Research Grants Awarded
Accurate, Reproducible And Automated Breast Histopathology Without The Use Of Stains Or Human Input
Investigator Initiated Research
Histopathology is central to diagnoses, treatment and research in breast cancer. Imprecision in identifying at-risk patients (screening), grading tumors (diagnosis) and predicting outcome (prognosis), however, provides less than optimal management of the disease. Suboptimal screening implies a large workload for pathology. Varied tumor morphology and molecular compositions present a diagnostic challenge that is complicated by the limited amounts of tissue available in modern biopsies. Imprecise and variable diagnoses complicate therapy while serving to limit prognosis capability; they also present a challenge in research into the etiology of the disease, development of therapies or drugs and evaluation of animal models. Hence, new technologies are urgently needed that can aid in histologic recognition in the laboratory, in intra-operative settings and for in-vivo guidance. The purpose of this proposal is to develop and validate spectroscopic imaging technology to address this need.
Combining chemical and morphologic information obtained from novel infrared spectroscopic imaging, the objective of this project is to establish automated, objective and accurate breast histopathologic recognition. The hypothesis of this study is that breast pathology practice can be significantly aided by the development of these tools for automated pathology and that progress towards better prognostication can be achieved by determining chemical and morphologic markers not currently used. A validated approach to pathology, that is ready for clinical trials, is expected to be the outcome. A second outcome is expected to be the design parameters to translate this technology for systems that can evaluate needle biopsies, perform intra-operative pathology and guide needle biopsy procedures accurately.
1) Develop high-resolution FT-IR imaging, which can be used for future intra-operative evaluations, and obtain data from fixed, surgical breast tissue samples
2) Demonstrate automated breast histopathology and determine limits of FT-IR imaging in diagnosing breast cancer
3) Detect epithelial cells in SLN sections and predict performance for intra-operative SLN evaluation, predict optimal design for in-situ pathology with a needle probe and needle biopsy evaluation.
Anonymized radical prostatectomy tissue specimens from the Feinberg School of Medicine, Northwestern University archives and associated clinical data will be selected under the guidance of an experienced, practicing pathologist (collaborator). Cases will span the range of clinical presentations from women who have been followed since diagnosis and for whom outcome data, death due to cancer or distant metastases development information is available. At least 375 formalin-fixed, paraffin-embedded tissue blocks will be processed as per standard histologic protocols and an unstained section will be imaged using FTIR spectroscopy. Sequential sections, stained with Hematoxylin and Eosin (H&E), will be used for pathologist review. FTIR imaging data will be recorded using a novel high-resolution approach. We will also develop a feedback loop and mathematical processing to control data quality. Quantitative image and spectral data analysis will be carried out on the FTIR spectral data by developing supervised (Bayesian) classification algorithms for automated computer analyses. Samples from each pathologic cohort will be divided into a training (calibration) set and validation set. Model optimization to determine the best prediction will be undertaken in the training set and the developed models will be tested in a blinded validation procedure to determine performance. We will use multivariate analysis and regression models to determine the relation between spectrally-derived markers, both spectral and morphometric, and breast cancer outcome. Given the size and quality of follow-up, the diversity of tissue and the large numbers of observations, we have sufficient robustness to provide a tool useful in practice. Results will be extended to predict performance of intra-operative and in-vivo devices. Parameters for an intra-operative imaging tool for margin determination and lymph node involvement will be determined. Performance of automated classifier in translation of the approach to biopsy tissue will be determined. Last, design parameters will be obtained for a tool to provide in-situ histologic guidance of needle biopsy procedures.
The primary goal of this project is to test and validate a novel imaging approach that provides structural and chemical data to objectively characterize breast cancer, without the use of reagents or human input. Being compatible with current pathologic practice, the approach to cancer diagnosis can aid in mitigating laboratory variation, workload and pathology errors. The approach also represents a unique opportunity to aid the prediction of aggressive breast cancer by permitting the visualization of panels of cells without immunohistochemical staining. Better diagnoses enabled by this project and translation of the advances here to future technologies will help in correct and timely diagnoses, in evaluating metastatic progression risk and in locating tumors in-vivo. In research, the imaging tools present an opportunity to increase significantly the throughput of histopathologic characterization and the consistency of such determinations, in turn leading to better understanding and treatment of the disease.
Breast cancer is the most commonly diagnosed cancer among US women and is a major cause of mortality. With 1 in 8 women expected to develop breast cancer during their lifetime, public awareness is heightened and screening programs are widespread. Suspicious results upon screening necessitate a biopsy to diagnose or rule out cancer. The recognition of tumors in biopsies, using dyes to stain tissue and a microscope to observe structures, is the current gold standard for diagnosis. The recognition of disease and predicting its progression is an imperfect science, however, and contains several open questions.
First, issues of variability in diagnoses, high workload and errors are important in pathologic exams, directly affecting patient therapy and increasing the social, legal and economic burden of the disease. Can we provide new technologies in pathology that improve clinical decisions by providing more accurate, consistent, objective and economical diagnoses? Second, we seek to examine if such technology can be applied in three distinct areas, ranging from screening to treatment:
1) Can the combination of chemistry and imaging provide accurate determinations that aid pathologists in making correct decisions for treatment, research and outcome predictions?
2) Can the concept of pathology without stains be extended to the operating theater to enable comprehensive diagnoses, such that no evidence of metastatic disease is missed in lymph nodes and that margins are accurately determined for excised tissue?
3) Can the same results be obtained in real-time by optics in a needle to guide biopsy? Can important pathologic diagnoses be accomplished in small needle biopsies with the same quality as surgical specimens?
This project seeks to provide a novel approach to aid the resolution of these questions by developing imaging technology. The unique aspect of this imaging technology is that it combines both structure and chemistry to provide not just images, but actual diagnoses in a form that is easily understood by clinicians.
We have developed Fourier transform infrared (FTIR) spectroscopic imaging, a new imaging technique that merges the power of microscopy with the chemical analysis capabilities of spectroscopy. Directly measuring the chemical composition of microscopic regions of tissue, a structural and chemical picture is obtained. Using computer algorithms, the picture can be translated into pathologic knowledge. Hence, no specialized chemical reagents or human interaction is required for this automated approach and the resulting diagnosis is objective, reproducible and consistent.
This research project is a collaboration between engineers and pathologists to develop a clinical tool for pathology that seeks to directly provide information to address the questions above. We propose to develop high-resolution and fast imaging that is needed for breast pathology. We will then select archival tissue with long term follow-up. The tissue will be processed as per standard pathology protocols and imaged using FTIR imaging without staining. A corresponding stained section will be used for routine pathology exams. Spectral and morphologic data from FTIR imaging will be used, first, to construct a prediction protocol for computer recognition of disease. A second set of samples will be imaged and will be used to validate the developed protocol. Results from the evaluation will be used to determine whether the system can accurately and sensitively detect epithelial cells in lymph nodes intra-operatively. The results also form the basis of analysis of needle biopsy material and predicting its accuracy. Last, the results will allow us here to design a needle probe that can evaluate breast lesions in-situ as guides to accurate biopsy.
This project is a critical step in providing a novel tool for pathology practice. The availability of objective, quantitative, consistent and automated recognition of pathologic status will aid in improving pathology practice, improving the quality of information available to oncologists and will provide a novel tool for breast cancer research. Since the methodology is entirely compatible with pathology, a large validation trial following the successful completion of this project, a matter of a few years, will provide sufficient information to determine if the technology can be translated to practice. The achievement of the goals of this project will directly provide valuable input to guide the patient and physician in making decisions that impact therapy and quality of life for a majority of patients at risk of breast cancer.