Assessment of the statistical significance of classifications in infrared spectroscopy based diagnostic models

Fourier transform IR imaging of primary tumors predicts lymph node metastasis of bladder carcinoma

The process of metastasis is complex and often impossible to be recognized in conventional clinical diagnosis. Lymph node metastasis (LNM) of bladder carcinoma (BC) is often associated with muscle-invasive tumors. To prevent and recognize LNM, the standard treatment includes radical cystectomy with lymph node dissection and histological examination. Here, we propose infrared (IR) microscopy as a tool for the prediction of LNM. For this purpose, IR images of bladder biopsies from patients with diagnosed non-metastatic early (E BC) and advanced (A BC), as well as metastatic advanced (M BC) bladder cancer were first collected. Furthermore, this dataset was complemented with images of the secondary tumors from the lymph nodes (M LN) of the M BC patients. Unsupervised clustering was used to extract tissue structures from IR images to create a data set comprising 382 IR spectra of non-metastatic bladder tumors and 241 metastatic ones. Based on that, we next established discrimination models using PLS-DA with repeated random sampling double cross-validation, and permutation test to perform the classification. The accuracy of BC metastasis prediction from IR bladder biopsies was 83 % and 78 % for early and advanced BC, respectively, herein demonstrating a proof-of-concept IR detection of BC metastasis. The analysis of spectral profiles additionally showed molecular composition similarity between metastatic bladder and lymph node tumors. We also determined spectral biomarkers of LNM that are associated with sugar metabolism, remodeling of extracellular matrix, and morphological features of cancer cells. Our approach can improve clinical decision-making in urological oncology.

Evaluation of grade and invasiveness of bladder urothelial carcinoma using infrared imaging and machine learning

Urothelial bladder carcinoma (BC) is primarily diagnosed with a subjective examination of biopsies by histopathologists, but accurate diagnosis remains time-consuming and of low diagnostic accuracy, especially for low grade non-invasive BC. We propose a novel approach for high-throughput BC evaluation by combining infrared (IR) microscopy of bladder sections with machine learning (partial least squares-discriminant analysis) to provide an automated prediction of the presence of cancer, invasiveness and grade. Cystoscopic biopsies from 50 patients with clinical suspicion of BC were histologically examined to assign grades and stages. Adjacent tissue cross-sections were IR imaged to provide hyperspectral datasets and cluster analysis segregated IR images to extract the average spectra of epithelial and subepithelial tissues. Discriminant models, which were validated using repeated random sampling double cross-validation, showed sensitivities (AUROC) ca. 85% (0.85) for the identification of cancer in epithelium and subepithelium. The diagnosis of non-invasive and invasive cases showed sensitivity values around 80% (0.84-0.85) and 76% (0.73-0.80), respectively, while the identification of low and high grade BC showed higher sensitivity values 87-88% (0.91-0.92). Finally, models for the discrimination between cancers with different invasiveness and grades showed more modest AUROC values (0.67-0.72). This proves the high potential of IR imaging in the development of ancillary platforms to screen bladder biopsies.

Infrared Spectroscopy of Blood

The magnitude of infectious diseases in the twenty-first century created an urgent need for point-of-care diagnostics. Critical shortages in reagents and testing kits have had a large impact on the ability to test patients with a suspected parasitic, bacteria, fungal, and viral infections. New point-of-care tests need to be highly sensitive, specific, and easy to use and provide results in rapid time. Infrared spectroscopy, coupled to multivariate and machine learning algorithms, has the potential to meet this unmet demand requiring minimal sample preparation to detect both pathogenic infectious agents and chronic disease markers in blood. This focal point article will highlight the application of Fourier transform infrared spectroscopy to detect disease markers in blood focusing principally on parasites, bacteria, viruses, cancer markers, and important analytes indicative of disease. Methodologies and state-of-the-art approaches will be reported and potential confounding variables in blood analysis identified. The article provides an up to date review of the literature on blood diagnosis using infrared spectroscopy highlighting the recent advances in this burgeoning field.

The Impact of Preprocessing Methods for a Successful Prostate Cell Lines Discrimination Using Partial Least Squares Regression and Discriminant Analysis Based on Fourier Transform Infrared Imaging

Fourier transform infrared spectroscopy (FT-IR) is widely used in the analysis of the chemical composition of biological materials and has the potential to reveal new aspects of the molecular basis of diseases, including different types of cancer. The potential of FT-IR in cancer research lies in its capability of monitoring the biochemical status of cells, which undergo malignant transformation and further examination of spectral features that differentiate normal and cancerous ones using proper mathematical approaches. Such examination can be performed with the use of chemometric tools, such as partial least squares discriminant analysis (PLS-DA) classification and partial least squares regression (PLSR), and proper application of preprocessing methods and their correct sequence is crucial for success. Here, we performed a comparison of several state-of-the-art methods commonly used in infrared biospectroscopy (denoising, baseline correction, and normalization) with the addition of methods not previously used in infrared biospectroscopy classification problems: Mie extinction extended multiplicative signal correction, Eiler’s smoothing, and probabilistic quotient normalization. We compared all of these approaches and their effect on the data structure, classification, and regression capability on experimental FT-IR spectra collected from five different prostate normal and cancerous cell lines. Additionally, we tested the influence of added spectral noise. Overall, we concluded that in the case of the data analyzed here, the biggest impact on data structure and performance of PLS-DA and PLSR was caused by the baseline correction; therefore, much attention should be given, especially to this step of data preprocessing.

Searching for a Better Match between Protein Secondary Structure Definitions and Protein FTIR Spectra

Obtaining protein secondary structure content from high-resolution structures requires definitions and thresholds for the various parameters involved, typically hydrogen bond energy or length/angle and backbone φ/ψ angles. Several definitions are currently used and can have a profound impact on secondary structure content. Fourier transform infrared (FTIR) spectroscopy has its own sensitivity to molecular geometry. It is, therefore, important to select a set of definitions that matches this sensitivity. Here, we used a new protein set consisting of 92 proteins designed for the calibration of spectroscopic methods. Spectra have been obtained from protein microarrays in a high throughput process. The potential for improving secondary structure predictions from FTIR spectra has been tested using 71 structures determined according to different definitions. The paper demonstrates that different secondary structure definitions result in large variations in secondary structure content that are not equivalent in view of the protein FTIR spectra. The prediction quality factor ζ can be improved by ca. 20-50% by selecting an adequate definition set. The results also indicate that the dictionary of secondary structure of proteins (DSSP) algorithm, which is currently widely used to evaluate protein secondary structure content, is a good choice when dealing with FTIR spectra.

Multimodal vibrational studies of drug uptake in vitro: Is the whole greater than the sum of their parts?

Herein, we investigated the use of multimodal Raman and infrared (IR) spectroscopic microscopy for the elucidation of drug uptake and subsequent cellular responses. Firstly, we compared different methods for the analysis of the combined data. Secondly, we evaluated whether the combined analysis provided enough benefits to justify the fusion of the data. A459 cells inoculated with doxorubicin (DOX) at different times were fixed and analysed using each technique. Raman spectroscopy provided high sensitivity to DOX and enabled an accurate estimation of the drug uptake at each time point, whereas IR provided a better insight into the resultant changes in the biochemical composition of the cell. In terms of benefits of data fusion, 2D correlation analysis allowed the study of the relationship between IR and Raman variables, whereas the joint analysis of IR and Raman enabled the correlation of the different variables to be monitored over time. In summary, the complementary nature of IR and Raman makes the combination of these vibrational techniques an appealing tool to follow drug kinetics and cellular response.

Toward Rapid Screening of Liver Grafts at the Operating Room Using Mid-infrared Spectroscopy

The estimation of steatosis in a liver graft is mandatory prior to liver transplantation, as the risk of graft failure increases with the level of infiltrated fat. However, the assessment of liver steatosis before transplantation is typically based on a qualitative or semiquantitative characterization by visual inspection and palpation and histological analysis. Thus, there is an unmet need for transplantation surgeons to have access to a diagnostic tool enabling an in situ fast classification of grafts prior to extraction. In this study, we have assessed an attenuated total reflection-Fourier transform infrared (ATR-FTIR) spectroscopic method compatible with the requirements of an operation room for the evaluation of the lipid contents in human livers. A set of 20 human liver biopsies obtained from organs intended for transplantation were analyzed by expert pathologists, ATR-FTIR spectroscopy, lipid biochemical analysis, and UPLC-ESI(+/-)TOFMS for lipidomic profiling. Comparative analysis of multisource data showed strong correlations between ATR-FTIR, clinical, and lipidomic information. Results show that ATR-FTIR captures a global picture of the lipid composition of the liver, along with information for the quantification of the triradylglycerol content in liver biopsies. Although the methodology performance needs to be further validated, results support the applicability of ATR-FTIR for the in situ determination of the grade of liver steatosis at the operation room as a fast, quantitative method, as an alternative to the qualitative and subjective pathological examination.

Infrared spectroscopy coupled to cloud-based data management as a tool to diagnose malaria: a pilot study in a malaria-endemic country

Background

Widespread elimination of malaria requires an ultra-sensitive detection method that can detect low parasitaemia levels seen in asymptomatic carriers who act as reservoirs for further transmission of the disease, but is inexpensive and easy to deploy in the field in low income settings. It was hypothesized that a new method of malaria detection based on infrared spectroscopy, shown in the laboratory to have similar sensitivity to PCR based detection, could prove effective in detecting malaria in a field setting using cheap portable units with data management systems allowing them to be used by users inexpert in spectroscopy. This study was designed to determine whether the methodology developed in the laboratory could be translated to the field to diagnose the presence of Plasmodium in the blood of patients presenting at hospital with symptoms of malaria, as a precursor to trials testing the sensitivity of to detect asymptomatic carriers.

Methods

The field study tested 318 patients presenting with suspected malaria at four regional clinics in Thailand. Two portable infrared spectrometers were employed, operated from a laptop computer or a mobile telephone with in-built software that guided the user through the simple measurement steps. Diagnostic modelling and validation testing using linear and machine learning approaches was performed against the gold standard qPCR. Sample spectra from 318 patients were used for building calibration models (112 positive and 110 negative samples according to PCR testing) and independent validation testing (39 positive and 57 negatives samples by PCR).

Results

The machine learning classification (support vector machines; SVM) performed with 92% sensitivity (3 false negatives) and 97% specificity (2 false positives). The Area Under the Receiver Operation Curve (AUROC) for the SVM classification was 0.98. These results may be better than as stated as one of the spectroscopy false positives was infected by a Plasmodium species other than Plasmodium falciparum or Plasmodium vivax, not detected by the PCR primers employed.

Conclusions

In conclusion, it was demonstrated that ATR-FTIR spectroscopy could be used as an efficient and reliable malaria diagnostic tool and has the potential to be developed for use at point of care under tropical field conditions with spectra able to be analysed via a Cloud-based system, and the diagnostic results returned to the user’s mobile telephone or computer. The combination of accessibility to mass screening, high sensitivity and selectivity, low logistics requirements and portability, makes this new approach a potentially outstanding tool in the context of malaria elimination programmes. The next step in the experimental programme now underway is to reduce the sample requirements to fingerprick volumes.

Predicting one-year outcome in first episode psychosis using machine learning

BACKGROUND

Early illness course correlates with long-term outcome in psychosis. Accurate prediction could allow more focused intervention. Earlier intervention corresponds to significantly better symptomatic and functional outcomes. Our study objective is to use routinely collected baseline demographic and clinical characteristics to predict employment, education or training (EET) status, and symptom remission in patients with first episode psychosis (FEP) at one-year.

METHODS AND FINDINGS

83 FEP patients were recruited from National Health Service (NHS) Glasgow between 2011 and 2014 to a 24-month prospective cohort study with regular assessment of demographic and psychometric measures. An external independent cohort of 79 FEP patients were recruited from NHS Glasgow and Edinburgh during a 12-month study between 2006 and 2009. Elastic net regularised logistic regression models were built to predict binary EET status, period and point remission outcomes at one-year on 83 Glasgow patients (training dataset). Models were externally validated on an independent dataset of 79 patients from Glasgow and Edinburgh (validation dataset). Only baseline predictors shared across both cohorts were made available for model training and validation. After excluding participants with missing outcomes, models were built on the training dataset for EET status, period and point remission outcomes and externally validated on the validation dataset. Models predicted EET status, period and point remission with receiver operating curve (ROC) area under the curve (AUC) performances of 0.876 (95%CI: 0.864, 0.887), 0.630 (95%CI: 0.612, 0.647) and 0.652 (95%CI: 0.635, 0.670) respectively. Positive predictors of EET included baseline EET and living with spouse/children. Negative predictors included higher PANSS suspiciousness, hostility and delusions scores. Positive predictors for symptom remission included living with spouse/children, and affective symptoms on the Positive and Negative Syndrome Scale (PANSS). Negative predictors of remission included passive social withdrawal symptoms on PANSS. A key limitation of this study is the small sample size (n) relative to the number of predictors (p), whereby p approaches n. The use of elastic net regularised regression rather than ordinary least squares regression helped circumvent this difficulty. Further, we did not have information for biological and additional social variables, such as nicotine dependence, which observational studies have linked to outcomes in psychosis.

CONCLUSIONS AND RELEVANCE

Using advanced statistical machine learning techniques, we provide the first externally validated evidence, in a temporally and geographically independent cohort, for the ability to predict one-year EET status and symptom remission in individual FEP patients.

High resolution FTIR imaging provides automated discrimination and detection of single malaria parasite infected erythrocytes on glass

New highly sensitive tools for malaria diagnostics are urgently needed to enable the detection of infection in asymptomatic carriers and patients with low parasitemia. In pursuit of a highly sensitive diagnostic tool that can identify parasite infections at the single cell level, we have been exploring Fourier transform infrared (FTIR) microscopy using a Focal Plane Array (FPA) imaging detector. Here we report for the first time the application of a new optic configuration developed by Agilent that incorporates 25× condenser and objective Cassegrain optics with a high numerical aperture (NA = 0.81) along with additional high magnification optics within the microscope to provide 0.66 micron pixel resolution (total IR system magnification of 61×) to diagnose malaria parasites at the single cell level on a conventional glass microscope slide. The high quality images clearly resolve the parasite's digestive vacuole demonstrating sub-cellular resolution using this approach. Moreover, we have developed an algorithm that first detects the cells in the infrared image, and secondly extracts the average spectrum. The average spectrum is then run through a model based on Partial Least Squares-Discriminant Analysis (PLS-DA), which diagnoses unequivocally the infected from normal cells. The high quality images, and the fact this measurement can be achieved without a synchrotron source on a conventional glass slide, shows promise as a potential gold standard for malaria detection at the single cell level.

New insights of Raman spectroscopy for oral clinical applications

Oral injuries are currently diagnosed by histopathological analysis of biopsy, which is an invasive procedure and does not give immediate results.

Ultra-filtration of human serum for improved quantitative analysis of low molecular weight biomarkers using ATR-IR spectroscopy

Monitoring of changes in the concentrations of the low molecular weight constituents enhanced by abundant proteins depletion.

Combining clinical variables to optimize prediction of antidepressant treatment outcomes

The outcome of treatment with antidepressants varies markedly across people with the same diagnosis. A clinically significant prediction of outcomes could spare the frustration of trial and error approach and improve the outcomes of major depressive disorder through individualized treatment selection. It is likely that a combination of multiple predictors is needed to achieve such prediction. We used elastic net regularized regression to optimize prediction of symptom improvement and remission during treatment with escitalopram or nortriptyline and to identify contributing predictors from a range of demographic and clinical variables in 793 adults with major depressive disorder. A combination of demographic and clinical variables, with strong contributions from symptoms of depressed mood, reduced interest, decreased activity, indecisiveness, pessimism and anxiety significantly predicted treatment outcomes, explaining 5-10% of variance in symptom improvement with escitalopram. Similar combinations of variables predicted remission with area under the curve 0.72, explaining approximately 15% of variance (pseudo R(2)) in who achieves remission, with strong contributions from body mass index, appetite, interest-activity symptom dimension and anxious-somatizing depression subtype. Escitalopram-specific outcome prediction was more accurate than generic outcome prediction, and reached effect sizes that were near or above a previously established benchmark for clinical significance. Outcome prediction on the nortriptyline arm did not significantly differ from chance. These results suggest that easily obtained demographic and clinical variables can predict therapeutic response to escitalopram with clinically meaningful accuracy, suggesting a potential for individualized prescription of this antidepressant drug.

Mid-infrared fibre evanescent wave spectroscopy of serum allows fingerprinting of the hepatic metabolic status in mice

Infrared spectroscopy from mice unveils a transition in the time course of steatosis.

Spectral pre and post processing for infrared and Raman spectroscopy of biological tissues and cells

This review presents the current understanding of the factors influencing the quality of spectra recorded and the pre-processing steps commonly employed to improve on spectral quality, as well as some of the most common techniques for classification and analysis of the spectral data for biomedical applications.

Developing and understanding biofluid vibrational spectroscopy: a critical review

Biofluid vibrational spectroscopy, a promising tool for rapid disease diagnosis.