Presentation on theme: "Interpretation of Histology and Cytology Glass Sides: Features Essential for Diagnosis Ulysses J. Balis, MD Director, Division of Pathology Informatics."— Presentation transcript:
Interpretation of Histology and Cytology Glass Sides: Features Essential for Diagnosis Ulysses J. Balis, MD Director, Division of Pathology Informatics Associate Professor of Pathology Department of Pathology University of Michigan email@example.com
Outline Historical context of optical microscopy use in pathology Inherent limitations in optical systems Overview of histologic features essential for diagnosis Overview of the WSI combined optical-digital dataflow model and the potential data artifacts that it generates Limitations in digital representation of optical imagery Potential real-world consequences of digitization of optical microscopy : effect on Interpretive accuracy Current use-cases where WSI modality workflow may not be equivalent to optical systems, and why this is so Closing Remarks
Historical Context of Optical Microscopy Use in Pathology Several hundred years of operational experience Modality has been largely stable for over a hundred years, with the advent of corrective optics in the late 19 th century Recent innovations, largely coupled to contrast and field-of- view improvements, have further extended the utility of optical microscopy Historically has been exempted from device certification or precise minimum performance requirements Appropriate functioning of microscopic equipment is based on the “guild model,” where “each artisan is expected to know their tools.”
Characterizing the Microscope’s Optical Properties Available magnifications Diffraction-limited capabilities Contrast Field of view Numerical Aperture Color Correction Spatial Aberration correction Lighting and uniformity issues
Contemporary Microscopy in Surgical Pathology No regulatory oversight of – Equipment minimum contrast and resolving capabilities – Periodic assessment for conformance to a minimum performance standard (again, recognizing that there is no standard) – Suitability of the modality for consistent diagnostic performance with various classes of subject matter No agreed upon minimum set of functional capabilities or consistently rendered ergonomic interface (compare and contrast with standard-work practices advocated and then implemented by the Anesthesia Safety Foundation) No initial or periodic competency testing of users’ abilities to calibrate the microscope to obtain its best possible diffraction-limited image (Köhler Illumination)
Inherent Limitations in Optical Systems Diffraction limitation of light – Length scale hypothesis of diagnosis – Line/feature-resolving capability – Ability to consistently realize a target point-spread function in mass-produced optics and optical instruments (functional variability between similar models) – Inverse relationship between depth of field capability and numerical aperture. Contrast – Section thickness and quality of cut – Histologic stain performance and variation Overall calibration of the microscope – Uniformity of light source – Normalized color temperature – Kohler illumination Field of view capabilities Additional Metric of Performance: Ergonomics of the
Overview of Histologic Features Essential for Diagnosis Length Scale Issues – Diagnosis by microscopy is a multi-resolution (magnification) process – Effective optical capture is essential across almost two orders of magnitude of spatial resolution (20x to 1000x total system magnification) Pattern Recognition of key histologic architecture at low magnification Pattern recognition of combined histological / architectural and cytological features at intermediate magnification Cytological feature classification at high magnification The microscopy-based diagnostic process typically commences at the lowest magnification and proceeds to successively higher optical power.
Overview of Histologic Features Essential for Diagnosis Histologic Structures – Larger (lower magnification) – Spans multiple cells – High variability in architecture for many diagnostic entities Cell patterns – Glandular structures – Often with significant depth of field – Cell patterns often constrained between cases and diagnostic entities
Overview of the WSI combined optical-digital dataflow model and the potential data artifacts that it generates Light source columniation Section or cellular transillumination Initial capture of microscopic image Image shuttle to digital imaging device/surface Analog to digital conversion WSI file storage (with or without perceptually loss-based compression) Rendering of final image on display system Total system image quality is a function of error terms imparted by each of the above process steps, with degradation in earlier steps propagating throughout the overall image rendering process.
Limitations in digital representation of optical imagery Excessively light or dark subject matter – No perceived texture or structure – With typical 8-bit /color channel acquisition systems (24 bits total)) the dynamic range of digitally captured micrographs is considerably less than that of direct visualization by optical light microscopy Lack of contrast in subject matter – No perceived texture or structure Inadequate spatial sampling – No perceived texture or structure for intended length-scale of subject matter (sub-sampling) The “Artisan” model is no longer effective, as pathologists and cytotechnologists are unqualified to quantitatively assess the performance of the digital data path of the WSI modality It then follows that adequate dynamic range and spatial resolution of the digital capture system are essential attributes for enabling digital images to be diagnostically equivalent to optical microscopy. It also follows that some absolute metrics of WSI system performance would be helpful in compensating for the inherent lack of an artisan model with the digital components inherent in the data flow model.
Potential Real-world Consequences of Digitization of Optical Microscopy : Effect on Interpretive Accuracy Length scale issues are always at play At low magnification, a low NA system might generate sufficient field of view but inadequate spatial resolution (exactly the historical challenge of NTSC-based telepathology, a decade ago) vs. optical microscopy, where low magnification images exhibited much higher equivalent NA, leading to higher resolution and greater diagnostic content. The compensatory response to such degraded digital subject matter has been the falling back to dependence on use of higher magnifications. This response, in turn, sacrifices field of view and thus, creates the following undesirable eventualities: – There is a need to review more fields of view – Low magnification review is absent of higher resolution cellular detail. – Higher magnification cellular detail is absent of the benefit of a larger field of view. With newer systems that capture a high-NA equivalent image at low magnification, (e.g. by numerical sub-sampling of high magnification datasets), it is possible to render low magnification images in very high quality (only limited by the display modality)
Current Example Use-cases Where WSI-Based Modality Workflow May Not Be Equivalent to Optical Systems, and Why This is So Hematopathology – Resolution-based limitations in the setting of relatively uncommon use of newer, oil-based whole slide scanners Pigmented Lesions – Inadequate dynamic range to completely capture diagnostically required contrast detail General Cytopathology – Inadequate spatial resolution and – Inadequate sampling of multiple focal planes in the setting of lesions with significant depth of field lymphoglandular lesions Rosette structures Other multi-cellular structures where cohesiveness plays a role
Closing Remarks Given that optical microscopy works remarkably well for use as a diagnostic tool, despite its relative lack of standardization and functional certification, there is inherent robustness in the human visual diagnostic interpretive process. When digitally rendered histology and cytology images exceed a viewer’s ability to adequately discriminate requisite diagnostic elements, there is already a natural tendency in most pathologists to defer to diagnosis by light microscopy, making use of WSI an inherently self-selecting diagnostic medium. As WSI technology continues to make strides in resolution, depth of field, dynamic range and display performance, use-cases currently unapproachable for diagnostic use will incrementally be added to the acceptable repertoire of use. The pathology field at large is as equally ignorant of the performance characteristics of our contemporary and disparate microscopy optical systems as are we of our newer WSI systems; more knowledge should be gathered in this area as well. Ultimately, WSI technology will likely exceed conventional optical systems in overall capability, owing to their ability to provision for arbitrarily high field of view in the setting of constrained high numerical aperture (affording the creation of so-called diagnostic walls). Adoption of WSI technology should be validated on an organ system/ diagnostic entity basis, by use of kappa statistical approaches in concert with ROC analysis.