Early detection of head and neck tumors is crucial for patient survival. Often, diagnoses are made based on endoscopic examination of the larynx followed by biopsy and histological analysis, leading to a high interobserver variability due to subjective assessment. In this regard, early non-invasive diagnostics independent of the clinician would be a valuable tool. A recent study has shown that hyperspectral imaging (HSI) can be used for non-invasive detection of head and neck tumors, as precancerous or cancerous lesions show specific spectral signatures that distinguish them from healthy tissue. However, HSI data processing is challenging due to high spectral variations, various image interferences, and the high dimensionality of the data. Therefore, performance of automatic HSI analysis has been limited and so far, mostly ex-vivo studies have been presented with deep learning. In this work, we analyze deep learning techniques for in-vivo hyperspectral laryngeal cancer detection. For this purpose we design and evaluate convolutional neural networks (CNNs) with 2D spatial or 3D spatio-spectral convolutions combined with a state-of-the-art Densenet architecture. For evaluation, we use an in-vivo data set with HSI of the oral cavity or oropharynx. Overall, we present multiple deep learning techniques for in-vivo laryngeal cancer detection based on HSI and we show that jointly learning from the spatial and spectral domain improves classification accuracy notably. Our 3D spatio-spectral Densenet achieves an average accuracy of 81%.
Here we present a study where we used in vivo hyperspectral imaging (HSI) for the detection of upper aerodigestive tract (UADT) cancer. Hyperspectral datasets were recorded in 100 patients before surgery in vivo. We established an automated data interpretation pathway that can classify the tissue into healthy and tumorous using, different deep learning techniques. Our method is based on convolutional neural networks (CNNs) with 2D spatial or 3D spatio-spectral convolutions combined with a state-of-the-art Densenet architecture. Using both the spatial and spectral domain improves classification accuracy notably. Our 3D spatio-spectral Densenet classification method achieves an average accuracy of over 80%.
Immunophenotyping of peripheral blood leukocytes (PBLs) is performed by flow cytometry (FCM) as the golden standard. Slide based cytometry systems for example laser scanning cytometer (LSC) can give additional information (repeated staining and scanning, morphology). In order to adequately judge on the clinical usefulness of immunophenotyping by LSC it is obligatory to compare it with the long established FCM assays. We performed this study to systematically compare the two methods, FCM and LSC for immunophenotyping and to test the correlation of the results. Leucocytes were stained with directly labeled monoclonal antibodies with whole blood staining method. Aliquots of the same paraformaldehyde fixed specimens were analyzed in a FACScan (BD-Biosciences) using standard protocols and parallel with LSC (CompuCyte) after placing to glass slide, drying and fixation by aceton and 7-AAD staining. Calculating the percentage distribution of PBLs obtained by LSC and by FCM shows very good correlation with regression coefficients close to 1.0 for the major populations (neutrophils, lymphocytes, and monocytes), as well as for the lymphocyte sub-populations (T-helper-, T-cytotoxic-, B-, NK-cells). LSC can be recommended for immunophenotyping of PBLs especially in cases where only very limited sample volumes are available or where additional analysis of the cells’ morphology is important. There are limitations in the detection of rare leucocytes or weak antigens where appropriate amplification steps for immunofluorescence should be engaged.
For immunophenotypic analysis more measurable parameters for the discrimination of leukocyte subsets are necessary. With a single scan six fluorochromes can be distinguished with the Laser Scanning Cytometer (LSC). Due to the number of PMTs the amount of simultaneously measurable fluorescences per scan is limited. Nevertheless, the amount of measurable colors can be improved to eight by appropriate change of the filter settings and two scans per specimen. Aim of this study was to use the special features of Slide based Cytometry (SBC) beyond filter change, remeasurement and merging to distinguish fluorochromes with similar emission spectra. The photosensitivity of fluorochromes that are excited and emit in a similar wavelength range may be very different. The number of measurable parameters per PMT was increased using photosensitivity of different fluorochromes as additional criteria. Peripheral blood leukocytes were stained with antibodies conjugated to the fluorochromes APC, APC-Cy5.5 and Alexa-Fluor 633 and mounted on conventional uncoated glass slides with Fluorescence mounting medium. Specimens were excited in the LSC with the HeNe (633nm) Laser and measured at different filter settings (670/20nm-filter for APC/ALEXA 633 and 710/20nm-filter for APC-Cy5.5). At this point, APC-Cy5.5 and APC/ALEXA633 were already distinguishable. In order to differentiate between APC and ALEXA633 photobleaching was performed by repeated excitation with the laser at 633nm. Control measurements proved that APC is much more sensitive against laser excitation, i.e. looses much more fluorescence intensity than ALEXA633. The separate measurements (before/after filter change and before/after bleaching) were merged into one file. The photostability of Alexa-Fluor 633 (1.02% bleach per scan) and APC (5.74% bleach per scan) are substantially different. Therefore, after bleaching and merging both fluorochromes can be distinguished and are regarded by the software as separate parameters. The fluorochromes APC/ALEXA633 and APC-Cy5.5 can be discriminated by changing the emission filters before bleach. By sequential photobleaching, change of filters and subsequent merging of the data the number of simultaneously measurable “colors” is substantially increased.
Slide-based cytometry (SBC) is a promising new development in clinical diagnostics. The Laser Scanning Cytometer (LSC) was the first such system that became commercially available. Over the years methods have been developed that can be applied in a broad variety of clinical diagnostic settings. The principle of SBC is that fluorochrome labeled specimens are immobilized on microscopic slides which are placed on a conventional epi-fluorescence microscope. Specimens are analyzed by one or two lasers. Data comparable to flow cytometry are generated. But in addition, the position of each individual event is recorded, a feature that allows to re-localize and to visualize each event that has been measured. The major advantage of LSC compared with other cytometric methods is the combination of two features: a) the minimal clinical sample volume needed, and b) the connection of fluorescence data and morphological information about the measured event. We have developed and will present examples of different techniques for application in clinical diagnosis: (1) Immunophenotyping using up to six different fluorochromes at a time, (2) analysis of minimal sample volumes using fine needle aspirate biopsies, and (3) analysis of cells in tissue sections. With these assays and assays developed by others, SBC has proven its wide spectrum of clinical applicability and can be introduced as a standard technology for multiple clinical settings.
The request for a more profound immunophenotyping and sometimes the lack of material demands more measurable fluorescence colors to increase the number of detectable antigens per specimen. Six different fluorescences are distinguishable in the Laser Scanning Cytometer (LSC). In the present study we wanted to increase this number to eight colors per measurement. Combined with an earlier study it is likely possible to measure n fluorescences i.e. n leukocyte subsets by a series of measurements followed by subsequent restraining steps. The new method is realized by s-ing the combination of filter change and a subsequent re-measurement for the distinction between the fluorescent dyes Cy5 and Cy5.5. The optical filters are replaced after the first measurement and the same specimen is remeasured without removing it from the microscope. For the second measurement a filter is inserted that detects Cy5.5 but not Cy5 (710/10nm). After the second measurement of the same specimen both data files are combined. With the aid of this feature it is possible to line out the differences between both measurements. If the data from the second measuring (Cy5.5 only) is subtracted from the first, Cy5 data is the result. After the first two measurements when eight different fluorescences (i.e. antigens or leukocyte subsets) were analyzed, the same cells are restained and a new measurement is performed. In theory, one can perform n re-measurements with eight fluorescences respectively. The information gained per specimen is only limited by the number of available antibodies and b sterical hindrance.
Immunophenotyping of peripheral blood leukocytes (PBL) is a very well documented application of Slide Based Cytometry (SBC). As for any other assay it is of highest importance to ensure that all cells which are relevant for an analysis are recognized. Unlike assays for cultured cells which have homogenous morphology immunophenotyping of PBLs is performed on cells with heterogeneous size and shape. Therefore, triggering on parameters related to cell morphology might lead to an incomplete analysis of just a subset of cells especially in pathological conditions. Several dyes stain DNA specifically in a wide variety of emission spectra. Many of them show some influence of the chromatin condensation and organization on the staining intensity. DNA dyes therefore can be used to differentiate between cell types having the same ploidy. This can be exploited for immunophenotyping since some dyes therefore can partially replace antibody staining. The concept of using DNA dyes in the setting of immunostaining has the following advantages: (1) nuclear staining provides a stable and easy triggering signal that guarantees both, that neither cells are excluded nor that debris or polluting particles are included into the analysis; (2) some DNA dyes differentiate between mononuclear and polymorphonuclear cells. A disadvantage of DNA dyes is that mostly cells have to be permeabilized. Because of this only one set of immunophenotypic markers can be stained, cells are fixed and permeabilized, and then nuclei are stained with the appropriate DNA dye. In the study we demonstrate the use of the most commonly available DNA dyes (7-AAD, To-Pro, To-To, PI etc.) in their applicability in immunophenotyping. An overview of spectral properties, fluorescence spill-over and optimal combinations with surface antigen staining will be shown. As in general for SBC only very small sample volumes are needed. This allows to serially analyze PBL in clinical settings that up to now could not be studied in detail such as in the critical ill patient, during major surgery, and in new-borns and infants.
In order to minimize hospitalization and morbidity with optimized therapy for a patient with a tumor of the parotid gland a malignancy must be confirmed or excluded as soon as possible. Up to now, non- and minimal-invasive methods do not yield this information. For fine needle aspirate biopsies (FNABs), analysis by a specialized cytologist yields subjective and qualitative but not objective and quantitative data. LSC is a semi-automated microscope-based technology and offers ideal prerequisites for the analysis of specimens fixed on a slide. We have established an assay for FNABs from parotid gland tumors. Cells are stained for cytokeratin and DNA. The analysis quantitatively determines the ploidy of the cells and the degree of condensation of the DNA; on this basis the percentage of cells undergoing mitosis can be determined. Subsequently the cells are stained by H&E and are re-localized on the slide at their fixed position. Micrographs are taken for objective documentation of the cells' morphology. Using this assay FNABs from parotid gland tumors were analyzed; tumors that were diagnosed as benign by routine histopathology showed no aneuploidy whereas malignant tumors were aneuploid. This preliminary study demonstrates the capacities of LSC for minimal-invasive assays yielding quantitative and objective data.
In lymphatic organs the quantitative analysis of the spatial distribution of leukocytes would give relevant information about alterations during diseases (leukemia, HIV, AIDS) and their therapeutic regimen. Analysis of them in solid tissues is difficult to perform but would yield important data in a variety of clinical and experimental settings. We have developed an automated analysis method for LSC suitable for archived or fresh biopsy material of human lymph nodes and tonsils. Sections are stained with PI for DNA and up to three antigens using direct or indirect immunofluorescence staining. Measurement is triggered on DNA-fluorescence (Argon Laser). Due to the heterogeneity in cell density measurements are repeatedly performed at different threshold levels (low threshold: regions of low cellular density, germinal centers; high threshold: dense regions, mantle zone). Data are acquired by single- (Ar) or dual-laser excitation (Ar-HeNe) in order to determine data from single- (FITC), up to triple-staining (FITC/PE-Cy5/APC). Percentage and cellular density of cell-subsets is quantified in different structural regions of the specimen. Comparison with manual analysis of identical specimens showed very good correlation. With LSC a semi-automated operator-independent and immunophenotyping of lymphatic tissues with simultaneously up to four antibodies is possible. This technique should yield new insight into processes during diseases and should help to quantify the success of therapeutic interventions.
Cardiac surgery with cardiopulmonary bypass (CPB) alters the leukocyte composition of the peripheral blood (PB). This response contributes to the sometimes adverse outcome with capillary leakage. Migration of activated cells to sites of inflammation, driven by chemokines is part of this response. In order to determine the chemotactic activity of patients serum during and after surgery we established an assay for PB leukocytes (PBL). PBL from healthy donors were isolated and 250,000 cells were placed into a migration chamber separated by a filter from a second lower chamber filled with patient serum. After incubation cells from top and bottom chamber were removed and stained with a cocktail of monoclonal antibodies for leukocyte subsets and analyzed on a flow cytometer (FCM). Cells at the bottom of the filter belong to the migrating compartment and were quantified by LSC after staining of nucleated cells. Increased chemotactic activity started at onset of anaesthesia followed by a phase of low activity immediately after surgery and a second phase of a high post-operative activity. The in vitro results correlated with results obtained by immunopenotyping of circulating PBL. Manipulation of the chemokine pattern might prove beneficial to prevent extravasation of cells leading to tissue damage. In chemotaxis assays with low amount of available serum the combined use of FCM and Laser Scanning LSC proved as an appropriate analytical tool.
LSC is a microscope-based technology. The principle of the instrument is that any specimen is immobilized on a microscope slide. Therefore the cells are not lost in a fluid stream but are kept on the slide and minimal specimens as low as 1.000 cells can be analyzed. Additionally cells are available for further analyses such as staining for another set of specific markers and re-analysis or cytological staining (H&E). This approach multiplies the information gained from a given sample. We have established an assay for immunophenotyping of peripheral blood leukocytes by LSC. Cells are prepared according to routine flow cytometry protocols with a first set of CD-antibodies and are fixed on microscope slides. As a stable trigger signal the nuclear DNA is stained by 7-aminoactinomycin-D. This guarantees that all nucleated cells and that only nucleated cells are included in the analysis, and many differentiate between lymphocytes and neutrophiles by staining intensity. After analysis cells are stained with a second set of CD-antibodies and analyzed again. This step can be repeated with a third set of CD-antigens. Since the location of the cells on the slide is fixed data from the analyses can be attributed to the same cell.