, corresponding to the first region of the saturated fluorescence signal, is greater than the threshold value. The threshold value is changed according to the detection characteristics for the fluorescence photons.

[0019] To achieve the above objective, in a fluorescence lifetime imaging method according to embodiments of the present invention, a signal generation unit generates a saturated fluorescence signal in which the fluorescence intensity in a first region is a threshold value based on fluorescence photons generated by irradiating a sample with excitation light. A signal correction unit including a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator generates a corrected fluorescence signal based on the saturated fluorescence signal. A signal processing unit generates fluorescence lifetime information based on the corrected fluorescence signal. A fluorescence lifetime image processing unit generates a fluorescence lifetime image based on the fluorescence lifetime information. The fluorescence intensity in the second region of the corrected fluorescence signal, corresponding to the first region of the saturated fluorescence signal, is greater than the threshold value, and the threshold value is changed according to the detection characteristics for the fluorescence photons. The fluorescence lifetime image is a two-dimensional image, the horizontal axis of the fluorescence lifetime image is information on the translational distance for the collection of the fluorescence photons, and the vertical axis of the fluorescence lifetime image is information on the rotational angle for the collection of the fluorescence photons.

[0020] To achieve the above objective, a fluorescence signal measuring device according to embodiments of the present invention includes a signal generation unit and a signal correction unit. The signal generation unit generates a saturated fluorescence signal in which the fluorescence intensity in a first region is a threshold value. The signal correction unit includes a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator, and generates a corrected fluorescence signal based on the saturated fluorescence signal. The fluorescence intensity in a second region of the corrected fluorescence signal, corresponding to the first region of the saturated fluorescence signal, is greater than the threshold value, and the signal generation unit includes a light generation unit that generates excitation light irradiated onto a sample and a detection unit that generates the saturated fluorescence signal based on fluorescence photons generated by irradiating the sample with the excitation light. The threshold value is changed according to the characteristics of the detection unit.

[0021] To achieve the above objective, a fluorescence lifetime measurement device according to embodiments of the present invention includes a signal generation unit, a signal correction unit, and a signal processing unit. The signal generation unit generates a saturated fluorescence signal in which the fluorescence intensity in a first region is a threshold value. The signal correction unit includes a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator, and generates a corrected fluorescence signal based on the saturated fluorescence signal. The signal processing unit generates fluorescence lifetime information based on the corrected fluorescence signal. The fluorescence intensity in a second region of the corrected fluorescence signal corresponding to the first region of the saturated fluorescence signal is greater than the threshold value, and the signal generation unit includes a light generation unit that generates excitation light irradiated onto a sample and a detection unit that generates the saturated fluorescence signal based on fluorescence photons generated by irradiating the sample with the excitation light. The threshold value is changed according to the characteristics of the detection unit.

[0022] To achieve the above objective, a fluorescence lifetime imaging device according to embodiments of the present invention comprises a signal generation unit, a signal correction unit, a signal processing unit, and a fluorescence lifetime image processing unit. The signal generation unit generates a saturated fluorescence signal in which the fluorescence intensity in a first region is a threshold value. The signal correction unit includes a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator, and generates a corrected fluorescence signal based on the saturated fluorescence signal. The signal processing unit generates fluorescence lifetime information based on the corrected fluorescence signal. The fluorescence lifetime image processing unit generates a fluorescence lifetime image based on the fluorescence lifetime information. The fluorescence intensity in a second region of the corrected fluorescence signal corresponding to the first region of the saturated fluorescence signal is greater than the threshold value. The signal generation unit includes a light generation unit that generates excitation light to be irradiated onto a sample, a collection unit capable of translation and rotational movement that collects fluorescence photons generated by irradiating the excitation light onto the sample, and a detection unit that generates the saturated fluorescence signal based on the fluorescence photons. The threshold value is changed according to the characteristics of the detection unit. The above fluorescence lifetime image is a two-dimensional image, the horizontal axis of the above fluorescence lifetime image is information on the parallel translation distance of the collection unit, and the vertical axis of the above fluorescence lifetime image is information on the rotational translation angle of the collection unit.

[0023] In the fluorescence signal measurement method using deep learning, the fluorescence lifetime measurement method using the same, the fluorescence lifetime imaging method, and the fluorescence signal measurement device, fluorescence lifetime measurement device, and fluorescence lifetime imaging device performing the same, according to the embodiments of the present invention as described above, a corrected fluorescence signal can be obtained based on a saturated fluorescence signal based on a convolutional neural network model in which adversarial learning has been performed. The accuracy of the fluorescence lifetime can be improved through the corrected fluorescence signal. Fluorescence signals and fluorescence lifetime images having a wide dynamic range can be obtained without separate hardware setup. For example, when imaging lesion tissue for pathological research and diagnosis, the accuracy of information regarding lesion tissue containing various tissue components can be improved. Therefore, as it is trained based on deep learning, the scope of universal application can be expanded to various fluorescence lifetime imaging systems without being limited to a specific system.

[0024] FIG. 1 is a flowchart illustrating a method for measuring a fluorescence signal according to embodiments of the present invention.

[0025] FIG. 2a is a block diagram showing a fluorescence signal measuring device according to embodiments of the present invention.

[0026] Figure 2b is a diagram showing an example of a saturated fluorescence signal of Figure 1.

[0027] Figure 2c is a diagram showing an example of a corrected fluorescence signal of Figure 1.

[0028] FIGS. 3 and 4 are block diagrams showing examples of signal correction units included in the fluorescence signal measuring device of FIG. 2a.

[0029] FIG. 5 is a block diagram showing an example of a convolutional neural network (CNN) model learning unit according to embodiments of the present invention.

[0030] FIGS. 6 and 7 are block diagrams showing examples of signal generation units included in the fluorescence signal measuring device of FIG. 2a.

[0031] FIGS. 8 and 9 are block diagrams showing a signal generation unit according to embodiments of the present invention.

[0032] FIG. 10 is a flowchart illustrating a method for measuring fluorescence lifetime according to embodiments of the present invention.

[0033] FIG. 11 is a block diagram showing a fluorescence lifetime measuring device according to embodiments of the present invention.

[0034] FIG. 12 is a flowchart illustrating a fluorescence lifetime imaging method according to embodiments of the present invention.

[0035] FIG. 13 is a block diagram showing a fluorescence lifetime imaging device according to embodiments of the present invention.

[0036] FIG. 14 is a block diagram showing an example of a convolutional neural network model according to embodiments of the present invention.

[0037] FIG. 15 is a block diagram showing an example of a residual dense block of FIG. 14.

[0038] FIG. 16 is a block diagram showing an example of the discriminator of FIG. 5.

[0039] FIG. 17 is a diagram illustrating the operation of generating saturated fluorescence signal data used for training the convolutional neural network model of FIG. 5.

[0040] FIG. 18 is a diagram illustrating the operation of generating a saturated fluorescence signal image based on a saturated fluorescence signal in the input image processing unit of FIG. 3.

[0041] FIGS. 19a, 19b, 19c, 19d, 20a, 20b, 21, 22a, 22b, 23a and 23b are drawings illustrating the performance of a fluorescence signal measurement method according to embodiments of the present invention.

[0042] FIG. 24 is a hardware configuration diagram of a fluorescence signal measuring device according to embodiments of the present invention.

[0043] Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the attached drawings. Identical components in the drawings are given the same reference numerals, and redundant descriptions of identical components are omitted.

[0044] FIG. 1 is a flowchart illustrating a method for measuring a fluorescence signal according to embodiments of the present invention.

[0045] Referring to FIG. 1, a fluorescence signal measurement method according to embodiments of the present invention is performed by a fluorescence signal measurement device. The structure of the fluorescence signal measurement device will be described later with reference to FIG. 2a, etc.

[0046] In a method for measuring a fluorescence signal according to embodiments of the present invention, a saturated fluorescence signal (e.g., SSIG in FIG. 2a) is generated in which the fluorescence intensity in a first region (e.g., REG_1 in FIG. 2b) is a threshold value (e.g., THV in FIG. 2b) (step S100). For example, a signal generating unit (1100 in FIG. 2a) can generate a saturated fluorescence signal (SSIG in FIG. 2a) based on fluorescence photons generated by irradiating excitation light onto a sample (1120 in FIG. 2a).

[0047] Subsequently, a compensated fluorescence signal (e.g., CSIG in FIG. 2a) is generated in which the fluorescence intensity in the second region (e.g., REG_2 in FIG. 2c) is greater than the threshold value (THV in FIG. 2b) (step S200). For example, the signal compensation unit (1200 in FIG. 2a) may include a convolutional neural network (CNN) model (1210 in FIG. 2a) in which adversarial learning is performed using a discriminator (e.g., 1230 in FIG. 5), and may generate a compensated fluorescence signal (CSIG in FIG. 2a) based on a saturated fluorescence signal (SSIG in FIG. 2a).

[0048] In one embodiment, the signal generating unit (1100 in FIG. 2a) may include a sensing unit (1130 in FIG. 2a), and the threshold value (THV in FIG. 2b) may be changed according to the sensing characteristics for fluorescent photons.

[0049] FIG. 2a is a block diagram showing a fluorescence signal measuring device according to embodiments of the present invention.

[0050] Referring to FIG. 2a, the fluorescent signal measuring device (1000) includes a signal generating unit (1100) and a signal correction unit (1200).

[0051] The signal generation unit (1100) generates a saturated fluorescence signal (SSIG) in which the fluorescence intensity in the first region (REG_1 in FIG. 2b) is a threshold value (e.g., THV in FIG. 2b). The signal generation unit (1100) includes a light generation unit (1110), a sample (1120), and a detection unit (1130).

[0052] The light generating unit (1100) is a module that generates excitation light to be irradiated onto a sample (1120) containing fluorescent molecules, and can generate pulse-shaped excitation light.

[0053] The detection unit (1130) receives fluorescent photons generated by excitation light being irradiated onto the sample (1120) and can convert the received fluorescent photons into an electrical signal. At this time, the magnification of the signal (e.g., amplification) can be adjusted. For example, a photo-multiplier tube (PMT) or an avalanche photo diode (APD) can be used as the detection unit (1130). Depending on the characteristics of the detection unit (1130), the measurable fluorescence intensity has a threshold value (THV in FIG. 2b). Fluorescence intensity above the threshold value (THV in FIG. 2b) causes a saturation phenomenon in which the detection unit (1130) measures the threshold value (THV in FIG. 2b). Due to this saturation phenomenon, when the fluorescence intensity is above the threshold value (THV in FIG. 2b), the detection unit (1130) cannot generate a reference fluorescent signal in which saturation does not occur. The detection unit (1130) can generate a saturated fluorescence signal (SSIG) in which the fluorescence intensity in the first region (REG_1 in FIG. 2b) is a threshold value (THV in FIG. 2b) based on fluorescence photons generated by irradiating the sample (1120) with excitation light.

[0054] The signal correction unit (1200) generates a corrected fluorescence signal (CSIG) in which the fluorescence intensity in the second region (REG_2 in FIG. 2c) is greater than the threshold value (THV in FIG. 2b). The signal correction unit (1200) includes a convolutional neural network (CNN) model (1210). Adversarial learning is performed in the convolutional neural network model (1210) using a discriminator (1230 in FIG. 5). The learning of the convolutional neural network model (1210) will be described later with reference to FIG. 5, etc., and the structure of the convolutional neural network model (1210) will be described later with reference to FIG. 14. The signal correction unit (1200) can generate a corrected fluorescence signal (CSIG) based on a saturated fluorescence signal (SSIG). The fluorescence intensity in the second region of the corrected fluorescence signal (REG_2 in Fig. 2c) corresponding to the first region of the saturated fluorescence signal (SSIG) (REG_1 in Fig. 2b) may be greater than the threshold value (THV in Fig. 2b).

[0055] Figure 2b is a diagram showing an example of a saturated fluorescence signal of Figure 1.

[0056] Referring to FIG. 2b, the fluorescence intensity in the first region (REG_1) of the saturated fluorescence signal (e.g., SSIG in FIG. 2a) may be at a threshold value (THV). For example, if the fluorescence intensity in the reference fluorescence signal is greater than or equal to the threshold value (THV), a saturation phenomenon may occur in which the fluorescence signal is measured at the threshold value (THV) in the first region (REG_1).

[0057] Figure 2c is a diagram showing an example of a corrected fluorescence signal of Figure 1.

[0058] Referring to FIG. 2c, the fluorescence intensity in the second region (REG_2) of the corrected fluorescence signal (e.g., CSIG in FIG. 2a) may be equal to or greater than the threshold value (THV). For example, the second region (REG_2) of the corrected fluorescence signal (CSIG in FIG. 2a) may correspond to the first region (REG_1) of the saturated fluorescence signal (SSIG in FIG. 2a).

[0059] FIGS. 3 and 4 are block diagrams showing examples of signal correction units included in the fluorescence signal measuring device of FIG. 2a. Descriptions that overlap with FIG. 2a are omitted.

[0060] Referring to FIG. 3, the signal correction unit (1201) may include an input image processing unit (1220) and a convolutional neural network model (1211).

[0061] The input image processing unit (1220) can generate a saturated fluorescence signal image (SSIG_IM) based on the saturated fluorescence signal (SSIG). For example, the saturated fluorescence signal (SSIG) can be merged by bundle to generate the saturated fluorescence signal image (SSIG_IM). The horizontal axis of the saturated fluorescence signal image (SSIG_IM) represents the number of bundles, the vertical axis represents time information, and the color information represents fluorescence intensity.

[0062] A convolutional neural network model (1211) can generate a corrected fluorescence signal image (CSIG_IM) based on a saturated fluorescence signal image (SSIG_IM). For example, the horizontal axis of the corrected fluorescence signal image (CSIG_IM) may represent the number of bundles, the vertical axis may represent time information, and the color information may represent fluorescence intensity. In the example of FIG. 3, the corrected fluorescence signal image (CSIG_IM) may be output as a corrected fluorescence signal (CSIG).

[0063] Referring to FIG. 4, the signal correction unit (1202) may include an input image processing unit (1221) and a convolutional neural network model (1212). Descriptions that overlap with FIG. 3 are omitted below.

[0064] The input image processing unit (1221) can generate a saturated fluorescence signal image (SSIG_IM) and a saturated location indicator image (SSIG_MA) based on the saturated fluorescence signal (SSIG). The saturated location indicator image (SSIG_MA) indicates a location where the fluorescence intensity of the saturated fluorescence signal image (SSIG_IM) is at a threshold value (THV in FIG. 2b). For example, the saturated location indicator image (SSIG_MA) may have data at a location where the fluorescence intensity is at a threshold value (THV in FIG. 2b) as first location data, and data at a location where the fluorescence intensity is not at a threshold value (THV in FIG. 2b) as second location data. The color at the location where the fluorescence intensity is at a threshold value (THV in FIG. 2b) may be a first color (e.g., white), and the color at the location where the fluorescence intensity is less than the threshold value (THV in FIG. 2b) may be a second color (e.g., black).

[0065] The convolutional neural network model (1212) can generate a corrected fluorescence signal image (CSIG_IM) based on the saturated fluorescence signal image (SSIG_IM) and the saturated position mark image (SSIG_MA). In the example of FIG. 4, the corrected fluorescence signal image (CSIG_IM) can be output as a corrected fluorescence signal (CSIG).

[0066] FIG. 5 is a block diagram illustrating an example of a convolutional neural network model learning unit according to embodiments of the present invention. Descriptions that overlap with FIG. 3 and 4 are omitted below.

[0067] Referring to FIG. 5, the convolutional neural network model learning unit (1500) includes an input image processing unit (1222), a convolutional neural network model (1213), and a discriminator (1230). The convolutional neural network model learning unit (1500) performs adversarial learning on the convolutional neural network model (1213).

[0068] In one embodiment, the discriminator (1230) may receive a saturated fluorescence signal image (SSIG_IM) and a saturated position indicator image (SSIG_MA) generated by the input image processing unit (1222), and may receive a corrected fluorescence signal image (CSIG_IM) generated by the convolutional neural network model (1213).

[0069] In one embodiment, the discriminator (1230) may include a Markovian discriminator. The structure of the Markovian discriminator will be described later with reference to FIG. 16.

[0070] FIGS. 6 and 7 are block diagrams showing examples of signal generation units included in the fluorescence signal measuring device of FIG. 2a. Descriptions that overlap with FIG. 2a are omitted.

[0071] Referring to FIG. 6, the signal generating unit (1101) may include a light generating unit (1110), a sample (1120), a dichroic mirror (1140), and a sensing unit (1131).

[0072] The dichroic mirror (1140) can reflect or pass fluorescent photons according to wavelength. Fluorescent photons reflected by the dichroic mirror (1140) are input to the detection unit (1131) via the first channel (CH1), and fluorescent photons that pass through the dichroic mirror (1140) are input to the detection unit (1131) via the second channel (CH2). The fluorescent signals spectrally separated into wavelength-specific channels can also be separated temporally by passing through optical delay lines composed of multimode optical fibers of different lengths for each channel.

[0073] Referring to FIG. 7, the signal generating unit (1102) may include a light generating unit (1110), a sample (1120), first and second dichroic mirrors (1140, 1141) and a sensing unit (1132).

[0074] Fluorescent photons reflected by the first dichroic mirror (1140) are input to the detection unit (1132) via the first channel (CH1), and fluorescent photons that pass through the first dichroic mirror (1140) are input to the second dichroic mirror (1141). Fluorescent photons reflected by the second dichroic mirror are input to the detection unit (1132) via the second channel (CH2), and fluorescent photons that pass through the second dichroic mirror (1141) are input to the detection unit (1132) via the third channel (CH3).

[0075] Meanwhile, although the above-described case in which the signal generating unit (1102) includes first and second dichroic mirrors (1140, 1141) has been exemplified, the present invention is not limited thereto. For example, the signal generating unit (1102) may include three or more dichroic mirrors and multiple channels.

[0076] FIGS. 8 and 9 are block diagrams illustrating a signal generation unit according to embodiments of the present invention. Descriptions that overlap with FIG. 2a are omitted.

[0077] Referring to FIG. 8, the signal generation unit (1103) may include a light generation unit (1110), a sample (1120), a collection unit (1150), and a detection unit (1133).

[0078] The collection unit (1150) can collect a number of fluorescent photons and excitation light generated by irradiating the sample (1120) with excitation light, and can be moved in parallel and rotated. The detection unit (1133) can generate a saturated fluorescent signal (SSIG) based on the fluorescent photons collected by the collection unit (1150).

[0079] Referring to FIG. 9, the signal generation unit (1104) may include a light generation unit (1110), a sample (1121), a collection unit (1151), and a detection unit (1133).

[0080] The sample (1121) may be a cardiovascular vessel, and the collecting unit (1151) may irradiate light onto the tissue and acquire the reflected light.

[0081] In one embodiment, the collecting portion (1151) may be a probe or catheter that can be inserted into a blood vessel and may be combined with a rotary joint device.

[0082] In another embodiment, the collection unit (1151) may be combined with an optical coherence tomography (OCT) device. The optical coherence tomography device images the tissue using light collected from the collection unit (1151), and can image the tissue by interfering the light reflected back from the mirror of the reference arm with the light scattered back from the tissue and using the interference signal. Accordingly, the information imaged through the optical coherence tomography device may include morphological information of the vascular tissue. The original optical coherence tomography image obtained through the collection unit (1151) is a 3D image of the Z-axis (parallel movement direction of the collection unit), the angle axis (θ-axis, rotational movement direction of the collection unit), and the radiation axis (r, depth direction), and can create a 2D maximum projection mapping image or an OCT polar coordinate image by extracting only the maximum brightness values ​​for each θ from the original 3D image. The horizontal axis of the 2D maximum projection mapping image may be an image in which the horizontal axis is the translation distance information (Z) of the collection unit (1151) and the vertical axis is the rotational translation angle information (θ) of the collection unit. The OCT polar coordinate image may be an image in a polar coordinate domain in which the horizontal axis is the angle axis and the vertical axis is the radiation axis. In the description, the image coordinate system is not distinguished and is simply referred to as an optical coherence tomography image.

[0083] FIG. 10 is a flowchart illustrating a method for measuring fluorescence lifetime according to embodiments of the present invention.

[0084] Referring to FIG. 10, the fluorescence lifetime measurement method according to embodiments of the present invention is performed by a fluorescence lifetime measurement device. The structure of the fluorescence lifetime measurement device will be described later with reference to FIG. 11.

[0085] In the fluorescence lifetime measurement method according to embodiments of the present invention, steps S100 and S200 may be substantially the same as those described above with reference to FIG. 1.

[0086] Subsequently, fluorescence lifetime information (e.g., FL_INF of FIG. 11) is generated using the corrected fluorescence signal (e.g., CSIG of FIG. 11) (step S300). For example, the fluorescence lifetime information (FL_INF of FIG. 11) can be generated using the least squares method for a function of the corrected fluorescence signal (CSIG of FIG. 11).

[0087] FIG. 11 is a block diagram showing a fluorescence lifetime measuring device according to embodiments of the present invention. Descriptions that overlap with FIG. 2a are omitted.

[0088] Referring to FIG. 11, the fluorescence lifetime measuring device (1001) includes a signal generation unit (1100), a signal correction unit (1200), and a signal processing unit (1300).

[0089] The signal processing unit (1300) can generate fluorescence lifetime information (FL_INF) based on the corrected fluorescence signal (CSIG) generated by the signal correction unit (1200).

[0090] In one embodiment, the signal processing unit (1300) can generate fluorescence lifetime information (FL_INF) through an analog mean delay method.

[0091] In one embodiment, the signal processing unit (1300) can deconvolve a fluorescence signal with a pre-measured instrumental response function (IRF) to eliminate characteristic effects of the fluorescence lifetime measuring device (1001), and generate fluorescence lifetime information (FL_INF) using the average delay time of the deconvolved fluorescence signal.

[0092] FIG. 12 is a flowchart illustrating a fluorescence lifetime imaging method according to embodiments of the present invention.

[0093] Referring to FIG. 12, the fluorescence lifetime measurement method according to embodiments of the present invention is performed by a fluorescence lifetime measurement device. The structure of the fluorescence lifetime measurement device will be described later with reference to FIG. 13.

[0094] In the fluorescence lifetime imaging method according to embodiments of the present invention, steps S100, S200, and S300 may be substantially the same as those described above with reference to FIGS. 1 and 10.

[0095] Subsequently, a fluorescence lifetime image (e.g., FL_IMG of FIG. 13) is generated using fluorescence lifetime information (e.g., FL_INF of FIG. 13) (step S400). For example, the fluorescence lifetime image (FL_IMG of FIG. 13) may be a two-dimensional image. The horizontal axis of the fluorescence lifetime image (FL_IMG of FIG. 13) may represent the translation distance (Z) information of the collection unit (1150 of FIG. 13), and the vertical axis of the fluorescence lifetime image may represent the rotation angle (θ) information of the collection unit (1150 of FIG. 13).

[0096] FIG. 13 is a block diagram showing a fluorescence lifetime imaging device according to embodiments of the present invention. Descriptions that overlap with FIG. 2a and 11 are omitted.

[0097] Referring to FIG. 13, the fluorescence lifetime imaging device (1002) includes a signal generation unit (1103), a signal correction unit (1200), a signal processing unit (1300), and a fluorescence lifetime image processing unit (1400).

[0098] The fluorescence lifetime image processing unit (1400) can generate a fluorescence lifetime image (FL_IMG) based on fluorescence lifetime information (FL_INF).

[0099] In one embodiment, the fluorescence lifetime image processing unit (1400) can obtain biochemical information of tissue components from the fluorescence lifetime image (FL_IMG) because it utilizes the fact that each molecule or substance has a unique time of emitting light (fluorescence lifetime) in response to excitation light.

[0100] In another embodiment, the fluorescence lifetime image processing unit (1400) can generate a fluorescence lifetime image (FL_IMG) of a plurality of channels (e.g., CH1, CH2, CH3 of FIG. 7) mapped to excitation lights having different wavelengths. For example, the fluorescence lifetime image (FL_IMG) may be composed of a first channel image containing collagen information, a second channel image containing elastin and macrophage information, and a third channel image containing lipid information. The fluorescence lifetime image (FL_IMG) is a two-dimensional image, and the horizontal axis of the fluorescence lifetime image (FL_IMG) may represent the parallel translation distance (Z) information of the collection unit (1150), and the vertical axis of the fluorescence lifetime image (FL_IMG) may represent the rotational translation angle (θ) information of the collection unit (1150).

[0101] FIG. 14 is a block diagram showing an example of a convolutional neural network model according to embodiments of the present invention.

[0102] Referring to FIG. 14, the convolutional neural network model (e.g., 1210 in FIG. 13) may include a residual dense network model. In other words, the convolutional neural network model (1210) may be implemented in the form of a residual dense network model.

[0103] A residual dense network model may include multiple computational layers, and the multiple computational layers may include multiple concatenation layers, multiple convolution layers, and multiple leaky RELU (rectified linear unit) layers. A residual dense network model may include multiple residual dense blocks.

[0104] In one embodiment, the convolutional neural network model receives input by performing concatenation on the saturated fluorescence signal image (SSIG_IM) and the saturated localization image (SSIG_MA). The data output from the convolutional layer is input to the next convolutional layer, and the data output from the next convolutional layer can be input to the residual density block.

[0105] FIG. 15 is a block diagram showing an example of a residual dense block of FIG. 14.

[0106] The residual density block may include multiple computation layers. The multiple computation layers may include multiple concatenation layers, multiple convolution layers, and multiple leaky RELU (rectified linear unit) layers.

[0107] In one embodiment, data input to a convolutional layer undergoes a Leaky RELU layer operation, and the output data can be input to the next convolutional layer through concatenation with the input data. This structure can be repeated in a residual density block.

[0108] Figure 16 is a block diagram showing an example of the discriminator of Figure 5.

[0109] Referring to FIG. 16, the discriminator (1230) may include a Markov discriminator. The Markov discriminator may include a plurality of convolution layers, a plurality of spectral normalization layers, and a plurality of leaky RELU (rectified linear unit) layers.

[0110] In one embodiment, the discriminator (e.g., 1230 in FIG. 5) may include a loss function operation unit (1240). The loss function is used to perform adversarial learning of a convolutional neural network model (e.g., 1210 in FIG. 13). The loss function may include a hinge generative adversarial network (GAN) loss function, an L1 loss function, and a loss function that forces the fluorescence intensity of a corrected fluorescence signal (e.g., CSIG in FIG. 13) to be greater than the fluorescence intensity of a saturated fluorescence signal (e.g., SSIG in FIG. 13).

[0111] FIG. 17 is a diagram illustrating the operation of generating saturated fluorescence signal data used for training the convolutional neural network model of FIG. 5.

[0112] Referring to Fig. 17, the first graph represents reference fluorescence signal data in which no saturation occurs. Random scaling is performed on the reference fluorescence signal data by channel (CH1, CH2, CH3). The second graph represents the fluorescence signal data after random scaling is performed based on the reference fluorescence signal data. Subsequently, when a threshold value (THV) is set, fluorescence intensities above the threshold value (THV) are changed to the threshold value (THV), causing saturation and generating saturated fluorescence signal data. The third graph represents saturated fluorescence signal data in which fluorescence intensity is limited to the threshold value (THV).

[0113] In one embodiment, the reference fluorescence signal data used for training a convolutional neural network model (1210 in FIG. 13) may be a multi-channel fluorescence signal of a human coronary artery and atherosclerotic lesion acquired using a catheter system combined with an optical coherence tomography (OCT) device and a fluorescence lifetime imaging (FLIm) device (e.g., 1002 in FIG. 13).

[0114] Meanwhile, although the reference fluorescence signal data used to train the convolutional neural network model (1210 in FIG. 13) has been exemplified as generating saturated fluorescence signal data with a single threshold value (THV), the present invention is not limited thereto. For example, saturated fluorescence signal data can be generated by setting a first threshold value (THV_1) and a second threshold value (THV_2) as the threshold values ​​(THV). Additionally, saturated fluorescence signal data can be generated by setting three or more threshold values ​​(THV).

[0115] FIG. 18 is a diagram illustrating the operation of generating a saturated fluorescence signal image based on a saturated fluorescence signal in the input image processing unit of FIG. 3.

[0116] Referring to Fig. 18, a saturated fluorescence signal (e.g., SSIG in Fig. 3) can be merged by bundle to generate a saturated fluorescence signal image (e.g., SSIG_IM in Fig. 3). The horizontal axis of the saturated fluorescence signal (SSIG in Fig. 3) represents time information, and the vertical axis represents fluorescence intensity. The horizontal axis of the saturated fluorescence signal image (SSIG_IM in Fig. 3) represents the number of bundles, the vertical axis represents time information, and the color information represents fluorescence intensity. In the example of Fig. 18, the color information of the saturated fluorescence signal image (SSIG_IM in Fig. 3) where saturation occurs can be displayed in black.

[0117] FIGS. 19a, 19b, 19c, 19d, 20a, 20b, 21, 22a, 22b, 23a and 23b are drawings illustrating the performance of a fluorescence signal measurement method according to embodiments of the present invention.

[0118] Referring to Fig. 19a, the average square root error of the fluorescence intensity of the saturated fluorescence signal before correction (SSIG in Fig. 13) and the corrected fluorescence signal after correction (CSIG in Fig. 13) is compared with the fluorescence intensity of the reference fluorescence signal according to the degree of saturation.

[0119] Referring to Fig. 19b, the relative measurement error of the fluorescence lifetime measured based on the saturated fluorescence signal before correction (SSIG in Fig. 13) and the corrected fluorescence signal after correction (CSIG in Fig. 13) and the fluorescence lifetime measured based on the reference fluorescence signal is shown according to the degree of saturation.

[0120] Referring to Figures 19c and 19d, when using Coumarin 120 ethanol solution and Rhodamine 6G ethanol solution, the fluorescence lifetime measured based on the saturated fluorescence signal before correction (SSIG in Figure 13) and the corrected fluorescence signal after correction (CSIG in Figure 13) is shown according to the degree of saturation.

[0121] Referring to FIG. 20a, the frame-by-frame average fluorescence lifetime of the saturated fluorescence signal before correction (SSIG in FIG. 13), the corrected fluorescence signal after correction (CSIG in FIG. 13), and the reference fluorescence signal is shown according to the parallel translation direction of the collection unit (1150 in FIG. 13).

[0122] Referring to Fig. 20b, the relative frequencies of the saturated fluorescence signal before correction (SSIG in Fig. 13), the corrected fluorescence signal after correction (CSIG in Fig. 13), and the reference fluorescence signal are shown according to the fluorescence lifetime.

[0123] Referring to FIG. 21, a human coronary artery fluorescence lifetime image (2000) acquired using a catheter system is shown. It shows a fluorescence lifetime image (2100) measured based on a saturated fluorescence signal (SSIG in FIG. 13) before correction, a fluorescence lifetime image (2200) measured based on a corrected fluorescence signal (CSIG in FIG. 13) after correction, and a fluorescence lifetime image (2300) measured based on a reference fluorescence signal. It can be confirmed that the accuracy of fluorescence lifetime measurement is improved as the saturation phenomenon in the parts where saturation occurred in the fluorescence lifetime image before correction (2100) disappears in the fluorescence lifetime image after correction (2200). It shows images (2400, 2500, 2600) fused with optical coherence tomography images corresponding to specific values ​​of translation distances of the fluorescence lifetime images (2100, 2200, 2300) and fluorescence lifetime information (e.g., FL_INF in FIG. 13).

[0124] Referring to FIGS. 22a and 22b, the relative frequencies of the pre-correction saturated fluorescence signal (SSIG in FIG. 13), post-correction fluorescence signal (CSIG in FIG. 13), and reference fluorescence signal obtained from mouse RAW 264.7 cells and rabbit aortic vascular tissue, which are fluorescence lifetime imaging systems not used for training the convolutional neural network model (1210), are shown according to fluorescence lifetime.

[0125] Referring to FIG. 23a, a fluorescence lifetime image (3000) obtained from mouse RAW 264.7 cells is shown. It shows a saturation level (3100) obtained from mouse RAW 264.7 cells, which is a fluorescence lifetime imaging system not used for training a convolutional neural network model (1210), a saturated fluorescence signal before correction (SSIG in FIG. 13), a corrected fluorescence signal after correction (CSIG in FIG. 13), and fluorescence lifetime images (3200, 3300, 3400) measured based on a reference fluorescence signal.

[0126] Referring to FIG. 23b, a fluorescence lifetime image (4000) obtained from the aortic vascular tissue of a rabbit is shown. It shows a saturation level (4100) obtained from the aortic vascular tissue of a rabbit, which is a fluorescence lifetime image system not used for training the convolutional neural network model (1210), a saturated fluorescence signal before correction (SSIG of FIG. 13), a corrected fluorescence signal after correction (CSIG of FIG. 13), and fluorescence lifetime images (4200, 4300, 4400) measured based on a reference fluorescence signal.

[0127] FIG. 24 is a hardware configuration diagram of a fluorescence signal measuring device according to embodiments of the present invention.

[0128] Referring to FIG. 24, the fluorescence signal measuring device (5000) may be implemented as a computing device operated by at least one processor. The fluorescence signal measuring device (1000) may include one or more processors (5100), a memory (5200) for loading a computer program executed by the processor (5100), a storage device (5300) for storing the computer program and various data, a communication interface (5400), and a bus (5500) connecting them. In addition, the fluorescence signal measuring device (5000) may further include various components.

[0129] The processor (5100) is a device that controls the operation of the fluorescent signal measuring device (5000) and may be a processor of various types that processes instructions included in a computer program, and may be configured to include at least one of, for example, a CPU (Central Processing Unit), an MPU (Micro Processor Unit), an MCU (Micro Controller Unit), a GPU (Graphic Processing Unit), or any type of processor well known in the art of the present disclosure.

[0130] The memory (5200) stores various data, instructions, and / or information. The memory (5200) may load a corresponding computer program from a storage device (5300) so that instructions described to execute the operation of the present disclosure are processed by a processor (5100). The memory (5200) may be, for example, ROM (read only memory), RAM (random access memory), etc.

[0131] The storage device (5300) can store computer programs and various data non-temporarily. The storage device (5300) may be configured to include non-volatile memory such as ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), flash memory, a hard disk, a removable disk, or any form of computer-readable recording medium well known in the art to which this disclosure belongs.

[0132] The communication interface (5400) may be a wired / wireless communication module that supports wired / wireless communication.

[0133] The bus (5500) provides communication between components of the fluorescent signal measuring device (5000).

[0134] A computer program includes instructions executed by a processor (5100) and is stored in a non-transitory computer-readable storage medium, and the instructions cause the processor (5100) to execute the operation of the present disclosure. The computer program may be downloaded over a network or sold as a product. A convolutional network model (1210) may be implemented as a computer program executed by the processor (5100).

[0135] Embodiments of the present invention can be usefully utilized in various fluorescence signal measurement systems to which deep learning is applied.

[0136] Although the present invention has been described above with reference to preferred embodiments, those skilled in the art will understand that various modifications and changes can be made to the invention without departing from the spirit and scope of the invention as set forth in the following claims.

Claims

1. A signal generation unit generates a saturated fluorescence signal in which the fluorescence intensity in a first region is a threshold value, based on fluorescence photons generated by irradiating a sample with excitation light; and A signal correction unit comprising a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator includes the step of generating a correction fluorescence signal based on the saturated fluorescence signal. The fluorescence intensity in the second region of the corrected fluorescence signal corresponding to the first region of the saturated fluorescence signal is greater than the threshold value, and A method for measuring a fluorescence signal in which the above threshold value changes according to the detection characteristics for the fluorescence photon.

2. In Paragraph 1, The above signal correction unit further includes an input image processing unit, and The step of generating the above-mentioned correction fluorescence signal is, The input image processing unit generates a saturated fluorescence signal image based on the saturated fluorescence signal; and The above convolutional neural network model includes the step of generating a corrected fluorescence signal image based on the saturated fluorescence signal image, and A method for measuring a fluorescence signal characterized in that the above-mentioned corrected fluorescence signal image is output as the above-mentioned corrected fluorescence signal.

3. In claim 2, the step of generating the correction fluorescence signal is, The above input image processing unit further includes the step of generating a saturated position display image based on the saturated fluorescence signal, and The above convolutional neural network model generates the corrected fluorescence signal image based on the saturated fluorescence signal image and the saturated position indicator image, and A method for measuring a fluorescence signal, characterized in that the above-described saturated position indicator image indicates a position where the fluorescence intensity of the above-described saturated fluorescence signal image is at the threshold value.

4. In Paragraph 1, The above signal generating unit includes a dichroic mirror, and The step of generating the above-mentioned saturated fluorescent signal is, A method for measuring a fluorescence signal characterized by including the step of the fluorescence photon passing through the dichroic mirror.

5. In Paragraph 1, The above signal generating unit includes a plurality of dichroic mirrors, and The step of generating the above-mentioned saturated fluorescent signal is, A method for measuring a fluorescence signal characterized by including the step of the fluorescence photon passing through the plurality of dichroic mirrors.

6. In Paragraph 1, The above convolutional neural network model includes a residual dense network model, and The above residual dense network model includes a plurality of computational layers, and The above plurality of computation layers are, Multiple concatenation layers; Multiple convolution layers; and A method for measuring a fluorescence signal characterized by including multiple leaky RELU (rectified linear unit) layers.

7. In Paragraph 1, The above discriminator is a Markovian discriminator, and The above Markov discriminator is, Multiple convolution layers; Multiple spectral normalization layers; and A method for measuring a fluorescence signal characterized by including multiple leaky RELU (rectified linear unit) layers.

8. In Paragraph 1, The method further includes the step of performing training on the convolutional neural network based on saturated fluorescence signal data, A method for measuring a fluorescence signal characterized by obtaining the above-mentioned saturated fluorescence signal data by performing random rescaling on the fluorescence signal data.

9. A signal generation unit generates a saturated fluorescence signal in which the fluorescence intensity in a first region is a threshold value, based on fluorescence photons generated by irradiating a sample with excitation light; A signal correction unit comprising a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator, the step of generating a correction fluorescence signal based on the saturated fluorescence signal; and The signal processing unit includes the step of generating fluorescence lifetime information based on the correction fluorescence signal, and The fluorescence intensity in the second region of the corrected fluorescence signal corresponding to the first region of the saturated fluorescence signal is greater than the threshold value, and A fluorescence lifetime measurement method in which the above threshold value changes according to the detection characteristics for the above fluorescent photons.

10. A signal generation unit generates a saturated fluorescence signal in which the fluorescence intensity in a first region is a threshold value, based on fluorescence photons generated by irradiating a sample with excitation light; A signal correction unit comprising a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator, and a step of generating a correction fluorescence signal based on the saturated fluorescence signal; A step in which a signal processing unit generates fluorescence lifetime information based on the correction fluorescence signal; and The fluorescence lifetime image processing unit includes the step of generating a fluorescence lifetime image based on the fluorescence lifetime information, and The fluorescence intensity in the second region of the corrected fluorescence signal corresponding to the first region of the saturated fluorescence signal is greater than the threshold value, and The above threshold value is changed according to the detection characteristics for the fluorescent photon, and A fluorescence lifetime imaging method wherein the above fluorescence lifetime image is a two-dimensional image, the horizontal axis of the above fluorescence lifetime image is information on the translational distance for the collection of the fluorescence photons, and the vertical axis of the above fluorescence lifetime image is information on the rotational translation angle for the collection of the fluorescence photons.

11. A signal generation unit that generates a saturated fluorescence signal in which the fluorescence intensity in the first region is a threshold value; and It includes a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator, and a signal correction unit that generates a correction fluorescence signal based on the saturated fluorescence signal. The fluorescence intensity in the second region of the corrected fluorescence signal corresponding to the first region of the saturated fluorescence signal is greater than the threshold value, and The above signal generating unit is, A light generating unit that generates excitation light to irradiate a sample; and It includes a detection unit that generates the saturated fluorescence signal based on fluorescent photons generated by irradiating the excitation light onto the sample, A fluorescent signal measuring device in which the above threshold value changes according to the characteristics of the above detection unit.

12. A signal generation unit that generates a saturated fluorescence signal in which the fluorescence intensity in the first region is a threshold value; A signal correction unit comprising a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator, and generating a correction fluorescence signal based on the saturated fluorescence signal; and It includes a signal processing unit that generates fluorescence lifetime information based on the above-mentioned corrected fluorescence signal, and The fluorescence intensity in the second region of the corrected fluorescence signal corresponding to the first region of the saturated fluorescence signal is greater than the threshold value, and The above signal generating unit is, A light generating unit that generates excitation light to irradiate a sample; and It includes a detection unit that generates the saturated fluorescence signal based on fluorescent photons generated by irradiating the excitation light onto the sample, The above threshold value is a fluorescence lifetime measuring device that changes according to the characteristics of the above-mentioned sensing unit.

13. A signal generation unit that generates a saturated fluorescence signal in which the fluorescence intensity in the first region is a threshold value; A signal correction unit comprising a convolutional neural network (CNN) model in which adversarial learning is performed using a discriminator, and which generates a correction fluorescence signal based on the saturated fluorescence signal; A signal processing unit that generates fluorescence lifetime information based on the above-mentioned corrected fluorescence signal; and It includes a fluorescence lifetime image processing unit that generates a fluorescence lifetime image based on the above fluorescence lifetime information, and The fluorescence intensity in the second region of the corrected fluorescence signal corresponding to the first region of the saturated fluorescence signal is greater than the threshold value, and The above signal generating unit is, A light generating unit that generates excitation light to irradiate a sample; A collection unit capable of parallel and rotational movement for collecting fluorescent photons generated by irradiating the above excitation light onto the above sample; and It includes a detection unit that generates the saturated fluorescence signal based on the above fluorescent photons, and The above threshold value is changed according to the characteristics of the detection unit, and A fluorescence lifetime imaging device wherein the above fluorescence lifetime image is a two-dimensional image, the horizontal axis of the above fluorescence lifetime image is information on the parallel translation distance of the collection unit, and the vertical axis of the above fluorescence lifetime image is information on the rotational translation angle of the collection unit.

Images