Human Pathologists Research Articles

Pathologist-level classification of histopathological melanoma images with deep neural networks

BackgroundThe diagnosis of most cancers is made by a board-certified pathologist based on a tissue biopsy under the microscope. Recent research reveals a high discordance between individual pathologists. For melanoma, the literature reports 25–26% of discordance for classifying a benign nevus versus malignant melanoma. Deep learning was successfully implemented to enhance the precision of lung and breast cancer diagnoses. The aim of this study is to illustrate the potential of deep learning to assist human assessment for a histopathologic melanoma diagnosis. MethodsSix hundred ninety-five lesions were classified by an expert histopathologist in accordance with current guidelines (350 nevi and 345 melanomas). Only the haematoxylin and eosin stained (H&E) slides of these lesions were digitalised using a slide scanner and then randomly cropped. Five hundred ninety-five of the resulting images were used for the training of a convolutional neural network (CNN). The additional 100 H&E image sections were used to test the results of the CNN in comparison with the original class labels. FindingsThe total discordance with the histopathologist was 18% for melanoma (95% confidence interval [CI]: 7.4–28.6%), 20% for nevi (95% CI: 8.9–31.1%) and 19% for the full set of images (95% CI: 11.3–26.7%). InterpretationEven in the worst case, the discordance of the CNN was about the same compared with the discordance between human pathologists as reported in the literature. Despite the vastly reduced amount of data, time necessary for diagnosis and cost compared with the pathologist, our CNN archived on-par performance. Conclusively, CNNs indicate to be a valuable tool to assist human melanoma diagnoses.

In Vivo Imaging With Confirmation by Histopathology for Increased Rigor and Reproducibility in Translational Research: A Review of Examples, Options, and Resources.

Preclinical noninvasive imaging can be an indispensable tool for studying animal models of disease. In vivo imaging to assess anatomical, functional, and molecular features requires verification by a comparison to the macroscopic and microscopic morphological features, since all noninvasive in vivo imaging methods have much lower resolution than standard histopathology. Comprehensive pathological evaluation of the animal model is underutilized; yet, many institutions have veterinary or human pathologists with necessary comparative pathology expertise. By performing a rigorous comparison to gross or histopathology for image interpretation, these trained individuals can assist scientists with the development of the animal model, experimental design, and evaluation of the in vivo imaging data. These imaging and pathology corroboration studies undoubtedly increase scientific rigor and reproducibility in descriptive and hypothesis-driven research. A review of case examples including ultrasound, nuclear, optical, and MRI is provided to illustrate how a wide range of imaging modalities data can be confirmed by gross or microscopic pathology. This image confirmation and authentication will improve characterization of the model and may contribute to decreasing costs and number of animals used and to more rapid translation from preclinical animal model to the clinic.

Automatic detection of mycobacterium tuberculosis using artificial intelligence.

Tuberculosis (TB) is a global issue that seriously endangers public health. Pathology is one of the most important means for diagnosing TB in clinical practice. To confirm TB as the diagnosis, finding specially stained TB bacilli under a microscope is critical. Because of the very small size and number of bacilli, it is a time-consuming and strenuous work even for experienced pathologists, and this strenuosity often leads to low detection rate and false diagnoses. We investigated the clinical efficacy of an artificial intelligence (AI)-assisted detection method for acid-fast stained TB bacillus. We built a convolutional neural networks (CNN) model, named tuberculosis AI (TB-AI), specifically to recognize TB bacillus. The training set contains 45 samples, including 30 positive cases and 15 negative cases, where bacilli are labeled by human pathologists. Upon training the neural network model, 201 samples (108 positive cases and 93 negative cases) were collected as test set and used to examine TB-AI. We compared the diagnosis of TB-AI to the ground truth result provided by human pathologists, analyzed inconsistencies between AI and human, and adjusted the protocol accordingly. Trained TB-AI were run on the test data twice. Examined against the double confirmed diagnosis by pathologists both via microscopes and digital slides, TB-AI achieved 97.94% sensitivity and 83.65% specificity. TB-AI can be a promising support system to detect stained TB bacilli and help make clinical decisions. It holds the potential to relieve the heavy workload of pathologists and decrease chances of missed diagnosis. Samples labeled as positive by TB-AI must be confirmed by pathologists, and those labeled as negative should be reviewed to make sure that the digital slides are qualified.

Automated histological classification of whole slide images of colorectal biopsy specimens.

BackgroundAn automated image analysis system, e-Pathologist, was developed to improve the quality of colorectal biopsy diagnostics in routine pathology practice.ObjectiveThe aim of the study was to evaluate the classification accuracy of the e-Pathologist image analysis software in the setting of routine pathology practice in two institutions.Materials and methodsIn total, 1328 colorectal tissue specimens were consecutively obtained from two hospitals (1077 tissues from Tokyo hospital, and 251 tissues from East hospital) and the stained specimen slides were anonymized and digitized. At least two experienced gastrointestinal pathologists evaluated each slide for pathological diagnosis. We compared the 3-tier classification results (carcinoma or suspicion of carcinoma, adenoma, and lastly negative for a neoplastic lesion) between the human pathologists and that of e-Pathologist.ResultsFor the Tokyo hospital specimens, all carcinoma tissues were correctly classified (n=112), and 9.9% (80/810) of the adenoma tissues were incorrectly classified as negative. For the East hospital specimens, 0 out of the 51 adenoma tissues were incorrectly classified as negative while 9.3% (11/118) of the carcinoma tissues were incorrectly classified as either adenoma, or negative. For the Tokyo and East hospital datasets, the undetected rate of carcinoma, undetected rate of adenoma, and over-detected proportion were 0% and 9.3%, 9.9% and 0%, and 36.1% and 27.1%, respectively.ConclusionsThis image analysis system requires some improvements; however, it has the potential to assist pathologists in quality improvement of routine pathological practice in the not too distant future.

Automated histological classification of whole-slide images of gastric biopsy specimens.

Automated image analysis has been developed currently in the field of surgical pathology. The aim of the present study was to evaluate the classification accuracy of the e-Pathologist image analysis software. A total of 3062 gastric biopsy specimens were consecutively obtained and stained. The specimen slides were anonymized and digitized. At least two experienced gastrointestinal pathologists evaluated each slide for pathological diagnosis. We compared the three-tier (positive for carcinoma or suspicion of carcinoma; caution for adenoma or suspicion of a neoplastic lesion; or negative for a neoplastic lesion) or two-tier (negative or non-negative) classification results of human pathologists and of the e-Pathologist. Of 3062 cases, 33.4% showed an abnormal finding. For the three-tier classification, the overall concordance rate was 55.6% (1702/3062). The kappa coefficient was 0.28 (95% CI, 0.26-0.30; fair agreement). For the negative biopsy specimens, the concordance rate was 90.6% (1033/1140), but for the positive biopsy specimens, the concordance rate was less than 50%. For the two-tier classification, the sensitivity, specificity, positive predictive value, and negative predictive value were 89.5% (95% CI, 87.5-91.4%), 50.7% (95% CI, 48.5-52.9%), 47.7% (95% CI, 45.4-49.9%), and 90.6% (95% CI, 88.8-92.2%), respectively. Although there are limitations and requirements for applying automated histopathological classification of gastric biopsy specimens in the clinical setting, the results of the present study are promising.

Pathological Diagnosis of Gastric Cancers with a Novel Computerized Analysis System

Background: Recent studies of molecular biology have provided great advances for diagnostic molecular pathology. Automated diagnostic systems with computerized scanning for sampled cells in fluids or smears are now widely utilized. Automated analysis of tissue sections is, however, very difficult because they exhibit a complex mixture of overlapping malignant tumor cells, benign host-derived cells, and extracellular materials. Thus, traditional histological diagnosis is still the most powerful method for diagnosis of diseases. Methods: We have developed a novel computer-assisted pathology system for rapid, automated histological analysis of hematoxylin and eosin (H and E)-stained sections. It is a multistage recognition system patterned after methods that human pathologists use for diagnosis but harnessing machine learning and image analysis. The system first analyzes an entire H and E-stained section (tissue) at low resolution to search suspicious areas for cancer and then the selected areas are analyzed at high resolution to confirm the initial suspicion. Results: After training the pathology system with gastric tissues samples, we examined its performance using other 1905 gastric tissues. The system's accuracy in detecting malignancies was shown to be almost equal to that of conventional diagnosis by expert pathologists. Conclusions: Our novel computerized analysis system provides a support for histological diagnosis, which is useful for screening and quality control. We consider that it could be extended to be applicable to many other carcinomas after learning normal and malignant forms of various tissues. Furthermore, we expect it to contribute to the development of more objective grading systems, immunohistochemical staining systems, and fluorescent-stained image analysis systems.

Pathology of the Laboratory Mouse

The fifth in a series of European workshops for veterinary and human pathologists, "Pathology of the Laboratory Mouse: An International Workshop on Challenges for High Throughput Phenotyping," was held in Bordeaux, France, from September 30 to October 1, 2010. In this report we outline the rationale for setting up this workshop series, summarize our experience, and suggest approaches for optimizing histopathology phenotyping for gene function discovery.

Substantial Interobserver Variation Among Morphopathological Diagnosis Of Canine Mammary Gland Tumours From Veterinary And Human Pathologists

Because the morphopathological diagnosis for the benignity or malignance of tumours has played a key role of therapeutic decisions and prognosis, the reproducibility and coherence among pathologists is important. To determine the level of interobserver variation for comparative studies among veterinary and human pathologists for the diagnosis of canine mammary gland tumours (MGTs), one hundred and thirty-six selected canine MGTs with histological slides were randomly divided into two groups: 68 cases were evaluated independently by three veterinary pathologists and one human pathologist; other 68 cases were reviewed blindly by one veterinary pathologists and three human pathologists. All the participating pathologists designated whether the canine MGTs are benign or malignant. Kappa (κ) statistics were calculated to evaluate the level of agreement. For three human pathologists involved in this study, moderate to good levels of agreement among their diagnosis (κ=0.68-0.81). However, three veterinary pathologists participated in this study show less than chance to slight levels agreement (κ= -0.06 to 0.25). In this study, human pathologists tend to diagnosed canine MGTs as benign lesions while veterinary pathologists tend to diagnose as malignant lesions. We conclude that substantially high interobserver variations do exist among pathologists and will greatly influence the research result and therefore; internationally accepted pathological diagnostic definitions with high accuracy and reproducibility among pathologists are required to assist global researchers in characterizing the tumour biology, natural history, and the comparison of treatment modalities of canine MGTs.

Classification of Canine Urinary Bladder Urothelial Tumours Based on the World Health Organization/International Society of Urological Pathology Consensus Classification

One hundred canine urinary bladder urothelial (transitional cell) tumours, including roughly equal numbers of benign and malignant forms, were retrospectively categorized in accordance with the newly described human consensus classification of the World Health Organization/International Society of Urological Pathology (WHO/ISUP). The tumours were reviewed and classified by three veterinary pathologists from Michigan State University and two human pathologists from the Armed Forces Institute of Pathology (AFIP). The current human WHO/ISUP classification system was considered to be readily applicable to the dog. Canine tumours, however, differed from human tumours in that the great majority showed extensive glandular differentiation (or metaplasia) and hyperplastic lesions tended to be more florid than those seen in human beings. The various diagnoses and grades assigned to the tumours were highly consistent between all reviewing pathologists. This paper presents the salient features of the new WHO and ISUP consensus classification and provides illustrations of the various tumour types that were directly applicable to the dog.

The interactive Web-based histology atlas system.

Molecular pathways and mechanisms underlying human cancer are frequently investigated in animal models. Studies to identify and dissect molecular pathways include the discovery of genes and their subsequent analysis in transgenic and 'knock-out' mice. A critical aspect in such investigations is the evaluation of organ integrity and histology upon the alteration or inactivation of specific genes. Results from such studies are usually published in scientific journals. However, due to print space and costs the display of large high quality images is limited. Furthermore, the printed media does not permit an easy comparison of histological images published in different journals and different years. The Internet provides a tool for the timely and inexpensive dissemination of scientific data to the research community. However, its potential for the analysis of histological images has not been explored. Here we present a web-based interactive histology atlas (http://histology.nih.gov) that permits the retrieval of annotated, high-resolution histology images via the Internet. This histology atlas also takes advantage of the interactive nature of the Internet to support the communication between different research groups. As an outline forum this atlas provides the framework to evaluate and understand cancer pathology, and to develop a consensus between veterinary and human pathologists.

Growth inhibition of soil-borne pathogenic fungi by typical sterols

The application of bark-compost to farm land has been reported to suppress soil-borne diseases, including Fusarium wilts. The “bacterial type” microflora of the compost, that includes many antagonists, has been considered to play the major role in suppression of soil-borne plant pathogens. To elucidate the suppressive effect of hemlock bark-compost on soil-borne pathogenic fungi, one of the antifungal agents in bark-compost was isolated by preparative HPLC. From 2D-nuclear magnetic resonance spectra and high resolution mass spectrometry, the antifungal agent was determined to be 24R-ethylcholest-5-en-3β-ol (silosterol). Sitosterol is a normal secondary plant metabolite found throughout the plant kingdom. Furthermore, using reverse-phase HPLC, we found considerable amounts of cholesterol, campesterol and stigmasterol in acetone extracts of the bark-compost. These typical sterols had strong in vitro antifungal activity against Fusarium oxysporum f. sp. cucumerinum. We investigated the antimicrobial spectrum of cholesterol. It was active against Cochliobolus miyabeanus, Gibberella zeae, Glomerella cingulata, Alternaria alternata and Rhizoctonia solani, fungi of interest to plant and human pathologists. It was inactive against yeasts and procaryotic organisms, based on a disk-diffusion technique. We propose that there might be a parallel between this selective antimicrobial activity of the sterol and the formation of a “bacterial type” microflora of the compost. These findings suggest that the typical sterols, such as cholesterol, may function as defensive substances against pathogenic fungi at the external surface of higher eucaryotes, including the cuticle of higher plants and in kelatinized squames of the mammalian epidermis.

Liver Tumor Induction.

The significance of the development of nodular liver lesions in rodents following the administration of test agents raises several questions which could be placed in one of two general categories: diagnostic and interpretational. From a diagnostic point of view, the proper classification of liver tumors into a benign and malignant category has to be based on the direct correlation between the morphology and the biologic behavior of the lesions. Therefore, extreme care should be taken to separate the malignant tumors from the benign and the benign neoplasia from the hyperplasia. The substitution of the term "neoplastic nodule" for hyperplastic nodule in rats is misleading. Most of these nodules, when induced under special experimental conditions, may regress or remodel and thus they are not neoplastic in nature. Chronic carcinogenicity bioassays should include "stop" type of treatment leaving enough of the observational time to establish the fate of induced nodular lesions. The induction of histochemically changed foci can serve only as an indication of potential hepatocarcinogenicity and should not be equated with the induction of bona fide cancer. The biologic interpretation of nodular liver lesions, especially in mice, needs further scrutiny because these lesions have a tendency to develop spontaneously with high incidence in some strains. This characteristic then raises the question as to the mechanism by which various agents augment and/or accelerate the development of such tumors. Is this action primarily promoting or initiating in nature or does it represent the induction of tumors de novo? The answer to this dilemma may have a decisive bearing on carcinogenic risk assessment and the type of regulatory action, since the promoting agents possess a threshold effect and the promoted changes may regress following withdrawal of treatment. The interpretation of hepatocarcinogenesis is further complicated by the fact that several factors, such as sex hormonal environment, increased mitotic activity following an excessive loss of parenchymal cells, degree of caloric intake, enzymatic complement, and animals' age at the time of the exposure to a test agent, may influence the outcome of liver tumor development by modulating "initiation" and/or "promotion" of carcinogenesis. Broad fluctuation in the historic incidence of liver tumors further compounds the complexity of the proper bioassay interpretation. The specifically designed experiments may have the objective to explore predominantly the initiating or promoting effects of the agent. Such protocols should be used whenever necessary to differentiate between these two mechanisms of action. In the Caucasians, the "spontaneous" development of the primary hepatocellular tumors is rare. The majority of these tumors are malignant and rapidly fatal. According to some human pathologists, the benign variety of liver tumors is rare and it does not represent necessarily a premalignant stage in tumor development. Carcinoma of the liver may occur in infancy, especially in males before the age of 2 years. This suggests a genetic causation or carcinogenic exposure in utero. One of the geographic factors which significantly enhances the incidence of hepatocellular carcinoma in humans is exposure to aflatoxin B1 which is apparently potentiated by concurrent liver cirrhosis. Because many more agents have been found to be hepatocarcinogenic in mice and rats than in men, a question arises as to the direct relevance of rodent studies to humans. A balanced assessment of the carcinogenicity of the agent could only be reached in considering both the pharmacokinetics and the development of malignant neoplasia in other organs. In the case of positive carcinogenicity assessment, the outcome of the mutagenicity bioassays can suggest genic (genotoxic) or paragenic (epigenetic) mode of action in mammalian systems.

Radiation Malformations Belonging to the Cyclopia–Arrhinencephalia–Otocephalia Group in the Mouse Foetus

ABSTRACT Mice were subjected to a single whole-body X-radiation of 200 r. on the 9th day of pregnancy and embryos and foetuses derived from them were examined from the 13th to 19th day of pregnancy. Besides some exencephalias, hydrocephaluses, harelips and cleft-palates, a spectrum of malformations grossly classified into the following three groups was observed at comparatively high frequencies (over 32 per cent.). The first group : an apparently univentricular telencephalon accompanied by a beak-like appearance of the snout and eyes approaching the median. The second group : a reduction in the size of the telencephalon associated with a tapering of the snout and a tendency to synophthalmia. The third group : the tapering snout which may be accompanied by fusing nostrils. Besides, two cases of microtia were found. Microscopic findings suggest that these three groups present a spectrum of malformations belonging to the cyclopia-arrhinencephalia-otocephalia group. However, no typical cyclopias were detected. During the most sensitive developmental stage, the 9th day of pregnancy, 3-5 brain vesicles are formed and the anterior neuropore is closed. That this phase of basic morphogenesis of the nervous system is most sensitive to teratogenic influences is in harmony with both the results of experimental embryologists employing lower vertebrates and those of human pathologists.