Time to Standardize? Time to Validate?

Meuten, Don; Munday, John S.; Hauck, Marlene

doi:10.1177/0300985817753869

Cited by 12 publications

(11 citation statements)

References 28 publications

(51 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Reproducibility is a crucial aspect of each diagnostic method, as also pointed out by Meuten 38 . A clear definition of the criteria and methods used for counting, which is essential in this regard, should thus try to reduce individual factors as much as possible.…”

Section: Discussionmentioning

confidence: 91%

See 1 more Smart Citation

Deep learning algorithms out-perform veterinary pathologists in detecting the mitotically most active tumor region

Aubreville

Bertram

Marzahl

et al. 2020

Sci Rep

View full text Add to dashboard Cite

Manual count of mitotic figures, which is determined in the tumor region with the highest mitotic activity, is a key parameter of most tumor grading schemes. It can be, however, strongly dependent on the area selection due to uneven mitotic figure distribution in the tumor section. We aimed to assess the question, how significantly the area selection could impact the mitotic count, which has a known high inter-rater disagreement. On a data set of 32 whole slide images of H&E-stained canine cutaneous mast cell tumor, fully annotated for mitotic figures, we asked eight veterinary pathologists (five board-certified, three in training) to select a field of interest for the mitotic count. To assess the potential difference on the mitotic count, we compared the mitotic count of the selected regions to the overall distribution on the slide. Additionally, we evaluated three deep learning-based methods for the assessment of highest mitotic density: In one approach, the model would directly try to predict the mitotic count for the presented image patches as a regression task. The second method aims at deriving a segmentation mask for mitotic figures, which is then used to obtain a mitotic density. Finally, we evaluated a two-stage object-detection pipeline based on state-of-the-art architectures to identify individual mitotic figures. We found that the predictions by all models were, on average, better than those of the experts. The two-stage object detector performed best and outperformed most of the human pathologists on the majority of tumor cases. The correlation between the predicted and the ground truth mitotic count was also best for this approach (0.963–0.979). Further, we found considerable differences in position selection between pathologists, which could partially explain the high variance that has been reported for the manual mitotic count. To achieve better inter-rater agreement, we propose to use a computer-based area selection for support of the pathologist in the manual mitotic count.

show abstract

Section: Discussionmentioning

confidence: 91%

“…Current grading schemes were mostly developed in retrospect 3,4,38 , with consideration of the survival rates of the patients and the manual mitotic count. Since the slides were not investigated for mitotic figures within the complete tumor area, the mitotically most active region was necessarily unknown and was possibly missed.…”

Section: Discussionmentioning

confidence: 99%

Deep learning algorithms out-perform veterinary pathologists in detecting the mitotically most active tumor region

Aubreville

Bertram

Marzahl

et al. 2020

Sci Rep

View full text Add to dashboard Cite

show abstract

“…For a prognostic marker to be clinically useful, it has to be shown to consistently predict prognosis when used by multiple different pathologists in multiple different diagnostic laboratories. In veterinary pathology, many prognostic grading schemes and other markers have been proposed, but few have been shown to consistently predict prognosis when widely used [21]. A common reason for low consistency is the inter-pathologist variability when subjectively evaluating histological criteria [22].…”

Section: Discussionmentioning

confidence: 99%

Immunostaining for p53 and p16CDKN2A Protein Is Not Predictive of Prognosis for Dogs with Malignant Mammary Gland Neoplasms

Munday

Ariyarathna

2019

Veterinary Sciences

View full text Add to dashboard Cite

Mammary gland tumors (MGTs) are common in dogs and show a variable clinical behavior that is difficult to predict. Currently, few immunohistochemical markers have been established to predict the prognosis of a canine MGT. However, p53 immunostaining has been variably reported to be prognostic for canine MGTs. Additionally, while p16CDK2NA protein (p16) immunostaining has been found to be prognostic for human breast cancers, this marker has never been evaluated as a prognostic marker for canine neoplasms. In the present study, the prognostic utility of p53 and p16 was evaluated in 35 canine malignant MGTs. It was observed that 19 (54%) dogs died due to their MGTs with an overall mean survival time (MST) of 882 days. Seven MGTs showed p53 immunostaining, but this was not significantly associated with death (4 of 7 vs. 15 of 28; p = 0.6) or MST (670 vs. 934 days; p = 0.57). Five dogs had MGTs with no p16 immunostaining, 28 MGTs had intermediate p16 immunostaining, and two MGTs had increased p16 immunostaining. Neither death due to MGT (4 of 5, 14 of 28, or 1 of 2; p = 0.28) nor MST (683, 927, and 307 days; p = 0.31) were significantly associated with p16 immunostaining. Interestingly, p53 immunostaining was significantly associated with an increase or loss of p16 immunostaining. This is the first time that p16 has been evaluated as a prognostic marker for canine neoplasms. While these results suggest that a proportion of canine MGTs develop by cellular mechanisms that alter both p53 and p16 expression, there was no evidence that defects in p53 or p16 alter the behavior of a MGT. Neither p53 nor p16 was found to significantly predict prognosis, although this could reflect the limited number of MGTs included in the study.

show abstract

“…A clear, objective, and precise definition of the outcome is expected to reduce variability of the data, improve understanding of the study methods and results, and make it easier for readers to apply the study findings to their own caseload. 29 Some of our current cancer grading schemes lack clarity or require subjective assessments or interpretations. "Severity of nuclear pleomorphism" and "degree of differentiation" are examples of criteria that cannot be consistently applied to tumor grading by different individuals.…”

Section: Definition Of the Outcomementioning

confidence: 99%

“…Even "percentage of necrosis" is of dubious consistency because the amount of necrosis depends on which parts of the tumor were examined. 29 For studies of new grading schemes, investigators should measure the interobserver variability, using both novice and experienced pathologists. Ideally, each evaluator should be given only the grading criteria used in the study and not provided with additional training, in order to test whether the criteria are sufficiently clear and can be reliably applied to daily practice.…”

Section: Definition Of the Outcomementioning

confidence: 99%

Observational Study Design in Veterinary Pathology, Part 2: Methodology

et al. 2018

View full text Add to dashboard Cite

Observational studies are a basis for much of our knowledge of veterinary pathology, yet considerations for conducting pathology-based observational studies are not readily available. In part 1 of this series, we offered advice on planning and carrying out an observational study. Part 2 of the series focuses on methodology. Our general recommendations are to consider using already-validated methods, published guidelines, data from primary sources, and quantitative analyses. We discuss 3 common methods in pathology research-histopathologic scoring, immunohistochemistry, and polymerase chain reaction-to illustrate principles of method validation. Some aspects of quality control include use of clear objective grading criteria, validation of key reagents, assessing sample quality, determining specificity and sensitivity, use of technical and biologic negative and positive controls, blinding of investigators, approaches to minimizing operator-dependent variation, measuring technical variation, and consistency in analysis of the different study groups. We close by discussing approaches to increasing the rigor of observational studies by corroborating results with complementary methods, using sufficiently large numbers of study subjects, consideration of the data in light of similar published studies, replicating the results in a second study population, and critical analysis of the study findings.

show abstract

Time to Standardize? Time to Validate?

Cited by 12 publications

References 28 publications

Deep learning algorithms out-perform veterinary pathologists in detecting the mitotically most active tumor region

Deep learning algorithms out-perform veterinary pathologists in detecting the mitotically most active tumor region

Immunostaining for p53 and p16CDKN2A Protein Is Not Predictive of Prognosis for Dogs with Malignant Mammary Gland Neoplasms

Observational Study Design in Veterinary Pathology, Part 2: Methodology

Contact Info

Product

Resources

About