Ki67 Quantitative Interpretation: Insights using Image Analysis

Background:

Proliferation markers, especially Ki67, are increasingly important in diagnosis and prognosis. The best method for calculating Ki67 is still the subject of debate.

Materials and Methods:

We evaluated an image analysis tool for quantitative interpretation of Ki67 in neuroendocrine tumors and compared it to manual counts. We expanded a primary digital pathology platform to include the Leica Biosystems image analysis nuclear algorithm. Slides were digitized using a Leica Aperio AT2 Scanner and accessed through the Cerner CoPath LIS interfaced with Aperio eSlideManager through Aperio ImageScope. Selected regions of interest (ROIs) were manually defined and annotated to include tumor cells only; they were then analyzed with the algorithm and by four pathologists counting on printed images. After validation, the algorithm was used to examine the impact of the size and number of areas selected as ROIs.

Results:

The algorithm provided reproducible results that were obtained within seconds, compared to up to 55 min of manual counting that varied between users. Benefits of image analysis identified by users included accuracy, time savings, and ease of viewing. Access to the algorithm allowed rapid comparisons of Ki67 counts in ROIs that varied in numbers of cells and selection of fields, the outputs demonstrated that the results vary around defined cutoffs that provide tumor grade depending on the number of cells and ROIs counted.

Conclusions:

Digital image analysis provides accurate and reproducible quantitative data faster than manual counts. However, access to this tool allows multiple analyses of a single sample to use variable numbers of cells and selection of variable ROIs that can alter the result in clinically significant ways. This study highlights the potential risk of hard cutoffs of continuous variables and indicates that standardization of number of cells and number of regions selected for analysis should be incorporated into guidelines for Ki67 calculations.

Visualizing histopathologic deep learning classification and anomaly detection using nonlinear feature space dimensionality reduction

BACKGROUND

There is growing interest in utilizing artificial intelligence, and particularly deep learning, for computer vision in histopathology. While accumulating studies highlight expert-level performance of convolutional neural networks (CNNs) on focused classification tasks, most studies rely on probability distribution scores with empirically defined cutoff values based on post-hoc analysis. More generalizable tools that allow humans to visualize histology-based deep learning inferences and decision making are scarce.

RESULTS

Here, we leverage t-distributed Stochastic Neighbor Embedding (t-SNE) to reduce dimensionality and depict how CNNs organize histomorphologic information. Unique to our workflow, we develop a quantitative and transparent approach to visualizing classification decisions prior to softmax compression. By discretizing the relationships between classes on the t-SNE plot, we show we can super-impose randomly sampled regions of test images and use their distribution to render statistically-driven classifications. Therefore, in addition to providing intuitive outputs for human review, this visual approach can carry out automated and objective multi-class classifications similar to more traditional and less-transparent categorical probability distribution scores. Importantly, this novel classification approach is driven by a priori statistically defined cutoffs. It therefore serves as a generalizable classification and anomaly detection tool less reliant on post-hoc tuning.

CONCLUSION

Routine incorporation of this convenient approach for quantitative visualization and error reduction in histopathology aims to accelerate early adoption of CNNs into generalized real-world applications where unanticipated and previously untrained classes are often encountered.

Integrated Pathology Informatics Enables High-Quality Personalized and Precision Medicine: Digital Pathology and Beyond

CONTEXT

- The critical role of pathology in diagnosis, prognosis, and prediction demands high-quality subspecialty diagnostics that integrates information from multiple laboratories.

OBJECTIVE

- To identify key requirements and to establish a systematic approach to providing high-quality pathology in a health care system that is responsible for services across a large geographic area.

DESIGN

- This report focuses on the development of a multisite pathology informatics platform to support high-quality surgical pathology and hematopathology using a sophisticated laboratory information system and whole slide imaging for histology and immunohistochemistry, integrated with ancillary tools, including electron microscopy, flow cytometry, cytogenetics, and molecular diagnostics.

RESULTS

- These tools enable patients in numerous geographic locations access to a model of subspecialty pathology that allows reporting of every specimen by the right pathologist at the right time. The use of whole slide imaging for multidisciplinary case conferences enables better communication among members of patient care teams. The system encourages data collection using a discrete data synoptic reporting module, has implemented documentation of quality assurance activities, and allows workload measurement, providing examples of additional benefits that can be gained by this electronic approach to pathology.

CONCLUSION

- This approach builds the foundation for accurate big data collection and high-quality personalized and precision medicine.

Raman spectroscopy: a real-time tool for identifying microcalcifications during stereotactic breast core needle biopsies

Microcalcifications are an early mammographic sign of breast cancer and a target for stereotactic breast needle biopsy. We present here a Raman spectroscopic tool for detecting microcalcifications in breast tissue based on their chemical composition. We collected ex vivo Raman spectra from 159 tissue sites in fresh stereotactic breast needle biopsies from 33 patients, including 54 normal sites, 75 lesions with microcalcifications and 30 lesions without microcalcifications. Application of our Raman technique resulted in a positive predictive value of 97% for detecting microcalcifications. This study shows that Raman spectroscopy has the potential to detect microcalcifications during stereotactic breast core biopsies and provide real-time feedback to radiologists, thus reducing non-diagnostic and false negative biopsies.

Multimodal spectroscopy detects features of vulnerable atherosclerotic plaque

Early detection and treatment of rupture-prone vulnerable atherosclerotic plaques is critical to reducing patient mortality associated with cardiovascular disease. The combination of reflectance, fluorescence, and Raman spectroscopy-termed multimodal spectroscopy (MMS)-provides detailed biochemical information about tissue and can detect vulnerable plaque features: thin fibrous cap (TFC), necrotic core (NC), superficial foam cells (SFC), and thrombus. Ex vivo MMS spectra are collected from 12 patients that underwent carotid endarterectomy or femoral bypass surgery. Data are collected by means of a unitary MMS optical fiber probe and a portable clinical instrument. Blinded histopathological analysis is used to assess the vulnerability of each spectrally evaluated artery lesion. Modeling of the ex vivo MMS spectra produce objective parameters that correlate with the presence of vulnerable plaque features: TFC with fluorescence parameters indicative of collagen presence; NC∕SFC with a combination of diffuse reflectance β-carotene∕ceroid absorption and the Raman spectral signature of lipids; and thrombus with its Raman signature. Using these parameters, suspected vulnerable plaques can be detected with a sensitivity of 96% and specificity of 72%. These encouraging results warrant the continued development of MMS as a catheter-based clinical diagnostic technique for early detection of vulnerable plaques.

Diagnosing breast cancer using Raman spectroscopy: prospective analysis

We present the first prospective test of Raman spectroscopy in diagnosing normal, benign, and malignant human breast tissues. Prospective testing of spectral diagnostic algorithms allows clinicians to accurately assess the diagnostic information contained in, and any bias of, the spectroscopic measurement. In previous work, we developed an accurate, internally validated algorithm for breast cancer diagnosis based on analysis of Raman spectra acquired from fresh-frozen in vitro tissue samples. We currently evaluate the performance of this algorithm prospectively on a large ex vivo clinical data set that closely mimics the in vivo environment. Spectroscopic data were collected from freshly excised surgical specimens, and 129 tissue sites from 21 patients were examined. Prospective application of the algorithm to the clinical data set resulted in a sensitivity of 83%, a specificity of 93%, a positive predictive value of 36%, and a negative predictive value of 99% for distinguishing cancerous from normal and benign tissues. The performance of the algorithm in different patient populations is discussed. Sources of bias in the in vitro calibration and ex vivo prospective data sets, including disease prevalence and disease spectrum, are examined and analytical methods for comparison provided.

A multimodal spectroscopy system for real-time disease diagnosis

The combination of reflectance, fluorescence, and Raman spectroscopy-termed multimodal spectroscopy (MMS)-provides complementary and depth-sensitive information about tissue composition. As such, MMS is a promising tool for disease diagnosis, particularly in atherosclerosis and breast cancer. We have developed an integrated MMS instrument and optical fiber spectral probe for simultaneous collection of all three modalities in a clinical setting. The MMS instrument multiplexes three excitation sources, a xenon flash lamp (370-740 nm), a nitrogen laser (337 nm), and a diode laser (830 nm), through the MMS probe to excite tissue and collect the spectra. The spectra are recorded on two spectrograph/charge-coupled device modules, one optimized for visible wavelengths (reflectance and fluorescence) and the other for the near-infrared (Raman), and processed to provide diagnostic parameters. We also describe the design and calibration of a unitary MMS optical fiber probe 2 mm in outer diameter, containing a single appropriately filtered excitation fiber and a ring of 15 collection fibers, with separate groups of appropriately filtered fibers for efficiently collecting reflectance, fluorescence, and Raman spectra from the same tissue location. A probe with this excitation/collection geometry has not been used previously to collect reflectance and fluorescence spectra, and thus physical tissue models ("phantoms") are used to characterize the probe's spectroscopic response. This calibration provides probe-specific modeling parameters that enable accurate extraction of spectral parameters. This clinical MMS system has been used recently to analyze artery and breast tissue in vivo and ex vivo.

Diagnosing breast cancer using diffuse reflectance spectroscopy and intrinsic fluorescence spectroscopy

Using diffuse reflectance spectroscopy and intrinsic fluorescence spectroscopy, we have developed an algorithm that successfully classifies normal breast tissue, fibrocystic change, fibroadenoma, and infiltrating ductal carcinoma in terms of physically meaningful parameters. We acquire 202 spectra from 104 sites in freshly excised breast biopsies from 17 patients within 30 min of surgical excision. The broadband diffuse reflectance and fluorescence spectra are collected via a portable clinical spectrometer and specially designed optical fiber probe. The diffuse reflectance spectra are fit using modified diffusion theory to extract absorption and scattering tissue parameters. Intrinsic fluorescence spectra are extracted from the combined fluorescence and diffuse reflectance spectra and analyzed using multivariate curve resolution. Spectroscopy results are compared to pathology diagnoses, and diagnostic algorithms are developed based on parameters obtained via logistic regression with cross-validation. The sensitivity, specificity, positive predictive value, negative predictive value, and overall diagnostic accuracy (total efficiency) of the algorithm are 100, 96, 69, 100, and 91%, respectively. All invasive breast cancer specimens are correctly diagnosed. The combination of diffuse reflectance spectroscopy and intrinsic fluorescence spectroscopy yields promising results for discrimination of breast cancer from benign breast lesions and warrants a prospective clinical study.

In vivo margin assessment during partial mastectomy breast surgery using raman spectroscopy

We present the first demonstration of in vivo collection of Raman spectra of breast tissue. Raman spectroscopy, which analyzes molecular vibrations, is a promising new technique for the diagnosis of breast cancer. We have collected 31 Raman spectra from nine patients undergoing partial mastectomy procedures to show the feasibility of in vivo Raman spectroscopy for intraoperative margin assessment. The data was fit with an established model, resulting in spectral-based tissue characterization in only 1 second. Application of our previously developed diagnostic algorithm resulted in perfect sensitivity and specificity for distinguishing cancerous from normal and benign tissues in our small data set. Significantly, we have detected a grossly invisible cancer that, upon pathologic review, required the patient to undergo a second surgical procedure. Had Raman spectroscopy been used in a real-time fashion to guide tissue excision during the procedure, the additional reexcision surgery might have been avoided. These preliminary findings suggest that Raman spectroscopy has the potential to lessen the need for reexcision surgeries resulting from positive margins and thereby reduce the recurrence rate of breast cancer following partial mastectomy surgeries.