Digital pathology is a computer-based imaging environment which enables the analysis and storage/retrieval of information from a digital slide. It is considered to be one of the most promising recent developments in diagnostic medicine, with the potential to provide better, quicker and more cost-effective diagnosis and prognosis, especially for managing diseases like cancers.
Digital pathology is closely associated with the term ‘virtual’ microscopy, since images are viewed remotely, without a microscope or slides, after being transferred over a hospital network or the Internet.

Whole slide imaging
A key technology driver of digital pathology is whole slide imaging (WSI), sometimes also described as whole slide digital imaging. WSI scans and converts specimen glass slides into digital images, which are made accessible by software for display on a computer monitor. Digitized slides allow analysis via computer algorithms, which automate the counting of structures and quantitatively classify tissue condition. This task is otherwise performed, painstakingly, by pathologists using a microscope to view, analyse and stain tissue slides.
The pace of development in virtual microscopy has accelerated. Recent advances in scanning technology allow for achieving over 100,000 dpi resolutions, in other words approaching the level of optical microscopes.

Cancer staging versus grading: the role of pathology
Digital pathology offers particular promise in the grading of tumours.  Tumour ‘grade’ is different from the ‘stage’ of a cancer.
Cancer stage refers to the size of a tumour and whether or not cancer cells have spread in the body.
Pathologists play a role in providing staging information, alongside physical examination, imaging and lab tests. Physicians select different combinations of each of these modalities for staging.
By contrast, grading is principally a pathologist’s area of expertise.

Grading scales
The grade describes a tumour and the likelihood of its growth and spread, based on the abnormality of tumour cells as seen under a microscope.
Tumours are typically graded on a 1-4 scale.  A lower grade indicates better prognosis, while higher grades tend to grow and spread more quickly, thus requiring more aggressive treatment. Grade 1 tumours are usually described as “well-differentiated” with tumour cells and tissue appearing close to normal. Grade 3 and 4 tumours look least like normal cells and tissue. They are often described as “poorly differentiated” or “undifferentiated,” and tend to grow and spread faster than tumours with a lower grade.

The pathologist and visual perspectives
One of the strongest arguments in favour of digital pathology is that microscopes require pathologists to always possess keen eyesight to determine the ‘differentiation’ required for assigning a grade to a tumour. Like many other healthcare professionals, pathologists are also burdened by heavy workloads. This, in turn, can impact on visual acuity and interpretation.

Digitized images, in contrast, can quantify the differentiation (and grading) process via algorithms, reduce the risk of human error and improve accuracy.

Glass slides and inconsistent interpretation
The challenge of consistency in pathology has been a vexing issue for some time. In March 2015, the ‘Journal of the American Medical Association’ (JAMA) published the results of a study to “quantify the magnitude of diagnostic disagreement among pathologists.” The study focused on pathologists interpreting breast biopsies from glass slides in eight US states and noted that although a breast pathology diagnosis provided “the basis for clinical treatment and management decisions,” its accuracy was “inadequately understood.” Among its findings – one in four cases did not show consonance of individual pathologists’ interpretations with expert consensus.
Disagreement with the reference diagnosis was statistically significant among pathologists who interpreted lower weekly case volumes or worked in smaller practices – confirming the observation about the inverse correlation between workload on the one side, and quality of eyesight on the other.

Digital pathology and second opinions
To address such variations in diagnoses, second opinions have become commonplace. However, for glass slides, a second opinion entails long lead times and complexities in a pathologist’s workflow (from glass packaging and transport, and at the other end, unpacking materials, verifying sample/reference case, registering the case in a laboratory information system etc.).  Many pathologists are forced to cope quietly with a difficult decision – weighing up the value of a second opinion against the extra waiting time for a patient.
Digital pathology streamlines access to second opinions, enabling quicker and more accurate delivery of diagnoses. Both these correlate strongly to successful treatment outcomes.  Telemedicine has taken explicit note about this potential. In 2014, the American Telemedicine Association published draft guidelines on the use of digital pathology in telemedicine.

Radiology and pathology: collaborative cousins
Pathology is involved in almost all cancer diagnoses.
Digital pathology is being seen as both a catalyst and enabler for more collaboration across specialties, beginning with radiology – one of the first fields to be digitized.
A September 2012 ‘BMC Medicine’ article titled ‘Integrating Pathology and Radiology Disciplines: An Emerging Opportunity?’ argues for an end to traditional pathology-radiology workflows where the two specialties “form the core of cancer diagnosis” but remain “ad hoc and occur in separate ‘silos’, even though “the opportunity for pathology-radiology integration to improve patient care is great, and more importantly, the tools to achieve this exist.”

DICOM and HIPAA
Until recently, digital pathology was hampered by a lack of standards for storing and transferring images, among other things, to be more in line with modern PACS systems storing radiology images. However, this was successfully addressed by DICOM (Digital Imaging and Communications in Medicine) supplement for digital pathology (No. 145), which was released in July 2010.
According to a May 2011 report in the ‘Journal of Pathology Informatics’, the DICOM supplement standard was hailed by “everyone involved in the field of digital pathology” since it made it easier for hospitals “to integrate digital pathology into their already established systems without adding too much overhead costs.”

Besides, it was seen to enable different vendors developing scanners “to upgrade their products to storage systems that are common across all systems.”
There is already sufficient integration between digital pathology systems (DPS) and anatomic pathology laboratory information systems (APLIS) to provide pathologists with access to images and image analysis data from either, and input it to a Patient Report. On the regulatory front, DPS vendors are also well placed to support HIPAA compliance by encrypting protected health information (PHI) metadata such as slide labels, hospital, patient and specimen information, etc.
A March 2007 issue of ‘Neuroimage’ points to another major attribute of digital pathology, namely the capacity for data mining.”

Digital pathology in medical education
So far, the key application of digital pathology has been in teaching. As the University of Minnesota observes, “virtual microscopes can transform traditional teaching methods by removing the reliance on physical space, equipment, and specimens to a model that is solely dependent upon computer-internet access. This rich database is enhanced with patient clinical presentations, laboratory data, comprehensive slide interpretations, and diagnoses.”
Also in the US, a  partnership between Oklahoma University Medical  Center (OUMC), the Children’s Hospital, and the University of Oklahoma College of Medicine, observes that digital pathology promotes efficiency and cost-effectiveness as a teaching tool as well as in using digital slides for consultation with patients referred to OUMC from other hospitals.

Barriers to digital pathology
Barriers to the more widespread implementation of digital pathology have also been recently assessed. These concern economics (mainly return on investment) and consistency and methodological robustness in WSI.

ROI
Unlike digital radiology which has a longer legacy and a stronger case for ROI (return on investment) – principally in terms of replacing film, the arguments for digital pathology are less obvious.
A study by a Swedish hospital found the following justifications for digital pathology: savings of time in administrative tasks (13%), slide review (6%) and supervision (3.1%), alongside an increase in efficiency of administrative tasks (100%), supervision (33%) and slide review (16%).
In terms of productivity per pathologist, the gain attained by digital pathology was 10%, while overall time savings were 24%.

Consistency in digital pathology interpretations
The second challenge facing digital pathology has been to determine the difference of interpretation of whole-slide images from glass-slide interpretation in difficult surgical cases, and the impact of such differences. This issue has been the subject of a study, with an article on the findings published in the December 2009 issue of the ‘Archives of Pathology & Laboratory Medicine’. 

Overall concordance between digital whole-slide and standard glass-slide interpretations was 91%, with agreement among digital, glass, and reference diagnoses in 85% of cases. 9% of digital cases were discordant with both reference and glass diagnoses. This was due to incorrect digital whole-slide interpretation, mainly because of issues such as fine resolution and navigating ability at high magnification.

FDA approval
One of the biggest obstructions to the growth of digital pathology has been the absence of approval by the US Food and Drug Administration (FDA) for primary diagnosis. Several EU countries allow pathologists to use WSI for primary diagnosis,  with some flexibility. For instance, in Sweden, slides are digitally scanned but also physically delivered to a consulting pathologist who has the choice to review the slides on screen, in the microscope, or both.
However, much of the technology development in digital pathology – as well as vendor interest – has originated in the US, and it is evident that freeing up digital pathology for primary diagnosis in that country would galvanize use worldwide.
US manufacturers have so far been able to market digital pathology technology for Research Use Only (RUO). Several vendors have also received one or more FDA 510 (k) clearances, with a key justification being manual and/or quantitative analysis of immunohistochemistry and/or in situ hybridization.
More developments are in the pipeline.
In February 2015, the FDA issued draft guidance for the technical performance assessment of digital pathology WSI devices. This followed an FDA Hematology and Pathology Devices Panel meeting six years previously to obtain industry feedback on replacing glass slides and conventional microscopy with whole slide images (WSI) for the purpose of rendering surgical pathology diagnosis.
On its part, the Digital Pathology Association (DPA) expects the draft guidance to lead to follow-on guidance and clarify the FDA’s expectations for WSI regulatory submissions, enabling increased access and adoption of digital pathology for clinical use.

Alzheimer’s disease (AD) is the most common cause of dementia. It is characterized by accumulation of β-amyloid (Aβ) peptides and tau proteins, loss of neurons and cognitive decline. It is difficult to diagnose AD in the early stages by clinical examination. Cerebrospinal fluid (CSF) biomarkers can be used to overcome this and could be useful in clinical practice, research and in trials of novel treatments.
 

by Philip Insel, Dr Rik Ossenkoppele and Dr Niklas Mattsson

Introduction
Alzheimer’s disease (AD) is a neurodegenerative disease that leads to cognitive impairment and ultimately dementia. The majority of the 45 million dementia patients world-wide have dementia due to AD [1]. In terms of brain pathology, AD is characterized by progressive accumulation of extracellular deposits of β-amyloid (Aβ) peptides in plaques and intracellular deposits of tau proteins in neurofibrillary tangles. The abnormal metabolism of Aβ and tau is believed to lead to impaired brain function, loss of synapses and neurons, and cognitive decline. The cognitive domain that is most severely impaired in most AD patients is episodic memory, but other domains such as language, visuo-spatial performance, behaviour and executive function may also become affected. In a subset of patients, deficits in non-memory domains are the dominating (early) features, as we will further discuss below.

Stages of Alzheimer’s disease
There is an increasing awareness among researchers and clinicians that AD starts many years before the onset of dementia. The first stage is thought to be asymptomatic, when pathologies accumulate silently in the brains of affected individuals for years or even decades before clinical symptoms emerge [2]. This is followed by an early clinical stage, which is characterized by objective cognitive impairment but still preserved function. This is often called mild cognitive impairment (MCI) due to AD, or prodromal AD [3]. The final stage is the classic stage, where AD has caused sufficient functional impairment for the patient to qualify for a dementia diagnosis.

Previously the procedure to diagnose AD was based solely on clinical examination and neuropsychological testing of the patient and interviews with proxies and caregivers. These methods are hampered by low sensitivity in the early stages of the disease and low specificity in the late stages of the disease. With the development of novel imaging and biochemistry technologies it is now possible to diagnose AD with the aid of direct evidence of relevant molecular pathologies in the brain. This is a conceptual leap that has revolutionized clinical AD research and is rapidly transforming clinical practice and clinical trial design. In this article we will discuss this development, with a focus on cerebrospinal fluid biomarkers for AD. Key points are summarized in Table 1.

The first steps
The amino acid sequences of Aβ and tau were identified in the mid-1980s [4, 5]. Early on it was suggested that a test for AD could be developed based on serum measurements of Aβ [4], but it was the discovery in the early 1990s that Aβ is secreted into the CSF that boosted the modern development of biochemical AD markers [6]. In the mid-1990s immunoassays (ELISAs) were developed for CSF Aβ1-42 (the prominent Aβ peptide isoform in Aβ plaques, typically reduced in AD patients compared to controls [7]), total-tau (T-tau, increased in AD patients compared to controls [8, 9]), and phosphorylated-tau (P-tau, increased in AD patients compared to other neurological diseases and controls [8]).

A window into the brain
Traditionally, AD could only be diagnosed at the dementia stage, and definite diagnosis was only possible through post-mortem analysis of brain tissue. The CSF biomarkers Aβ1-42, T-tau and P-tau (and imaging technologies not covered in this article) have made it possible to approach identification of AD brain pathology in living patients. Several studies have found that CSF Aβ1-42 is strongly related to the presence of brain Aβ pathology, quantified at autopsy [10] or in vivo using positron emission tomography imaging with tracers specific for fibrillar Aβ [11]. Likewise, although with less strong associations, CSF P-tau correlates with neocortical tangle pathology [12, 13], whereas CSF T-tau is more non-specifically increased in a number of neurological diseases, with the magnitude of the increase correlating with the size of the damaged tissue and the clinical outcome [14, 15]. CSF biomarkers enable detection of AD pathology in patients in early stages of the disease, when only mild symptoms or no clinical symptoms are present. As clinical diagnosis alone is inadequate in those early disease stages, biomarkers may be critical for an accurate diagnosis. CSF biomarkers may also increase the diagnostic accuracy of the diagnosis in advanced clinical stages, by providing biological evidence of AD related pathology. This helps in identifying the clinical syndrome and differentiating it from other neurological diseases.

Challenges
The field is rapidly advancing to overcome some hurdles that have prevented widespread implementation of CSF biomarkers. Problems with between-laboratory and between-assay variability in measurements have been noticed [16] and are being tackled by the development of certified reference procedures (based on selected reaction monitoring mass spectrometry [17]) and certified reference materials (created by the Institute of Reference Materials and Methods [18]). Development of fully automated assays will further bring down the variability and facilitate implementation of CSF biomarkers outside of expert centres (see www.neurochem.gu.se/TheAlzAssQCprogram for updated comparisons of different assay systems in a global quality control programme). In some countries a remaining obstacle is the unwillingness of medical practitioners to perform lumbar punctures, especially outside of highly specialized clinics. More training is needed to increase the familiarity of doctors with this procedure, and more education is needed to inform staff and patients that the procedure is safe. Headache is the only common complication (2–5 % incidence), but this is usually benign and treatable by common analgesics. Severe complications are extremely rare.

Alzheimer’s disease variants
CSF biomarkers of Aβ and tau may be particularly helpful to assist the diagnostic process in patients with a non-amnestic presentation of AD who may show substantial clinical overlap with patients experiencing non-AD types of dementia. Recently, there has been an increased awareness of these atypical presentations such as posterior cortical atrophy (PCA, ‘visual variant AD’ [19]), logopenic variant primary progressive aphasia (‘language variant AD’ [20]), and the behavioural/dysexecutive variant of AD [21]. As previous studies with small sample sizes have yielded conflicting results, we performed a study in 176 patients selected for abnormal CSF Aβ biomarkers to assess whether CSF T-tau and P-tau differ between atypical variants of AD [22]. Bootstrapping showed that the prevalence of abnormal T-tau and P-tau was ~80–90%, roughly equally distributed across AD phenotypes. This suggests that CSF T-tau and P-tau are equally useful in all clinical phenotypes of AD, which is compatible with current National Institute on Aging–Alzheimer’s Association (NIA-AA) and International Working Group for New Research Criteria for the Diagnosis of AD (IWG-2) diagnostic criteria.

Biomarkers in clinical trials
Biomarker measurement is a recent addition to AD clinical trials. The use of biomarkers is thought to improve trial design both in terms of subject selection and measurement of disease progression. It becomes particularly important in trials of disease-modifying treatments to recruit only those subjects with the target pathology of the therapy. Several recent failed trials may have been hindered by the inclusion of subjects without the underlying pathology [23]. Biomarkers are also less affected by measurement error compared with clinical outcomes and thus offer certain advantages in measuring progression over time, especially in early stages of disease [24]. In these early stages of the disease, before the onset of clinical symptoms, the ability of biomarkers to predict future pathology and cognitive decliners will aid in identifying those most in need of treatment. This will become especially important if intervention in the earliest stages of the disease, prior to substantial neurodegeneration, offers the best chance of a treatment to be effective.

Early treatment trials of anti-Aβ therapies employ thresholds to ensure recruitment of subjects with a minimal level of Aβ pathology. This threshold is frequently taken to be the level of amyloid pathology that most accurately distinguishes cases of AD from cognitively-normal controls [25]. However, if earlier treatment has a higher likelihood of success, identifying subjects with normal amyloid levels who are likely to have elevated levels in the future may be a further step toward early intervention. A recent study demonstrated that amyloid-negative subjects with low levels of CSF Aβ1-42 were much more likely to become amyloid-positive in the near term [26]. Individuals with CSF Aβ1-42 levels in the low normal range may be optimal candidates for early intervention trials aimed at halting further Aβ accumulation.

Conclusions
CSF biomarkers have helped to transform the diagnosis of AD from a clinical diagnosis to a biomarker-informed diagnosis based on molecular evidence of the underlying neuropathology. This has implications for research, where CSF biomarkers enable researchers to characterize subjects at all levels of cognitive function, in clinical practice, where CSF biomarkers aid doctors in diagnosis of AD versus other causes of cognitive impairment, and in the design of clinical trials, where CSF biomarkers may be used to enrich study populations and construct sensitive measures of outcomes to increase study power.

References
1. Alzheimer’s disease International: World Alzheimer Report 2015 (http://www.alz.co.uk/research/world-report-2015).
2. Jansen WJ, Ossenkoppele R, Knol DL, Tijms BM, et al. Prevalence of cerebral amyloid pathology in persons without dementia: a meta-analysis.
JAMA 2015; 313: 1924–1938.
3. Albert MS, DeKosky ST, Dickson D, Dubois B, et al. The diagnosis of mild cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. 2011; 7: 270–279.
4. Glenner GG, Wong CW. Alzheimer’s disease: initial report of the purification and characterization of a novel cerebrovascular amyloid protein. Biochem Biophys Res Commun. 1984; 120: 885–890.
5. Grundke-Iqbal I, Iqbal K, Quinlan M, Tung YC, et al. Microtubule-associated protein tau. A component of Alzheimer paired helical filaments. J Biol Chem. 1986; 261: 6084–6089.
6. Seubert P, Vigo-Pelfrey C, Esch F, Lee M, et al. Isolation and quantification of soluble Alzheimer’s beta-peptide from biological fluids. Nature 1992; 359: 325–327.
7. Motter R, Vigo-Pelfrey C, Kholodenko D, Barbour R, et al. Reduction of beta-amyloid peptide42 in the cerebrospinal fluid of patients with Alzheimer’s disease. Ann Neurol. 1995; 38: 643–648.
8. Blennow K, Wallin A, Agren H, Spenger C, et al. Tau protein in cerebrospinal fluid: a biochemical marker for axonal degeneration in Alzheimer disease? Mol Chem Neuropathol. 1995; 26: 231–245.
9. Vandermeeren M, Mercken M, Vanmechelen E, Six J, et al. Detection of tau proteins in normal and Alzheimer’s disease cerebrospinal fluid with a sensitive sandwich enzyme-linked immunosorbent assay. J Neurochem. 1993; 61: 1828–1834.
10. Strozyk D, Blennow K, White LR, Launer LJ. CSF Abeta 42 levels correlate with amyloid-neuropathology in a population-based autopsy study. Neurology. 2003; 60: 652–656.
11. Fagan AM, Mintun MA, Mach RH, Lee SY, et al. Inverse relation between in vivo amyloid imaging load and cerebrospinal fluid Abeta42 in humans. Ann Neurol. 2006; 59: 512–519.
12. Tapiola T, Alafuzoff I, Herukka SK, Parkkinen L, et al. Cerebrospinal fluid {beta}-amyloid 42 and tau proteins as biomarkers of Alzheimer-type pathologic changes in the brain. Arch Neurol. 2009; 66: 382–389.
13. Seppala TT, Nerg O, Koivisto AM, Rummukainen J, et al. CSF biomarkers for Alzheimer disease correlate with cortical brain biopsy findings. Neurology. 2012; 78: 1568–1575.
14. Hesse C, Rosengren L, Andreasen N, Davidsson P, et al. Transient increase in total tau but not phospho-tau in human cerebrospinal fluid after acute stroke. Neurosci Lett. 2001; 297: 187–190.
15. Otto M, Wiltfang J, Tumani H, Zerr I, et al. Elevated levels of tau-protein in cerebrospinal fluid of patients with Creutzfeldt-Jakob disease. Neurosci Lett. 1997; 225: 210–212.
16. Mattsson N, Andreasson U, Persson S, Carrillo MC, et al. CSF biomarker variability in the Alzheimer’s Association quality control program. Alzheimers Dement J Alzheimers Assoc. 2013; 9: 251–261.
17. Leinenbach A, Pannee J, Dülffer T, Huber A, et al. Mass Spectrometry-Based Candidate Reference Measurement Procedure for Quantification of Amyloid-β in Cerebrospinal Fluid. Clin Chem. 2014; 60: 987–994.
18. Mattsson N, Zetterberg H. What is a certified reference material? Biomark Med. 2012; 6: 369–370.
19. Crutch SJ, Lehmann M, Schott JM, Rabinovici GD, et al. Posterior cortical atrophy. Lancet Neurol. 2012; 11: 170–178.
20. Gorno-Tempini ML, Hillis AE, Weintraub S, Kertesz A, et al. Classification of primary progressive aphasia and its variants. Neurology 2011; 76: 1006–1014.
21. Ossenkoppele R, Pijnenburg YAL, Perry DC, Cohn-Sheehy BI, et al. The behavioural/dysexecutive variant of Alzheimer’s disease: clinical, neuroimaging and pathological features. Brain J Neurol. 2015; 138: 2732–2749.
22. Ossenkoppele R, Mattsson N, Teunissen CE, Barkhof F, et al. Cerebrospinal fluid biomarkers and cerebral atrophy in distinct clinical variants of probable Alzheimer’s disease. Neurobiol Aging 2015; 36: 2340–2347.
23. Karran E, Hardy J. Antiamyloid therapy for Alzheimer’s disease—are we on the right road? N Eng J Med. 2014; 370: 377–378.
24. Hendrix SB. Measuring clinical progression in MCI and pre-MCI populations: Enrichment and optimizing clinical outcomes over time. Alzheimer’s Res Ther. 2012; 4: 24.
25. Shaw LM, Vanderstichele H, Knapik‐Czajka M, Clark CM, et al. Cerebrospinal fluid biomarker signature in Alzheimer’s disease neuroimaging initiative subjects. Ann Neurol. 2009; 65: 403–413.
 26. Mattsson N, Insel PS, Donohue M, Jagust W, et al. Predicting reduction of cerebrospinal fluid β-amyloid 42 in cognitively healthy controls. JAMA Neurol. 2015; 72: 554–560.

The authors
Philip Insel^1,2,3 MSs; Rik Ossenkoppele^4,5 PhD; Niklas Mattsson*¹ MD, PhD
¹ Clinical Memory Research Unit, Faculty of Medicine, Lund University, Lund, Sweden
² Center for Imaging of Neurodegenerative Diseases, Department of Veterans Affairs Medical Center, San Francisco, CA, USA
³ Department of Radiology and Biomedical Imaging, University of California, San Francisco, CA, USA
⁴ Alzheimer Center and Department of Neurology, Neuroscience Campus Amsterdam, VU University Medical Center, Amsterdam, the Netherlands
⁵ Memory and Aging Center, University of California San Francisco, San Francisco, CA, USA

*Corresponding author
E-mail: Niklas.mattsson@med.lu.se