Improvements in cell block processing: The Cell-Gel method

La Fortune KA, Randolph ML, Wu HH, Cramer HM. Cancer 2017; 125(4): 267–276

BACKGROUND: The ability to produce adequate cell blocks profoundly impacts the diagnostic usefulness of cytology specimens. Cell blocks are routinely processed from fine-needle aspiration specimens or concentrated fluid samples. Obtaining directed passes for the sole purpose of producing a cell block is common practice, particularly when the cytopathologist anticipates the need for ancillary immunocytochemical stains and/or molecular studies.
METHODS: The authors developed an effective and inexpensive process for producing cell blocks that consistently yields abundant cellular material, which they have termed the Cell-Gel method. This method can be simplified into 3 main steps: (1) preparing the sample; (2) constructing the cell block; and (3) processing the cell block. Highlights of the protocol include using a hemolytic fixative for sample preparation and disposable base moulds for cell block construction.
RESULTS: The cell block failure rate in the current study decreased from 18% with the HistoGel Tube method (January 2014 – December 2014) to 6% with the Cell-Gel method (January 2015 – December 2016). The authors evaluated 110 cell blocks processed with the HistoGel Tube method and 110 cell blocks processed with the Cell-Gel method, for a total evaluation of 220 cell blocks.
CONCLUSIONS: The authors have developed an effective and inexpensive protocol for producing cell blocks that consistently yields abundant cellular material. The Cell-Gel method uses a hemolytic fixative and disposable base moulds to produce adequate cell blocks. When the method was implemented, the cell block failure rate of the study laboratory decreased by approximately 67%.

Lung carcinoma predictive biomarker testing by immunoperoxidase stains in cytology and small biopsy specimens: advantages and limitations

Zhou F, Moreira AL. Arch Pathol Lab Med 2016; 140(12): 1331–1337

CONTEXT: In the burgeoning era of molecular genomics, immunoperoxidase (IPOX) testing grows increasingly relevant as an efficient and effective molecular screening tool. Patients with lung carcinoma may especially benefit from the use of IPOX because most lung carcinomas are inoperable at diagnosis and only diagnosed by small tissue biopsy or fine-needle sampling. When such small specimens are at times inadequate for molecular testing, positive IPOX results still provide actionable information.
OBJECTIVE: To describe the benefits and pitfalls of IPOX in the detection of biomarkers in lung carcinoma cytology specimens and small biopsies by summarizing the currently available commercial antibodies, pre-analytic variables, and analytic considerations.
DATA SOURCES: PubMed.
CONCLUSIONS: Commercial antibodies exist for IPOX detection of aberrant protein expression due to EGFR L858R mutation, EGFR E746_A750 deletion, ALK rearrangement, ROS1 rearrangement, and BRAF V600E mutation, as well as PD-L1 expression in tumour cells. Automated IPOX protocols for ALK and PD-L1 detection were recently approved by the Food and Drug Administration as companion diagnostics for targeted therapies, but consistent interpretive criteria remain to be elucidated, and such protocols do not yet exist for other biomarkers. The inclusion of cytology specimens in clinical trials would expand patients’ access to testing and treatment, yet there is a scarcity of clinical trial data regarding the application of IPOX to cytology, which can be attributed to trial designers’ lack of familiarity with the advantages and limitations of cytology. The content of this review may be used to inform clinical trial design and advance IPOX validation studies.

DUX4 immunohistochemistry is a highly sensitive and specific marker for CIC-DUX4 fusion-positive round cell tumor

Siegele B, Roberts J, Black JO, Rudzinski E et al. Am J Surg Pathol 2017; 41(3): 423–429

The histologic differential diagnosis of pediatric and adult round cell tumours is vast and includes the recently recognized entity CIC-DUX4 fusion-positive round cell tumour. The diagnosis of CIC-DUX4 tumour can be suggested by light microscopic and immunohistochemical features, but currently, definitive diagnosis requires ancillary genetic testing such as conventional karyotyping, fluorescence in situ hybridization, or molecular methods. We sought to determine whether DUX4 expression would serve as a fusion-specific immunohistochemical marker distinguishing CIC-DUX4 tumour from potential histologic mimics. A cohort of CIC-DUX4 fusion-positive round cell tumours harbouring t(4;19)(q35;q13) and t(10;19)(q26;q13) translocations was designed, with additional inclusion of a case with a translocation confirmed to involve the CIC gene without delineation of the partner. Round cell tumours with potentially overlapping histologic features were also collected. Staining with a monoclonal antibody raised against the C-terminus of the DUX4 protein was applied to all cases. DUX4 immunohistochemistry exhibited diffuse, crisp, strong nuclear staining in all CIC-DUX4 fusion-positive round cell tumours (5/5, 100% sensitivity), and exhibited negative staining in nuclei of all of the other tested round cell tumours, including 20 Ewing sarcomas, 1 Ewing-like sarcoma, 11 alveolar rhabdomyosarcomas, 9 embryonal rhabdomyosarcomas, 12 synovial sarcomas, 7 desmoplastic small round cell tumours, 3 malignant rhabdoid tumours, 9 neuroblastomas, and 4 clear cell sarcomas (0/76, 100% specificity). Thus, in our experience, DUX4 immunostaining distinguishes CIC-DUX4 tumours from other round cell mimics. We recommend its use when CIC-DUX4 fusion-positive round cell tumour enters the histologic differential diagnosis.


Role of quantitative p16INK4A mRNA assay and digital reading of p16INK4A immunostained sections in diagnosis of cervical intraepithelial neoplasia

Vasiljević N, Carter PD, Reuter C et al. Int J Cancer 2017; 141(4): 829–836

Visual interpretation of cervical biopsies is subjective and variable, generally showing fair to moderate inter-reader agreement in distinguishing high from low grade cervical intraepithelial neoplasia (CIN). We investigated the performance of two objective p16 quantitative tests in comparison with visual assessment: (i) p16-mRNA assay and (ii) digital analysis of sections stained for p16 protein. The primary analysis considered 232 high-risk human papilloma virus positive (HPV+) samples from diagnostic cervical specimens. A p16 RT-qPCR (p16-mRNA assay) was run on mRNA extracted from formalin-fixed paraffin-embedded sections. Two p16 immunohistochemistry (IHC) readings, a visual read by a histopathologist (Visual IHC) and a digital read of a high-resolution scan (Digital IHC), were done on adjacent sections. The worst reviewed CIN grade (agreed by at least two histopathologists) from up to two biopsies and a loop excision was taken, with CIN2/3 as the primary endpoint. Visual IHC attained a specificity of 70% (95%CI 61–77) for 85% (95%CI 77–91%) sensitivity. The four-point Visual IHC staining area under the curve (AUC) was 0.77 (95%CI 0.71–0.82), compared with 0.71 (95%CI 0.64–0.77) for p16-mRNA and 0.67 (95%CI 0.60–0.74) for Digital IHC. Spearman rank-order correlations were: visual to p16-mRNA 0.41, visual to digital 0.49 and p16-mRNA to digital: 0.22. The addition of p16-mRNA assay to visual reading of p16 IHC improved the AUC from 0.77 to 0.84 (P=0.0049). p16-mRNA testing may be complementary to visual IHC p16 staining for a more accurate diagnosis of CIN, or perhaps a substitute in locations with a lack of skilled pathologists.

Biomarkers for pathology diagnosis of uterine cervix malignant glandular lesions

Lee S, Rose MS, Sahasrabuddhe VV et al. Int J Gynecol Pathol 2017;36(4): 310–322

Immunohistochemistry is widely used to support a pathology diagnosis of cervical adenocarcinoma despite the absence of a systematic review and meta-analysis of the published data. This systematic review and meta-analysis was performed to investigate the sensitivity and specificity of immunohistochemistry biomarkers in the tissue-based diagnosis of cervical adenocarcinoma histotypes compared with normal endocervix and benign glandular lesions. The systematic review and meta-analysis used a PICOT framework and QUADAS-2 to evaluate the quality of included studies. The literature search spanned 40 years and ended June 30, 2015. Abstracts of identified records were independently screened by two of the authors who then conducted a full-text review of selected articles. Sensitivity and specificity of immunohistochemistry expression in malignant glandular lesions of the cervix classified per WHO 2003 compared with 5 benign comparators (normal/benign endocervix, and benign endocervical, endometrioid, gastric, and mesonephric lesions) were calculated. Of 902 abstracts screened, 154 articles were selected for full review. Twenty-five articles with results for 36 biomarkers were included. The only biomarker with enough studies for a meta-analysis was p16 and the definition of positive p16 staining among them was variable. Nevertheless, any positive p16 expression was sensitive, ranging from 0.94 to 0.98 with narrow confidence intervals (CIs), for adenocarcinoma in situ (AIS) and mucinous adenocarcinomas in comparison with normal/benign endocervix and benign endocervical and endometrioid lesions. Specificity for AIS and mucinous adenocarcinomas was also high with narrow CIs compared with benign endocervical lesions. The specificity was high for AIS, 0.99 (0.24, 1.0), and mucinous adenocarcinoma, 0.95 (0.52, 1.0), compared with normal/benign endocervix but with wider CIs, and low with very wide CIs compared with benign endometrioid lesions: 0.31 (0.00, 0.99) and 0.34 (0.00, 0.99), respectively. Results from single studies showed that p16, p16/Ki67 dual stain, ProExC, CEA, ESA, HIK1083, Claudin 18, and ER loss in perilesional stromal cells were useful with high (≥0.75) sensitivity and specificity estimates in ≥1 malignant versus benign comparisons. None of the biomarkers had highly useful sensitivity and specificity estimates for AIS, mucinous adenocarcinomas, or minimal deviation adenocarcinoma/gastric adenocarcinoma compared with benign gastric or mesonephric lesions or for mesonephric carcinoma compared with normal/benign endocervix, benign endocervical, endometrial, or mesonephric lesions. Any expression of p16 supports a diagnosis of AIS and mucinous adenocarcinomas in comparison with normal/benign endocervix and benign endocervical lesions. The majority of studies did not separate mosaic/focal p16 staining from diffuse staining as a distinct pattern of p16 overexpression and this may have contributed to the poor performance of p16 in distinguishing AIS and mucinous adenocarcinomas from benign endometrioid lesions. Single studies support further investigation of 8 additional biomarkers that have highly useful sensitivity and specificity estimates for ≥1 malignant glandular lesions compared with ≥1 of the 5 benign comparators.

GATA3 expression in triple-negative breast cancers

Byrne DJ, Deb S, Takano EA, Fox SB. Histopathology 2017; 71(1): 63–71

AIMS: GATA-binding protein 3 (GATA3) is a well-studied transcription factor found to be essential in the development of luminal breast epithelium and has been identified in a variety of tumour types, including breast and urothelial carcinomas, making it a useful immunohistochemistry marker in the diagnosis of both primary and metastatic disease.
METHODS AND RESULTS: We investigated GATA3 protein expression in a 106 primary triple-negative breast carcinomas (100 basal-like, six non-basal-like) using Cell Marque mouse monoclonal anti-GATA3 (L50-823). Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) was used to quantify mRNA expression in 22 triple-negative breast cancers (TNBCs) (20 primary and two cell lines), four luminal (three primary and one cell line) and five human epidermal growth factor receptor 2 (HER2) (four primary and one cell line) amplified tumours. In 98 TNBCs where IHC was assessable, 47 (48%) had a 1+ or greater staining with 20 (21%) having high GATA3 expression when using a weighted scoring.
CONCLUSION: Our study has demonstrated that GATA3 expression is common in primary triple-negative breast carcinomas. It also suggests that although GATA3 is an estrogen receptor (ER) regulated gene, it still proves useful in differentiating between primary and metastatic tumours in patients with a history of breast cancer regardless of its molecular subtype.

Accuracy of fine needle cytology in histological prediction of papillary thyroid carcinoma variants: a prospective study

Cipolletta Campanile A, Malzone MG, Losito NS et al. 2017; 28(3): 187–197

Fine needle cytology (FNC) is a crucial procedure in the preoperative diagnosis of thyroid tumours. Papillary thyroid carcinoma (PTC), in its classic variant (cPTC), is the most common malignant neoplasm of the thyroid. Several histological variants of PTC have been described, each one with its own characteristics and prognosis. The ability of FNC to identify the variants represents a challenge even for a skilled pathologist. The aim of this study was to evaluate the diagnostic cytological accuracy of FNC in PTC and to look for specific features that could predict the different variants. This was a single centre prospective study on 128 patients who received a diagnosis of PTC on FNC. The smears were blindly reviewed by two cytopathologists to create a frequency score (0, 1, 2, 3) of the features for each variant. The cytological parameters were divided into three groups: architectural, nucleo-cytoplasmic, and background features. Univariate analysis was performed by chi-square test with Yates correction and Fisher exact test as appropriate. Multiple regression analysis was performed among the variables correlated at the linear correlation. The correlation study between cytology and histology showed an accuracy of FNC in classic, follicular, and oncocytic PTC variants of 63.5, 87.5, and 87%, respectively. Familiarity with cytological features may allow an early diagnosis of a given PTC variant on FNC samples. This is fundamental in a preoperative evaluation for the best surgical approach and subsequent treatment.

A fluid, game-changing combination of mathematical tools and Big Data seems ready to disrupt the field of radiology. However, it also promises to pave the way for what may turn out to be potentially-dramatic advances in healthcare.
There is some irony here. Data was once seen as a liability, to maintain and pay for. It is now being considered a potentially major asset. The key to this turnaround in perspectives lies in increasingly sophisticated, deep learning algorithms, advanced analytics and artificial intelligence which interpret the Big Data and make it usable.

Explosion in image numbers and volume
There is no hyperbole in the use of the term Big Data, as far as radiology is concerned. In recent years, there has been a veritable explosion in the stock of medical images. Emergency room radiologists often examine up to 200 cases a day, and each patient’s imaging studies can be around 250 GB of data. At the upper end, a ‘pan scan’ CT of a trauma patient can render 4,000 images. Currently, about 450-500 petabytes of medical imaging data are generated per year, but this is accelerating. Decisions are made on the basis of small parts of imaging data, the proverbial tip of the iceberg. Much of the information in this data has still to be deciphered and used.

Medical imaging and disease
Medical imaging provides important information on anatomy and organ function as well as detecting diseases states. Its analysis covers a gamut of areas from image acquisition and compression, to  transmission, enhancement, segmentation, de-noising and reconstruction. 
Technology has enabled often-dramatic leaps in image resolution, size and availability. Sophisticated picture archiving and communications systems (PACS) have allowed for the merger of patient images from different modalities and their integration with other patient data for analysis and use in a clinical setting.

Limits to vision – from digital to analogue
So far, radiology information to identify disease or other clinical conditions is presented in the form of images. Although scanners digitize data into pixels, this is reconstructed into shapes and shades or colours for display in a form that can be understood by the human brain. 
This is where the ‘tip of the iceberg’ statement above comes into play. Medical scanners encode an image pixel in 56 bits, equivalent to 72,000 trillion ‘shades’. However, the scanner reduces the data amount to 16 bits, just 65,536 shades, for the human eye. As a result, 40 bits of information is lost, in just one pixel.
At some point in the future, it seems likely that radiologists use numbers rather than images to numerically define and detect patterns of diseases. The process may in fact have already begun.

Imaging analytics and deep learning
Such trends are being fuelled by rapid advances in imaging analytics. Smart, deep learning (DL) algorithms, which analyse pixels and other digital data bytes within an image, have the capacity to detect specific patterns associated with a pathology and provide conclusions in terms of a numerical metric.
One example of the use of numbers as a diagnostic definition concerns the use of algorithms in CT images to calculate bone density. The result is compared to a reference number, which au tomatically trigger alerts on low bone density. Avoiding the need for another dedicated examination, a physician can determine if a patient needs calcium supplements or another preventative measure.
Such algorithms also learn over time, and become better at what they do, resulting in even greater speed and more confidence in the future. Such a process has been driven by the steady acceleration, over the years, in computer processing speed. Indeed, while training an algorithm at the turn of the century took 2-3 months, the same results can now be achieved and iterated within minutes.

Neural systems and algorithms
Technically, deep learning produces a direct mapping from raw inputs to outputs such as image classes. Many DL algorithms are inspired by biologic neural systems. They are different from traditional machine learning, which requires manual feature extraction from inputs, and face limitations to use in the face of the large volumes of information associated with Big Data.
Big Data’s virtuous circle
Many DL algorithms directly seek to harness Big Data in radiology. Gigantic (and fast-growing) image libraries are being accessed for investigation to develop, test, validate and continuously refine algorithms, with the aim of covering a whole range of pathologies.
For radiologists, analytic results from an examination can be comprehensively evaluated against similar data obtained over a long period of time and evaluated to suggest appropriate diagnosis in current scenarios.

Such a virtuous cycle of algorithms and Big Data have become the focus for a host of major medical technology vendors as well as start-ups. However, the key enabling players are radiology departments, who own the data repositories and are uniquely placed to curate the data, in other words, organize it from fragments and make it available for running analytical algorithms.
The above process has, in some senses, been jump-started by previous efforts to data mine reports from radiology departments as they transitioned from PACS to enterprise imaging. The next step in this Big Data-driven opportunity will consist of linking information in radiology reports to the pixels of medical images.

The pixel goldmine
Few doubt any more that pixels are a goldmine, holding wholly new insights into a medical image and how best they could be utilized, not just by radiologists but other clinicians offering patient care. Alongside data mined from electronic medical records, quantitative pixel-based analysis algorithms are increasingly likely to be used to find patterns in images.

Big Data-based screening algorithms, for example, can be used to highlight subtle, multi-dimensional changes in a nodule or a lesion. This can be followed by applications such as curved planar or 3D multi-planar reconstructions, or dynamic contrast enhancement (DCE) texture analysis on highly targeted data subsets, instead of making the time-consuming effort of querying a complete imaging dataset. 

Specific examples of such an approach might include diagnosis of lesions in the liver and identification of disease-free liver parenchyma. Another would be volume analysis of lung tumours and solitary pulmonary nodules to decide temporal evolution of lesion. Big data based pattern analysis modules can detect areas of opacities, honeycombing, reticular densities and fibrosis, and thereby provide a list of differentials, using computer aided diagnostic tools.
For tumours, in general, radiologists can run algorithms to check contrast enhancement characteristics, and such metrics can be compared to prior results as well as other pathology data to provide a specific differential list.

Decision support systems
One decision support system based on Big Data assists physicians in providing treatment planning for patients suffering from traumatic brain injury (TBI). The algorithm couples demographic data and medical records of the patient to specific features extracted from CT scans in order to predict intracranial pressure (ICP) levels.
Google’s entry into this field seeks to address real world limitations – not just in terms of human capacities but also trained medical personnel. Its first deep learning imaging algorithm sought to recognize diabetic retinopathy, the fastest growing cause of blindness in poor countries, where a shortage of specialists meant many patients lost their sight before diagnosis.

The promise of AI
Google’s algorithm is based on artificial intelligence (AI), seen as an especially promising catalyst for advances in such areas.
AI-based algorithms, for example,  can calculate the volume of bleed on the basis of multiple brain CT slices in stroke patients, with the size of bleed volume indicating urgency as well as care pathway. Another recent algorithm assesses recent infarcts on CT, which can be missed if they are hyper-acute (less than 8-12 hours old), and is therefore relevant to all patients with sudden onset weakness. The University of California in San Francisco has been testing an algorithm to identify pneumothorax in chest radiographs of surgery patients, before they exit the OR (operating room).  The aim is to not only avoid the huge costs of a collapsed lung but also ensure that the OR is freed from being used for an otherwise-avoidable procedure.
AI is also being considered for workflow management and triaging. In the near future, it is almost certain that images are screened as data is acquired by a scanner, to distinguish between ‘normal’ and ‘abnormal’ images, prioritize cases according to the likelihood of disease and alerting radiologists to conditions that require urgent attention. The results are tangible and impressive. One algorithm has helped physicians to shrink the time for cardiac diagnoses from 30 minutes to 15 seconds.

Certain vendors are leveraging AI to correlate findings on properties like morphology, cell density or physiological characteristics to expert radiologist’s reports, while taking additional clinical data such as biopsy results into account. Others use reasoning protocols as well as visual technologies such as virtual rendering to analyse medical images. This is then combined with data from a patient’s medical record to offer radiologists and clinicians decision-making support.

AI and the radiologist
So far, algorithms and emerging metrics are expected to be largely used as a complement to decisions made by radiologists.
However, at some point in the future, it seems plausible that radiologists no longer need to look at images at all. Instead, they would simply analyse outcomes of the algorithms.
Once again, AI is at play here. Apart from deep learning algorithms, radiology can claim to be witness to the first successes with the emerging science of ‘swarm’ AI, which helps form a diagnostic consensus by turning groups of human experts into super experts.  Swarm AI is directly based on nature, which sees species accomplishing more by participating in a flock, school or colony (a ‘swarm’) than they can individually. One report, published in ‘Public Library of Science (PLOS)’, stated that swarm intelligence could improve other types of medical decision-making, ”including many areas of diagnostic imaging.”
In December 2015, a study in ‘IET Systems Biology’ reported about a swarm intelligence algorithm which assisted “in the identification of metastasis in bone scans and micro-calcifications on mammographs.” The authors, from universities in the UK and India, also reported about the use of the algorithm in assessing CT images of the aorta and in chest X-ray. They proposed a hybrid swarm intelligence approach to detect tumour regions in an abnormal MR brain image.

The future: human-machine symbiosis
AI is unlikely to become a replacement for radiologists, but a tool to help them. According to Curt Langlotz, MD, PhD, professor of radiology and biomedical informatics at Stanford, the “human-machine system always performs better than either alone.”

Interventional radiology is a subspecialty providing minimally invasive image-guided diagnosis and treatment of disease. The number of procedures performed by interventional radiologists is extensive, ranging from the purely diagnostic such as angiography and cholangiography to the therapeutic, covering vascular and ablative applications. In recent years there has been a shift away from diagnostic angiography with the arrival on the market of high performance CT and MRI angiography systems which provide reliable and non-invasive alternatives.
Radiologists are by no means the only medical specialty performing interventional techniques as cardiologists and vascular surgeons have been quite successful in developing interventional skills, so much so that interventional cardiology has grown into a discipline of its own. In fields such as peripheral arterial disease treatment for example, it would seem that interventional radiologists have lost out to other specialties even though in some European countries like Germany they still have a significant share of this work. There is, however, a wide range of other areas, especially in interventional neuroradiology and oncology where interventional radiologists hold a quasi-monopoly.
In Europe, the Cardiovascular and Interventional Radiological Society of Europe (CIRSE) established in 2013 the first edition of the European Curriculum and Syllabus for Interventional Radiology which aimed at harmonizing training across European countries while supporting the European Board of Interventional Radiology (EBIR) examination in interventional radiology. CIRSE also works together with the European Society of Radiology (ESR) to attract more clinically-oriented medical students to interventional radiology. The steady growth of the ageing population in Europe and other industrialized countries and the resulting higher incidence of strokes and cancer cases combined with the multiplication and development of interventional techniques will boost the demand for interventional radiologists in the near future. Already now, there is a shortage of interventional radiologists in some countries, notably the UK where 25% of hospitals cannot provide minimally invasive procedures to their patients on a 24/7 basis because of a lack of recruitment of interventional radiologists in the National Health Service (NHS). This situation is having a clearly detrimental effect on patient care in some parts of the country. According to the Royal College of Radiologists (RCR), there are just 414 interventional radiologists in the NHS whereas 735 of them would be needed to provide 24/7 on-call service everywhere. In comparison, France has about 1,250 interventional radiologists while Germany totals over 1,000. In emerging countries, the shortages can be huge, such as in India where there are only 596 registered interventional radiologists, i.e. one per every 2,18 million population. The challenge for radiologists is to recognize the value of being close to the patient and embrace clinical care.