Wide-field fluorescence lifetime imaging of cancer.

Sample preparation

Fresh surgical specimens from a number of organs (colon, stomach, bladder, liver, pancreas) were collected from patients undergoing surgical resection for malignant disease. FLIM imaging preceded fixation with formaldehyde. The specimens were either opened (colon, stomach, bladder), cut open (pancreas) or cut into serial slices (liver) by an experienced histopathologist to reveal the malignancy. The specimens were then washed in Hanks’ Balanced Salt Solution to remove blood, stool and other debris from the tissue surface, and imaged within 60 minutes of resection. After imaging, the biopsies were photographed and subsequently fixed in formaldehyde. Histological correlation with the photograph and FLIM image, and histopathological analysis was performed by an experienced histopathologist. This study was performed under a protocol approved by the Hammersmith and Queen Charlotte’s & Chelsea Hospitals Research Ethics Committee. All patients gave their informed consent to the use of tissue for medical research.

Fluorescence Imaging Instrumentation and Protocol

Experiments were carried out on a compact FLIM system designed and built by the Photonics Group at Imperial College London and installed in a laboratory space in the Histopathology Department to allow rapid ex vivo measurements of the resection specimens immediately after resection. The principles of operation and of the instrumentation used in time-domain FLIM have been described elsewhere [ 26 , 35 ]. In brief, a 10 ps pulsed frequency-tripled Nd:YVO 4 laser (Spectra-Physics Inc, Vanguard 350-HMD355) with an 80 MHz repetition rate provided pulsed UV excitation light at 355 nm. The laser output was delivered to the sample under investigation by a 1 m length of optical fibre (Thorlabs, FT-200-UMT) and a divergent holographic diffuser (Physical Optics Corp, 60° full width half maximum). The emitted fluorescence was imaged via a 375 nm long-pass filter (Schott, GG375) onto a gated optical intensifier (GOI) (Kentech Instruments Ltd, HRI) coupled to a CCD camera (Hamamatsu, ORCA ER) ( Fig. 1 ). The GOI is used to intensify the fluorescence signal and produce ultrashort (1000 ps) time-gated ‘snapshots’ that are recorded by the CCD camera. By varying the delay between the excitation pulse and the opening of the gated intensifier, images of the exponentially decreasing fluorescence decay were recorded at 25 different time points at 250 ps intervals. A typical decay acquisition took 4 seconds to record. For each sample, 5 decays were acquired sequentially and then averaged to minimise the impact of photobleaching on the lifetime measurement. The total FLIM acquisition time was ~20 s. The CCD camera was also used to record white light reflectance images of the sample under investigation. Image acquisition was controlled by a LabVIEW (National Instruments Corporation, Austin, Texas, USA) program developed in-house by members of the Photonics Group. Data processing, lifetime calculation and FLIM map generation For each pixel in the series of time-gated images, a lifetime value was computed using an iterative weighted non-linear least-squares (WNLLS) algorithm to fit a mono-exponential decay curve to the data. The background signal due to CCD camera readout noise and offset was subtracted prior to fitting. False colour lifetime maps of the specimens were produced by assigning each pixel in the image a colour according to its lifetime value. The fluorescence intensity threshold was adjusted so that signals from outside the specimen were below threshold and excluded from lifetime processing. FLIM maps were also merged with fluorescence integrated intensity images to produce images displaying both lifetime and intensity information. To calculate lifetime values from different regions within the same image, manually delineated region of interest masks were drawn from the white light reflectance images and applied to the processing of the fluorescence decays. All data processing was performed with FLIMProcessing, a LabVIEW program written in-house by members of the Photonics Group.

Rapid lifetime determination fluorescence lifetime imaging

The method for rapid lifetime determination FLIM has previously been described in detail, see reference [ 33 ] and references therein. In brief, a low-jitter electronic delay generator synchronised to the CCD camera’s frame grabber status signal was used to adjust the relative delay between time-gated images and allowed the delay to be altered within 2.5 ms. Alternating frames, at delays of 0.2 and 5.7 ns with respect to the excitation pulse, were acquired with an exposure time of 100 ms, resulting in a frame rate of ~7.7 Hz. The detection gate-width was set at 2.4 ns.

Statistical analysis

We compared the mean lifetime and mean intensity values between the paired carcinoma and non-carcinoma regions for each sample with a Wilcoxon Signed-Rank test, as the data is nonparametric. The overlap of distributions of fluorescence intensity or fluorescence lifetime values from carcinoma and non-carcinoma regions for each sample was expressed quantitatively in terms of the area under the curve (AUC) of a receiver operating characteristic curve. The AUC is the probability that the fluorescence intensity/lifetime will correctly identify a randomly selected ‘positive’ pixel (e.g. presence of cancer) compared to a randomly chosen ‘negative’ one (e.g. normal tissue). The possible values of AUC therefore run from 0.5, indicating random assignment and therefore no diagnostic use, to either 0 or 1 (depending on the relative position of the histograms) for an ideal indicator.

Fluorescence Imaging Instrumentation and Protocol

Experiments were carried out on a compact FLIM system designed and built by the Photonics Group at Imperial College London and installed in a laboratory space in the Histopathology Department to allow rapid ex vivo measurements of the resection specimens immediately after resection. The principles of operation and of the instrumentation used in time-domain FLIM have been described elsewhere [ 26 , 35 ]. In brief, a 10 ps pulsed frequency-tripled Nd:YVO 4 laser (Spectra-Physics Inc, Vanguard 350-HMD355) with an 80 MHz repetition rate provided pulsed UV excitation light at 355 nm. The laser output was delivered to the sample under investigation by a 1 m length of optical fibre (Thorlabs, FT-200-UMT) and a divergent holographic diffuser (Physical Optics Corp, 60° full width half maximum). The emitted fluorescence was imaged via a 375 nm long-pass filter (Schott, GG375) onto a gated optical intensifier (GOI) (Kentech Instruments Ltd, HRI) coupled to a CCD camera (Hamamatsu, ORCA ER) ( Fig. 1 ). The GOI is used to intensify the fluorescence signal and produce ultrashort (1000 ps) time-gated ‘snapshots’ that are recorded by the CCD camera. By varying the delay between the excitation pulse and the opening of the gated intensifier, images of the exponentially decreasing fluorescence decay were recorded at 25 different time points at 250 ps intervals. A typical decay acquisition took 4 seconds to record. For each sample, 5 decays were acquired sequentially and then averaged to minimise the impact of photobleaching on the lifetime measurement. The total FLIM acquisition time was ~20 s. The CCD camera was also used to record white light reflectance images of the sample under investigation. Image acquisition was controlled by a LabVIEW (National Instruments Corporation, Austin, Texas, USA) program developed in-house by members of the Photonics Group.