How does a pathologist make a diagnosis?

Seeing the random forest through the decision trees. Supporting learning health systems from histopathology with machine learning models: Challenges and opportunities

This paper discusses some overlooked challenges faced when working with machine learning models for histopathology and presents a novel opportunity to support "Learning Health Systems" with them. Initially, the authors elaborate on these challenges after separating them according to their mitigation strategies: those that need innovative approaches, time, or future technological capabilities and those that require a conceptual reappraisal from a critical perspective. Then, a novel opportunity to support "Learning Health Systems" by integrating hidden information extracted by ML models from digitalized histopathology slides with other healthcare big data is presented.

Whole slide image representation in bone marrow cytology

One of the goals of AI-based computational pathology is to generate compact representations of whole slide images (WSIs) that capture the essential information needed for diagnosis. While such approaches have been applied to histopathology, few applications have been reported in cytology. Bone marrow aspirate cytology is the basis for key clinical decisions in hematology. However, visual inspection of aspirate specimens is a tedious and complex process subject to variation in interpretation, and hematopathology expertise is scarce. The ability to generate a compact representation of an aspirate specimen may form the basis for clinical decision-support tools in hematology. In this study, we leverage our previously published end-to-end AI-based system for counting and classifying cells from bone marrow aspirate WSIs, which enables the direct use of individual cells as inputs rather than WSI patches. We then construct bags of individual cell features from each WSI, and apply multiple instance learning to extract their vector representations. To evaluate the quality of our representations, we conducted WSI retrieval and classification tasks. Our results show that we achieved a mAP@10 of 0.58 ±0.02 in WSI-level image retrieval, surpassing the random-retrieval baseline of 0.39 ±0.1. Furthermore, we predicted five diagnostic labels for individual aspirate WSIs with a weighted-average F1 score of 0.57 ±0.03 using a k-nearest-neighbors (k-NN) model, outperforming guessing using empirical class prior probabilities (0.26 ±0.02). We present the first example of exploring trainable mechanisms to generate compact, slide-level representations in bone marrow cytology with deep learning. This method has the potential to summarize complex semantic information in WSIs toward improved diagnostics in hematology, and may eventually support AI-assisted computational pathology approaches.

Applied machine learning in hematopathology

An increasing number of machine learning applications are being developed and applied to digital pathology, including hematopathology. The goal of these modern computerized tools is often to support diagnostic workflows by extracting and summarizing information from multiple data sources, including digital images of human tissue. Hematopathology is inherently multimodal and can serve as an ideal case study for machine learning applications. However, hematopathology also poses unique challenges compared to other pathology subspecialities when applying machine learning approaches. By modeling the pathologist workflow and thinking process, machine learning algorithms may be designed to address practical and tangible problems in hematopathology. In this article, we discuss the current trends in machine learning in hematopathology. We review currently available machine learning enabled medical devices supporting hematopathology workflows. We then explore current machine learning research trends of the field with a focus on bone marrow cytology and histopathology, and how adoption of new machine learning tools may be enabled through the transition to digital pathology.

Deep Learning on Histopathological Images for Colorectal Cancer Diagnosis: A Systematic Review

Colorectal cancer (CRC) is the second most common cancer in women and the third most common in men, with an increasing incidence. Pathology diagnosis complemented with prognostic and predictive biomarker information is the first step for personalized treatment. The increased diagnostic load in the pathology laboratory, combined with the reported intra- and inter-variability in the assessment of biomarkers, has prompted the quest for reliable machine-based methods to be incorporated into the routine practice. Recently, Artificial Intelligence (AI) has made significant progress in the medical field, showing potential for clinical applications. Herein, we aim to systematically review the current research on AI in CRC image analysis. In histopathology, algorithms based on Deep Learning (DL) have the potential to assist in diagnosis, predict clinically relevant molecular phenotypes and microsatellite instability, identify histological features related to prognosis and correlated to metastasis, and assess the specific components of the tumor microenvironment.

Quantum Mechanics and Surgical Pathology: A Brief Introduction

Quantum mechanics (QM) and surgical pathology might seem totally unrelated fields of science. Because QM or particle physics explains the very basic structure and function of nature, there are growing interconnections between the fundamentals and applications of QM and biologic sciences. QM is not only applied to the structure of atoms but also probes the structure of biologic molecules, explains their mutational changes and has provided an insight into the basic mechanisms of many different biologic systems. Many of the current applications in biologic sciences, medicine, and surgical pathology rely on the principles of QM. Because surgical pathology uses quantum phenomena such as light and studies disease's alterations that are ultimately governed by quantum changes at nanoscale levels, QM will have potential future implications for the progress of surgical pathology. These might include quantum-enhanced refinements in light, ancillary tools, and interpretation assistance computerized systems. The future of applying the concepts, discoveries, and tools of QM in surgical pathology might create something analogous to quantum biology; that is, quantum pathology or "QuPath."

The "Make Surgical Pathology Easy" project: learning Pathology through tailored digital infographics - the case for renovation of an old teaching method

The term ‘infographics’ is a blend of the two words “information” and “graphics”. Infographics can be described as ‘information visualizations’, conceived as visual translation of data including text, numbers, graphs, charts, drawings and so on. Visual representations are a fundamental part of scientific communication. They match the need to organize different pieces of information in a coherent and synthetic structure and constitute one of the most effective methods scientists rely on to divulge their findings. In particular, infographics provide an overview of key points regarding specific topics in a form that promotes quick learning and knowledge retention. They can be presented in printed or digital formats, being the latter particularly suitable for a global-scale diffusion via social media or websites.

In recent years, many pathologists have started developing digital infographics as a strategy for providing free educational contents on Facebook, Twitter or websites. In the present review, we focus on the value of digital infographics to summarize various aspects of Surgical and General Pathology. They shed light on diagnostic criteria, differentials and predictive/prognostic markers for many diseases, being a useful learning tool both for residents and practicing pathologists. In this paper, the model of infographics ideation, processing and sharing to an online audience is described and the impact of infographics on knowledge processes in Pathology is investigated.

It's elemental! Siliceous diatom frustules producing sarcoid-like granulomas in the subcutis

Silica granulomas have been described on the skin and are rare; however, this is the first report of a sarcoid-like granulomatous reaction because of siliceous diatom frustules in the subcutis, making this an unprecedented case. A 41-year-old female presented with a subcutaneous nodule on the right forearm clinically suggestive of cyst, foreign body, or lipoma. Excisional biopsy revealed subcutis with a sarcoid-like granulomatous reaction in a background of fibrosis, containing abundant semitransparent, exquisitely geometric particles, mildly refringent under polarized light, highlighted by phase contrast microscopy; special stains were negative for microorganisms. Definitive characterization of the peculiar fragments was accomplished by confocal laser microscopy, scanning electron microscopy, and energy dispersive X-ray spectroscopy, revealing them as diatom frustules made of silicon dioxide (SiO₂ ) or silica. Diatoms are unicellular algae, their skeletons (frustules) made of silica have collected on the bottom of rivers, lakes, and oceans for thousands or millions of years and form what we know as diatomite or diatomaceous earth, which is widely used in different industries and easily available in the market. The mechanism whereby diatom frustules gained access to the patient's subcutis is enigmatic.

The leading role of pathology in assessing the somatic molecular alterations of cancer: Position Paper of the European Society of Pathology

Molecular pathology is an essential part of pathology complementing conventional morphological tools to obtain a correct integrated diagnosis with appropriate assessment of prognosis and prediction of response to therapy, particularly in cancer. There is a concern about the situation of molecular pathology in some areas of Europe, namely, regarding the central role of pathologists in assessing somatic genomic alterations in cancer. In some countries, there are attempts that other laboratory medicine specialists perform the molecular analysis of somatic alterations in cancer, particularly now when next generation sequencing (NGS) is incorporated into clinical practice. In this scenario, pathologists may play just the role of “tissue providers,” and other specialists may take the lead in molecular analysis. Geneticists and laboratory medicine specialists have all background and skills to perform genetic analysis of germline alterations in hereditary disorders, including familial forms of cancers. However, interpretation of somatic alterations of cancer belongs to the specific scientific domain of pathology. Pathologists are necessary to guarantee the quality of the results, for several reasons: (1) The identified molecular alterations should be interpreted in the appropriate morphologic context, since most of them are context-specific; (2) pre-analytical issues must be taken into consideration; (3) it is crucial to check the proportion of tumor cells in the sample subjected to analysis and presence of inflammatory infiltrate and necrosis should be monitored; and 4) the role of pathologists is crucial to select the most appropriate methods and to control the turnaround time in which the molecular results are delivered in the context of an integrated diagnosis. Obviously, there is the possibility of having core facilities for NGS in a hospital to perform the sequence analysis that are open to other specialties (microbiologists, geneticists), but also in this scenario, pathologists should have the lead in assessing somatic alterations of cancer. In this article, we emphasize the importance of interpreting somatic molecular alterations of the tumors in the context of morphology. In this Position Paper of the European Society of Pathology, we strongly support a central role of pathology departments in the process of analysis and interpretation of somatic molecular alterations in cancer.

Machine learning applications in the diagnosis of leukemia: Current trends and future directions

Machine learning (ML) offers opportunities to advance pathological diagnosis, especially with increasing trends in digitalizing microscopic images. Diagnosing leukemia is time-consuming and challenging in many areas globally and there is a growing trend in utilizing ML techniques for its diagnosis. In this review, we aimed to describe the literature of ML utilization in the diagnosis of the four common types of leukemia: acute lymphocytic leukemia (ALL), chronic lymphocytic leukemia (CLL), acute myeloid leukemia (AML), and chronic myelogenous leukemia (CML). Using a strict selection criterion, utilizing MeSH terminology and Boolean logic, an electronic search of MEDLINE and IEEE Xplore Digital Library was performed. The electronic search was complemented by handsearching of references of related studies and the top results of Google Scholar. The full texts of 58 articles were reviewed, out of which, 22 studies were included. The number of studies discussing ALL, AML, CLL, and CML was 12, 8, 3, and 1, respectively. No studies were prospectively applying algorithms in real-world scenarios. Majority of studies had small and homogenous samples and used supervised learning for classification tasks. 91% of the studies were performed after 2010, and 74% of the included studies applied ML algorithms to microscopic diagnosis of leukemia. The included studies illustrated the need to develop the field of ML research, including the transformation from solely designing algorithms to practically applying them clinically.

Artificial Intelligence and Digital Pathology: Challenges and Opportunities

In light of the recent success of artificial intelligence (AI) in computer vision applications, many researchers and physicians expect that AI would be able to assist in many tasks in digital pathology. Although opportunities are both manifest and tangible, there are clearly many challenges that need to be overcome in order to exploit the AI potentials in computational pathology. In this paper, we strive to provide a realistic account of all challenges and opportunities of adopting AI algorithms in digital pathology from both engineering and pathology perspectives.

The Effect of Mental Rotation on Surgical Pathological Diagnosis

PURPOSE

Pathological diagnosis involves very delicate and complex consequent processing that is conducted by a pathologist. The recognition of false patterns might be an important cause of misdiagnosis in the field of surgical pathology. In this study, we evaluated the influence of visual and cognitive bias in surgical pathologic diagnosis, focusing on the influence of "mental rotation."

MATERIALS AND METHODS

We designed three sets of the same images of uterine cervix biopsied specimens (original, left to right mirror images, and 180-degree rotated images), and recruited 32 pathologists to diagnose the 3 set items individually.

RESULTS

First, the items found to be adequate for analysis by classical test theory, Generalizability theory, and item response theory. The results showed statistically no differences in difficulty, discrimination indices, and response duration time between the image sets.

CONCLUSION

Mental rotation did not influence the pathologists' diagnosis in practice. Interestingly, outliers were more frequent in rotated image sets, suggesting that the mental rotation process may influence the pathological diagnoses of a few individual pathologists.

Learning-Based Visual Saliency Model for Detecting Diabetic Macular Edema in Retinal Image

This paper brings forth a learning-based visual saliency model method for detecting diagnostic diabetic macular edema (DME) regions of interest (RoIs) in retinal image. The method introduces the cognitive process of visual selection of relevant regions that arises during an ophthalmologist's image examination. To record the process, we collected eye-tracking data of 10 ophthalmologists on 100 images and used this database as training and testing examples. Based on analysis, two properties (Feature Property and Position Property) can be derived and combined by a simple intersection operation to obtain a saliency map. The Feature Property is implemented by support vector machine (SVM) technique using the diagnosis as supervisor; Position Property is implemented by statistical analysis of training samples. This technique is able to learn the preferences of ophthalmologist visual behavior while simultaneously considering feature uniqueness. The method was evaluated using three popular saliency model evaluation scores (AUC, EMD, and SS) and three quality measurements (classical sensitivity, specificity, and Youden's J statistic). The proposed method outperforms 8 state-of-the-art saliency models and 3 salient region detection approaches devised for natural images. Furthermore, our model successfully detects the DME RoIs in retinal image without sophisticated image processing such as region segmentation.

What should not be reported as atypia in urine cytology

The term "atypia," although not well characterized, is widely used in diagnostic surgical and cytopathology. Because there are no guidelines regarding when to use this term, in the majority of cases, it is used as a "wastebasket." This definitely applies to urine cytology, where the reported rate of atypia ranges from 1.9% to 23%. This review lists a number of cytomorphologic findings in urine cytology that are associated with known and specific causes. Urine specimens in which the morphologic changes can be attributed to particular etiologic factors should no longer be classified as "atypical." These include urine specimens showing reactive umbrella cells or seminal vesicle cells, reactive changes due to stones, cytologic changes characteristic of infectious processes or therapy effect, instrumented urines with pseudopapillary clusters, and urinary diversion specimens.

A multicenter study directly comparing the diagnostic accuracy of gene expression profiling and immunohistochemistry for primary site identification in metastatic tumors

Metastatic tumors with an uncertain primary site can be a difficult clinical problem. In tens of thousands of patients every year, no confident diagnosis is ever issued, making standard-of-care treatment impossible. Gene expression profiling (GEP) tests currently available to analyze these difficult-to-diagnose tumors have never been directly compared with the diagnostic standard of care, immunochemistry (IHC). This prospectively conducted, blinded, multicenter study compares the diagnostic accuracy of GEP with IHC in identifying the primary site of 157 formalin-fixed paraffin-embedded specimens from metastatic tumors with known primaries, representing the 15 tissues on the GEP test panel. Four pathologists rendered diagnoses by selecting from 84 stains in 2 rounds. GEP was performed using the Pathwork Tissue of Origin Test. Overall, GEP accurately identified 89% of specimens, compared with 83% accuracy using IHC (P=0.013). In the subset of 33 poorly differentiated and undifferentiated carcinomas, GEP accuracy exceeded that of IHC (91% to 71%, P=0.023). In specimens for which pathologists rendered their final diagnosis with a single round of stains, both IHC and GEP exceeded 90% accuracy. However, when the diagnosis required a second round, IHC significantly underperformed GEP (67% to 83%, P<0.001). GEP has been validated as accurate in diagnosing the primary site in metastatic tumors. The Pathwork Tissue of Origin Test used in this study was significantly more accurate than IHC when used to identify the primary site, with the most pronounced superiority observed in specimens that required a second round of stains and in poorly differentiated and undifferentiated metastatic carcinomas.

Standardized categorical reporting of cytopathology results: the strengths and weaknesses of a constantly evolving and expanding system

Since the success of the Bethesda nomenclature system in standardizing Pap smear results, there has been growing interest in adopting Bethesda-like standardized categorical formats in areas of nongynecologic cytopathology. Standardized categorical reporting may have several advantages over descriptive reporting, in enhancing cytopathologist-clinician communication and inter-institutional exchange of information, providing better guidance for treatment planning, and facilitating statistical analysis for research purposes or quality control studies. On the other hand, descriptive reporting may be more effective as a tool of communication between cytopathologists, may better express the uncertainty of the observer in diagnostically difficult and equivocal cases and may better serve the purposes of training and continuing education of cytopathologists. Future studies on the pros and cons of the different reporting systems used in cytopathology may provide further insight on these issues. The most problematic areas need to be identified and optimal solutions decided. Despite the ongoing debate on the optimal reporting format in cytopathology, there is general agreement on the need for high quality cytology reports (whether descriptive or standardized) in terms of their diagnostic accuracy, clarity and clinical value.

Comparison of histopathology to gene expression profiling for the diagnosis of metastatic cancer

Background

Determining the primary site of metastatic cancer with confidence can be challenging. Pathologists commonly use a battery of immunohistochemical (IHC) stains to determine the primary site. Gene expression profiling (GEP) has found increasing use, particularly in the most difficult cases. In this pilot study, a direct comparison between GEP and IHC-guided methods was performed.

Methods

Ten archived formalin-fixed paraffin embedded metastatic tumor samples for which the primary site had been clinically determined were selected. Five pathologists who were blinded to the diagnosis were asked to determine the primary site using IHC and other stains selected from a panel of 84 stains. Each pathologist was provided patient sex, biopsy site and gross sample description only. Slides were digitized using ScanScope®XT at 0.25 μm/pixel. Each evaluating pathologist was allowed to provide a diagnosis in three stages: initial (after reviewing the H&E image), intermediate (after reviewing images from the first batch of stains) and final diagnosis (after the second batch of stains if requested). GEP was performed using the only FDA-cleared test for this intended use, the Pathwork Tissue of Origin Test. No sample information was provided for GEP testing except for patient sex. Results were reported as the tumor tissue type with the highest similarity score.

Results

In this feasibility study, GEP determined the correct primary site in 9 of the 10 cases (90%), compared to the IHC-guided method which determined the correct primary site for 32 of 50 case evaluations (average 64%, range 50% to 80%). The five pathologists directing the IHC-guided method ordered an average of 8.8 stains per case (range 1 to 18). GEP required an average of 3 slides per case (range 1 to 4).

Conclusions

Results of the pilot study suggest that GEP provides correct primary site identification in a higher percentage of metastatic cases than IHC-guided methods, and uses less tissue. A larger comparative effectiveness study using this study design is needed to confirm the results.

Virtual slides

The virtual slide(s) for this article can be found here: http://www.diagnosticpathology.diagnomx.eu/vs/1749854104745508

A pathologist-in-the-loop IHC antibody test selection using the entropy-based probabilistic method

Background:

Immunohistochemistry (IHC) is an important tool to identify and quantify expression of certain proteins (antigens) to gain insights into the molecular processes in a diseased tissue. However, it is a challenge for pathologists to remember the discriminative characteristics of the growing number of such antigens across multiple diseases. The complexity of their expression patterns, fueled by continuous discoveries in molecular pathology, gives rise to a combinatorial explosion that places an unprecedented burden on a practicing pathologist and therefore increases cost and variability of IHC studies.

Materials and Methods:

To tackle these issues, we have developed antibody test optimized selection method, a novel informatics tool to help pathologists in improving the IHC antibody selection process. The method uses extensions of Shannon's information entropies and Bayesian probabilities to dynamically build an efficient diagnostic tree.

Results:

A comparative analysis of our method with the expert and World Health Organization classification guidelines showed that the proposed method brings threefold reduction in number of antibody tests required to reach a diagnostic conclusion.

Conclusion:

The developed method can significantly streamline the antibody test selection process, decrease associated costs and reduce inter- and intrapathologist variability in IHC decision-making.

Learning regions of interest from low level maps in virtual microscopy

Virtual microscopy can improve the workflow of modern pathology laboratories, a goal limited by the large size of the virtual slides (VS). Lately, determination of the Regions of Interest has shown to be useful in navigation and compression tasks. This work presents a novel method for establishing RoIs in VS, based on a relevance score calculated from example images selected by pathologist. The process starts by splitting the Virtual Slide (VS) into a grid of blocks, each represented by a set of low level features which aim to capture the very basic visual properties, namely, color, intensity, orientation and texture. The expert selects then two blocks i.e. A typical relevant (irrelevant) instance. Different similarity (disimilarity) maps are then constructed, using these positive (negative) examples. The obtained maps are then integrated by a normalization process that promotes maps with a similarity global maxima that largely exceeds the average local maxima. Each image region is thus entailed with an associated score, established by the number of closest positive (negative) blocks, whereby any block has also an associated score. Evaluation was carried out using 8 VS from different tissues, upon which a group of three pathologists had navigated. Precision-recall measurements were calculated at each step of any actual navigation, obtaining an average precision of 55% and a recall of about 38% when using the available set of navigations.

A supervised visual model for finding regions of interest in basal cell carcinoma images

This paper introduces a supervised learning method for finding diagnostic regions of interest in histopathological images. The method is based on the cognitive process of visual selection of relevant regions that arises during a pathologist's image examination. The proposed strategy emulates the interaction of the visual cortex areas V1, V2 and V4, being the V1 cortex responsible for assigning local levels of relevance to visual inputs while the V2 cortex gathers together these small regions according to some weights modulated by the V4 cortex, which stores some learned rules. This novel strategy can be considered as a complex mix of "bottom-up" and "top-down" mechanisms, integrated by calculating a unique index inside each region. The method was evaluated on a set of 338 images in which an expert pathologist had drawn the Regions of Interest. The proposed method outperforms two state-of-the-art methods devised to determine Regions of Interest (RoIs) in natural images. The quality gain with respect to an adaptated Itti's model which found RoIs was 3.6 dB in average, while with respect to the Achanta's proposal was 4.9 dB.

An experimental study of pathologist's navigation patterns in virtual microscopy

In virtual microscopy, a sequential process of captures of microscopical fields, allows to construct a virtual slide which is visualized using a specialized software, called the virtual microscopy viewer. This tool allows useful exploration of images, composed of thousands of microscopical fields of view at different levels of magnification, emulating an actual microscopical examination. The aim of this study was to establish the main pathologist's navigation patterns when exploring virtual microscopy slides, using a graphical user interface, adapted to the pathologist's workflow. Four pathologists with a similar level of experience, graduated from the same pathology program, navigated six virtual slides. Different issues were evaluated, namely, the percentage of common visited image regions, the time spent at each and its coincidence level, that is to say, the region of interest location. In addition, navigation patterns were also assessed, i.e., mouse movement velocities and linearity of the diagnostic paths. Results suggest that regions of interest are determined by a complex combination of the visited area, the time spent at each visit and the coincidence level among pathologists. Additionally, linear trajectories and particular velocity patterns were found for the registered diagnostic paths.