Automated identification of glomeruli and synchronised review of special stains in renal biopsies by machine learning and slide registration: a cross-institutional study.

AIMS: Machine learning in digital pathology can improve efficiency and accuracy via prescreening with automated feature identification. Studies using uniform histological material have shown promise. Generalised application requires validation on slides from multiple institutions. We used machine learning to identify glomeruli on renal biopsies and compared performance between single and multiple institutions. METHODS AND RESULTS: Randomly selected, adequately sampled renal core biopsy cases (71) consisting of four stains each (haematoxylin and eosin, trichrome, silver, periodic acid Schiff) from three institutions were digitised at ×40. Glomeruli were manually annotated by three renal pathologists using a digital tool. Cases were divided into training/validation (n = 52) and evaluation (n = 19) cohorts. An algorithm was trained to develop three convolutional neural network (CNN) models which tested case cohorts intra- and inter-institutionally. Raw CNN search data from each of the four slides per case were merged into composite regions of interest containing putative glomeruli. The sensitivity and modified specificity of glomerulus detection (versus annotated truth) were calculated for each model/cohort. Intra-institutional (3) sensitivity ranged from 90 to 93%, with modified specificity from 86 to 98%. Interinstitutional (1) sensitivity was 77%, with modified specificity 97%. Combined intra- and inter-institutional (1) sensitivity was 86%, with modified specificity 92%. CONCLUSIONS: Feature detection sensitivity degrades when training and test material originate from different sites. Training using a combined set of digital slides from three institutions improves performance. Differing histology methods probably account for algorithm performance contrasts. Our data highlight the need for diverse training sets for the development of generalisable machine learning histology algorithms.

Computational Cytology: Lessons Learned from Pap Test Computer-Assisted Screening.

BACKGROUND: In the face of rapid technological advances in computational cytology including artificial intelligence (AI), optimization of its application to clinical practice would benefit from reflection on the lessons learned from the decades-long journey in the development of computer-assisted Pap test screening. SUMMARY: The initial driving force for automated screening in cytology was the overwhelming number of Pap tests requiring manual screening, leading to workflow backlogs and incorrect diagnoses. Several companies invested resources to address these concerns utilizing different specimen processing techniques and imaging systems. However, not all companies were commercially prosperous. Successful implementation of this new technology required viable use cases, improved clinical outcomes, and an acceptable means of integration into the daily workflow of cytopathology laboratories. Several factors including supply and demand, Food and Drug Administration (FDA) oversight, reimbursement, overcoming learning curves and workflow changes associated with the adoption of new technology, and cytologist apprehension, played a significant role in either promoting or preventing the widespread adoption of automated screening technologies. Key Messages: Any change in health care, particularly those involving new technology that impacts clinical workflow, is bound to have its successes and failures. However, perseverance through learning curves, optimizing workflow processes, improvements in diagnostic accuracy, and regulatory and financial approval can facilitate widespread adoption of these technologies. Given their history with successfully implementing automated Pap test screening, cytologists are uniquely positioned to not only help with the development of AI technology for other areas of pathology, but also to guide how they are utilized, regulated, and managed.

Using Image Registration and Machine Learning to Develop a Workstation Tool for Rapid Analysis of Glomeruli in Medical Renal Biopsies.

BACKGROUND: Prescreening of biopsies has the potential to improve pathologists' workflow. Tools that identify features and display results in a visually thoughtful manner can enhance efficiency, accuracy, and reproducibility. Machine learning for detection of glomeruli ensures comprehensive assessment and registration of four different stains allows for simultaneous navigation and viewing. METHODS: Medical renal core biopsies (4 stains each) were digitized using a Leica SCN400 at ×40 and loaded into the Corista Quantum research platform. Glomeruli were manually annotated by pathologists. The tissue on the 4 stains was registered using a combination of keypoint- and intensity-based algorithms, and a 4-panel simultaneous viewing display was created. Using a training cohort, machine learning convolutional neural net (CNN) models were created to identify glomeruli in all stains, and merged into composite fields of views (FOVs). The sensitivity and specificity of glomerulus detection, and FOV area for each detection were calculated. RESULTS: Forty-one biopsies were used for training (28) and same-batch evaluation (6). Seven additional biopsies from a temporally different batch were also evaluated. A variant of AlexNet CNN, used for object recognition, showed the best result for the detection of glomeruli with same-batch and different-batch evaluation: Same-batch sensitivity 92%, "modified" specificity 89%, average FOV size represented 0.8% of the total slide area; different-batch sensitivity 90%, "modified" specificity 98% and average FOV size 1.6% of the total slide area. CONCLUSIONS: Glomerulus detection in the best CNN model shows that machine learning algorithms may be accurate for this task. The added benefit of biopsy registration with simultaneous display and navigation allows reviewers to move from one machine-generated FOV to the next in all 4 stains. Together these features could increase both efficiency and accuracy in the review process.

Advanced imaging technology applications in cytology.

Novel techniques have been developed to image cells at cellular and subcellular levels. They allow images to be analyzed with ultra-high resolution, in 2D and/or 3D. Several of these tools have been tested on cytology specimens demonstrating emerging applications that are likely to change the field of cytopathology. This review covers several of these advanced imaging methods. The use of optical coherence tomography to perform optical biopsies during endoscopic ultrasound procedures or visualize cells within effusion samples is discussed. The potential for quantitative phase microscopy to accurately screen Pap test slides or resolve indeterminate diagnoses in urine cytology is reviewed. The article also examines the application of 3D cytology using LuCED for lung cancer detection in sputum samples and the feasibility of imaging flow and mass cytometry to measure multiple biomarkers at the single cell level. Although these novel technologies have great potential, further research is necessary to validate their routine use in cytopathology practice.

Test Characteristics of Specific p16 Clones in the Detection of High-grade Squamous Intraepithelial Lesions (HSIL).

p16 immunohistochemistry is recommended by the CAP-ASCCP Lower Anogenital Squamous Terminology (LAST) Standardization Project for human papillomavirus associated Lesions as an adjunct to morphologic assessment in the diagnosis of high-grade squamous intraepithelial lesion. This study evaluates the performance of different p16 clones as compared with E6H4 (CINtec) in detecting high-grade squamous intraepithelial lesion. The 54 high-quality articles addressing the performance of p16 identified by work group 4 of the LAST Project were evaluated for: specific p16 clone, scoring method, number of cases, anatomic site, and histologic diagnoses. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated for each clone. Two-proportion z tests (pooled) were used to evaluate significance. In total, 32 of the 54 studies met the inclusion criteria. The most commonly used clone was E6H4 (17 studies, 3507 cases) with smaller numbers (1-4) of studies evaluating the following: 16P04, JC8, 16P07, G175-405, K5334, K5336, and 7962. p16 clones 16P04 and JC8 performed better than E6H4 with 16P04 exhibiting statistically significant higher sensitivity (94% vs. 87% for E6H4), specificity (94% vs. 81%), and positive predictive value (96% vs. 69%) while JC8 exhibited higher specificity (91% vs. 81%) and positive predictive value (88% vs. 69%). 16P07 performed similarly to E6H4 and the other 4 clones did not perform as well as E6H4. p16 clones 16P04, JC8, and 16P07 clones perform as well or better than the widely used p16 clone E6H4 (CINtec). However, further studies are indicated to determine the reproducibility of these findings and the impact of interlaboratory variation on test performance.

Dr. Bibbo's Presidential Address on Automation in Cytology: Were Her Predictions Right, Wrong, or Somewhere in the Middle?

In 1983, Dr. Marluce Bibbo gave the Presidential Address at the Annual Meeting of the American Society of Cytology in Denver, CO, USA. The lecture was entitled "Analytic and Quantitative Cytology," a field in which Dr. Bibbo was intimately involved. In the presentation, she included a summary of 30 years of work already accomplished, the present state of the art, and musings about issues encountered, potential resolutions, progress that needed to be made, and her perception of how the field needed to evolve in order to become ultimately successful as a clinical service. This commentary looks back 34 years, with observations about Dr. Bibbo's predictions and how the field of cytology automation did actually evolve in the decades following her address. New challenges are identified and possible paths forward are discussed.

The Bethesda System for Reporting Cervical Cytology: A Historical Perspective.

The aims of The Bethesda System (TBS) were to provide effective communication from the laboratory to the clinical provider; facilitate cytologic-histologic correlation; facilitate research into the epidemiology, biology, and pathology of cervical disease; and provide reproducible and reliable data for national and international statistical analysis comparisons. Dr. Diane Solomon and colleagues' contribution to cervical cancer screening, diagnosis, and management began with the inception and dissemination of TBS for reporting cervical cytology in 1988, as detailed in the accompanying article [Solomon et al.: Acta Cytol 1989;33:567-574]. The significance of TBS for the further development and implementation of standardized terminology in pathology, and the research/management of cervical cancer have continued to evolve over the past three decades. TBS has always been a multidisciplinary effort and acknowledgement needs to be given to several stakeholders who, over the years, have contributed to its success. It has been our privilege and honor to have carried on the legacy of this seminal work, even as molecular methods are being closely integrated into cervical cancer screening, triage, and prevention.

Bethesda Interobserver Reproducibility Study-2 (BIRST-2): Bethesda System 2014.

INTRODUCTION: In concert with the 2014 update to the Bethesda System for Reporting Cervical Cytology, a Web-based image interobserver study was performed to evaluate concordance with the "expert panel" interpretation, as was done during the Bethesda 2001 update. The aim was to identify cytomorphologic features and Bethesda reporting categories that represent sources of poor interobserver agreement and see how the trends compared to the first Bethesda Interobserver Reproducibility Study (BIRST). MATERIALS AND METHODS: Participants were recruited online through national and international cytopathology professional societies. Study participants evaluated 84 previously unpublished web images chosen from the third Bethesda Atlas image set, prior to the release of the atlas. These images spanned all reporting categories and included typical and borderline cytomorphology. Demographic information was collected on level of training, practice patterns, and experience of the participants. Participation was restricted to those correctly answering 2 basic cytopathology questions, ensuring minimal knowledge of gynecologic cytopathology. RESULTS: A total of 1290 unique individuals attempted access to this Web-based study and 833 correctly answered the two qualifying questions. Of these, 518 respondents completed the survey. Participant origin included: 59% United States, 41% international; 48% cytotechnologists, 41% pathologists, 5% fellows, and 6% other. Practice types were: 39% academic institutions, 29% private hospitals, and 16% commercial laboratories. Overall, the mean participant agreement with the exact Bethesda panel interpretation was 62.8%. The best agreement was found for negative for intraepithelial lesion or malignancy (NILM; 74%) and low-grade squamous intraepithelial lesion (LSIL; 86%) categories. Squamous cell carcinoma (SCC) (63%), high-grade squamous intraepithelial lesion (HSIL; 60%), atypical squamous cells of undetermined significance (ASC-US; 62%) and atypical squamous cells, cannot exclude high-grade squamous intraepithelial lesion (ASC-H; 60%) showed slightly lower concordance with the panel interpretations. Cervical glandular lesions were more problematic (33%). Anal samples performed similarly to their gynecologic counterparts. There was similar diagnostic agreement across participant certifications and practice type (academic versus non-academic). Performance was higher for United States and other North America-based participants (P = 0.0104). This significance may be attributed to a language bias, as the survey was only offered in English. CONCLUSIONS: Similar to the BIRST-1 study conducted in 2001, the most important factor for diagnostic agreement by cytotechnologists, pathologists, and trainees was the a priori difficulty of an image rather than participant training, certification, or experience. Participants showed better general diagnostic agreement with the expert panel interpretations of the material in BIRST-2 than in BIRST-1. Agreement was highest for Bethesda categories of NILM, LSIL, HSIL, and SCC. Concordance for even the borderline ASC-US and ASC-H categories exhibited remarkable improvement in the BIRST-2.

Fine-Needle Aspiration of Epithelial-Myoepithelial Carcinoma of the Parotid Gland With Prominent Adenoid Cystic Carcinoma-Like Cribriform Features: Avoiding a Diagnostic Pitfall.

OBJECTIVES: Epithelial-myoepithelial carcinoma (EMC) is a rare salivary gland malignancy associated with an overall good prognosis. Fine-needle aspiration (FNA) typically shows a dual population of myoepithelial and ductal cells. Rarely, acellular matrix globules are present, raising a differential diagnosis of adenoid cystic carcinoma (AdCC), a more aggressive salivary gland malignancy associated with a poor long-term prognosis. We report an FNA case of EMC containing a predominant pattern of AdCC-like spherical globules. METHODS: We compare features of an unusual case of EMC with those of cribriform AdCC to arrive at morphologic clues to the correct diagnosis. RESULTS: Distinguishing features of EMC on FNA include (1) a prominent population of myoepithelial cells vs the predominance of basaloid cells in AdCC and (2) cohesive matrix globules with a peripheral rim of pale-staining basement membrane material compared with the dyscohesive matrix globules of AdCC. Immunochemical markers (S100, CD117, and MyB) are also useful. CONCLUSIONS: Although EMC and AdCC can both contain spherical matrix globules, close evaluation of the cytomorphology of the globules and their relationship to surrounding cells provides a clue to distinguish the two neoplasms.

Automated Extraction of Formalin-Fixed, Paraffin-Embedded Tissue for High-Risk Human Papillomavirus Testing of Head and Neck Squamous Cell Carcinomas Using the Roche Cobas 4800 System.

CONTEXT: -Testing for high-risk human papillomavirus (HR-HPV) in head and neck squamous cell carcinomas (HNSCCs) is important for both prognostication and clinical management. Several testing platforms are available for HR-HPV; however, effective alternative automated approaches are needed. OBJECTIVE: -To assess the performance of the automated Roche cobas 4800 HPV real-time polymerase chain reaction-based system on formalin-fixed, paraffin-embedded HNSCC specimens and compare results with standard methods of in situ hybridization (ISH) and p16 immunohistochemistry. DESIGN: -Formalin-fixed, paraffin-embedded samples of HNSCC were collected from archival specimens in the Department of Pathology, Massachusetts General Hospital (Boston), and prepared using the automated system by deparaffinization and dehydration followed by tissue lysis. Samples were integrated into routine cervical cytology testing runs by cobas. Corresponding formalin-fixed, paraffin-embedded samples were evaluated for HR-HPV by ISH and p16 by immunohistochemistry. Discrepant cases were adjudicated by polymerase chain reaction. RESULTS: -Sixty-two HNSCC samples were analyzed using the automated cobas system, ISH, and immunohistochemistry. Fifty-two percent (n = 32 of 62) of formalin-fixed, paraffin-embedded tumors were positive for HR-HPV by cobas. Eighty-eight percent (n = 28 of 32) of cases were the HPV 16 subtype and 12% (n = 4 of 32) were other HR-HPV subtypes. Corresponding testing with ISH was concordant in 92% (n = 57 of 62) of cases. Compared with the adjudication polymerase chain reaction standard, there were 3 false-positive cases by cobas. CONCLUSIONS: -Concordance in HNSCC HR-HPV status between cobas and ISH was more than 90%. The cobas demonstrated a sensitivity of 100% and a specificity of 91% for detection of HR-HPV. Advantages favoring cobas include its automation, cost efficiency, objective results, and ease of performance.

Practical issues related to uterine pathology: in situ and invasive cervical glandular lesions and their benign mimics: emphasis on cytology-histology correlation and interpretive pitfalls.

In situ and invasive neoplastic glandular lesions of the uterine have cytologic correlates that must be distinguished from a variety of benign and reactive conditions. Careful study of the cytologic features allows discrimination in the majority of cases; however, the designation of 'atypical glandular cells' is reserved for equivocal cases that cannot be readily resolved. In this article, the cytologic features of endocervical adenocarcinoma in situ and invasive endocervical adenocarcinoma will be presented, highlighting their correlation to the well-known histologic features. Variants of the usual type of endocervical neoplasms that have important clinical and differential diagnostic features, including mucinous adenocarcinoma and clear-cell carcinoma, will be discussed. There are a number of common cytologic mimics of endocervical neoplasms, including tubal metaplasia, directly sampled (abraded) endometrium, and high-grade squamous intraepithelial lesion involving endocervical glands. The cytologic features of these entities and their differentiation from endocervical neoplasia will be explored. Finally, the role of ancillary studies such as human papillomavirus testing in the management of glandular lesions of the cervix will be integrated into the discussion.

Trends in Cervical Cytology Screening and Reporting Practices: Results From the College of American Pathologists 2011 PAP Education Supplemental Questionnaire.

CONTEXT: The College of American Pathologists periodically surveys laboratories to determine changes in cytopathology practices. We report the results of a 2011 gynecologic cytology survey. OBJECTIVE: To provide a cross-sectional survey of gynecologic cytology practices in 2010. DESIGN: In 2011, a survey was sent to 1604 laboratories participating in the College of American Pathologists gynecologic cytology interlaboratory comparison education program and proficiency testing programs requesting data from 2010 on the following topics: terminology/reporting, cytotechnologist workload, quality assurance, reagents, and ancillary testing. RESULTS: Six hundred and twenty-five laboratories (39%) replied to the survey. The nonstandard use of "low-grade squamous intraepithelial lesion cannot exclude high-grade squamous intraepithelial lesion" is used by most laboratories to report the presence of low-grade squamous intraepithelial lesion with possibility of high-grade squamous intraepithelial lesion. Most laboratories also report the presence or absence of cells from the transformation zone. Most respondents do not limit cytotechnologist screening workload during the work shift. Only about one-third of laboratories (188 of 582; 32%) use image-assisted screening devices. Rapid prescreening as a quality assurance measure is used by only 3.5% (21 of 594) of the laboratories. When used for screening, most laboratories use the imager for retrospective review of slides to detect human locator and interpretive errors. Most laboratories receive both liquid-based cytology samples (mainly ThinPrep, Hologic, Marlborough, Massachusetts) and conventional Papanicolaou tests. Expiration dates of liquid-based cytology test vials are not usually recorded. CONCLUSIONS: The field of gynecologic cytology is evolving rapidly. These survey results offer a snapshot of national gynecologic cytology practices in 2010.

Automated 3-dimensional morphologic analysis of sputum specimens for lung cancer detection: Performance characteristics support use in lung cancer screening.

BACKGROUND: The LuCED Lung Test comprises an automated 3-dimensional morphologic analysis of epithelial cells in sputum. For each cell, 594 morphology-based features are measured to drive algorithmic classifiers that quantitatively assess whether neoplastic cells are present. The current interim clinical study involves sputum samples from patients with known benign and malignant outcomes to assess the feasibility of LuCED as an adjunctive test after suspicious low-dose computed tomography (LDCT) results or as an independent screening test for lung cancer. METHODS: Sputum samples were fixed, enriched for epithelial cells, and analyzed with a 3-dimensional cell scanner called Cell-CT. Candidate abnormal cells were identified by the classifiers for manual review. The sensitivity, specificity, and negative and positive predictive values were calculated for the detection of neoplastic cases. RESULTS: A total of 91 sputum samples from patients with confirmed lung cancer (49 patients) and patients with no known malignancy (42 patients) were evaluated. After cytology review, sensitivity in the positive group was 91.8%, and specificity was 95.2%. Specificity was not 100% because there were 2 cases in which abnormal cells were identified by the Cell-CT that were confirmed as such at the time of manual cytology review. However, at the time of last follow-up, malignancy had not been detected in these 2 cases. Modeling in a population with a 1% prevalence of lung cancer, the positive and negative predictive values would be 95.4% and 99.9%, respectively. CONCLUSIONS: LuCED testing is highly sensitive and specific for the detection of lung cancer and has potential value as an adjunctive test after suspicious LDCT findings or as a primary screening test in which LuCED-positive cases would be triaged to diagnostic CT. Further prospective studies currently are underway to evaluate its full usefulness.

Whole slide imaging for human epidermal growth factor receptor 2 immunohistochemistry interpretation: Accuracy, Precision, and reproducibility studies for digital manual and paired glass slide manual interpretation.

BACKGROUND: The use of digital whole slide imaging for human epidermal growth factor receptor 2 (HER2) immunohistochemistry (IHC) could create improvements in workflow and performance, allowing for central archiving of specimens, distributed and remote interpretation, and the potential for additional computerized automation. PROCEDURES: The accuracy, precision, and reproducibility of manual digital interpretation for HER2 IHC were determined by comparison to manual glass slide interpretation. Inter- and intra-pathologist reproducibility and precision between the glass slide and digital interpretations of HER2 IHC were determined in 5 studies using DAKO HercepTest-stained breast cancer slides with the Philips Digital Pathology System. In 2 inter-method studies, 3 pathologists interpreted glass and digital slides in sequence or in random order with a minimum of 7 days as a washout period. These studies also measured inter-observer reproducibility and precision. Another two studies measured intra-pathologist reproducibility on cases read 10 times by glass and digital methods. One additional study evaluated the effects of adding IHC control slides with each run, using 1 pathologist interpreting glass and digital slides randomized from the sets above along with appropriate controls for each slide in the set. RESULTS: The overall results show that there is no statistical difference between the variance of performance when comparing glass and digital HER2 interpretations; and there were no effects noted when control tissues were evaluated in conjunction with the test slides. CONCLUSIONS: The results show that there is an equivalence of result when interpreting HER2 IHC slides in breast cancer by either glass slides or digital images. Digital interpretation can therefore be safely and effectively used for this purpose.

The Pap Test and Bethesda 2014: "The reports of my demise have been greatly exaggerated. (after a quotation from Mark Twain)".

The Bethesda System for gynecologic cytopathology has standardized reporting terminology over the past quarter of a century. In doing so, it has allowed for improved communication among practitioners around the world, facilitated large research projects and clinical trials, and has provided the basis from which uniformly accepted risk-based management strategies have been developed. Over time, changes in terminology and the underlying science require revisions, and the third edition of the Bethesda "Atlas" is the result of a yearlong effort to provide such an update. New material was developed and proposed via an Internet bulletin board, allowing for wide commentary that was compiled and incorporated. New images were collected, which comprised many new examples of equivocal presentations and mimics. New background material on the scientific basis supporting the terminology categories, the most current management algorithms, and comprehensive references were included. The effort completely refurbishes this standard reference that forms an inexpensive and, therefore, widely available resource for the world's cytology community.