Digital validation of breast biomarkers (ER, PR, AR, and HER2) in cytology specimens using three different scanners

Artificial intelligence algorithm accurately assesses oestrogen receptor immunohistochemistry in metastatic breast cancer cytology specimens: A pilot study

OBJECTIVE

The Visiopharm artificial intelligence (AI) algorithm for oestrogen receptor (ER) immunohistochemistry (IHC) in whole slide images (WSIs) has been successfully validated in surgical pathology. This study aimed to assess its efficacy in cytology specimens.

METHODS

The study cohort comprised 105 consecutive cytology specimens with metastatic breast carcinoma. ER IHC WSIs were seamlessly integrated into the Visiopharm platform from the Image Management System (IMS) during our routine digital workflow, and an AI algorithm was employed for analysis. ER AI scores were compared with pathologists' manual consensus scores. Optimization steps were implemented and evaluated to reduce discordance.

RESULTS

The overall concordance between pathologists' scores and AI scores was excellent (99/105, 94.3%). Six cases exhibited discordant results, including two false-negative (FN) cases due to abundant histiocytes incorrectly counted as negatively stained tumour cells by AI, two FN cases owing to weak staining, and two false-positive (FP) cases where pigmented macrophages were erroneously counted as positively stained tumour cells by AI. The Pearson correlation coefficient of ER-positive percentages between pathologists' and AI scores was 0.8483. Optimization steps, such as lowering the cut-off threshold and additional training using higher input magnification, significantly improved accuracy.

CONCLUSIONS

The automated ER AI algorithm demonstrated excellent concordance with pathologists' assessments and accurately differentiated ER-positive from ER-negative metastatic breast carcinoma cytology cases. However, precision in identifying tumour cells in cytology specimens requires further enhancement.

[Ecology: The blind spot in pathology research]

INTRODUCTION

The 2015 Paris Agreement has been the first restrictive agreement in the fight against climate change. The newer generations of pathologists, who feel more anxiety due to environmental problems than their predecessors, are asked to publish research works while they are harder and harder to and in a context of demographical tensions. We wanted to measure the rise of ecology research in pathology since the Paris Agreement.

MATERIAL & METHODS

Over a ten years study period (2013-2022), we have identified via PubMed the number of articles in which forty-three terms taken from the sustainable development vocabulary appeared in ten renowned international pathology journals, selected for their SJR index from ScimagoJr and their impact factor, plus the Annales de pathologie, and compared their means of incidence between the 2013-2015 (m1) and 2016-2022 (m2) periods. The same process has been applied for "artificial intelligence", "deep learning" and "digital pathology".

RESULTS

A total of 1336 articles have been identified. Only "digital pathology" (fromm1=8,33 to m2=23,29; p=0,010) and "deep learning" (fromm1=0 to m2=10,14; p=0,034) saw their incidence rise significantly. A significant decrease has been observed with "biological" (fromm1=70,00 to m2=56,86; p=0,020).

DISCUSSION-CONCLUSIONS

Pathology reacts to trends but research in ecology has remained in the blind spot since 2015. However there seems to be an awakening as editorials, articles and communications in congress have blossomed the last two years.

Validation of cytopathology specimens for digital pathology

INTRODUCTION

Digital cytopathology is being progressively implemented in centres worldwide, but impediments such as the three-dimensionality of specimens and the size of scanned images have prevented its use from becoming widespread. This study aimed to validate the use of digital whole slide image scanning of cytopathology samples for routine sign-out.

METHODS

Specimens were scanned using the Leica Aperio GT 450 System. The following sample types were used: liquid-based cytology, direct conventional smears from fine needle aspirates and cytospins. Cases were validated by the same pathologist who originally rendered the conventional diagnosis, with a washout of at least 3 months. Final digital diagnoses were compared to the original analogical diagnoses, and cases were considered concordant up to a one-degree difference between the original and digital diagnoses. Reasons for the unsuccessful scanning of slides were also noted. The technical procedures followed the College of American Pathologists' guidelines for digital pathology validation.

RESULTS

A total of 730 slides from 383 cases (337 female, 51 male; median age 42) were successfully scanned. These cases consisted of the following sample types: 81 (21.1%) conventional smears, 240 (62.7%) liquid-based cytology samples and 62 (16.2%) cytospins. There were only five discordant cases, with a 98.7% agreement between original and digital diagnoses using the difference rate of up to one degree. Seventy-seven slides (10.5%) had to be rescanned due to technical problems. The main reasons for unsuccessful scanning were paucicellular samples (44; 57.1%), the thickness of the smears (18; 23.4%) and issues with the coverslip (15; 19.5%).

CONCLUSION

Cytological specimens can be successfully scanned and used for digital pathology, with excellent agreement with the original diagnoses.

Multilayer outperforms single-layer slide scanning in AI-based classification of whole slide images with low-burden acid-fast mycobacteria (AFB)

Manual screening of Ziehl-Neelsen (ZN)-stained slides that are negative or contain rare acid-fast mycobacteria (AFB) is labor-intensive and requires repetitive refocusing to visualize AFB candidates under the microscope. Whole slide image (WSI) scanners have enabled implementation of AI to classify digital ZN-stained slides as AFB+ or AFB-. By default, these scanners acquire a single-layer WSI. However, some scanners can acquire a multilayer WSI with a z-stack and an extended focus image layer embedded. We developed a parameterized WSI classification pipeline to assess whether multilayer imaging improves ZN-stained slide classification accuracy. A CNN built into the pipeline classified tiles in each image layer to form an AFB probability score heatmap. Features extracted from the heatmap were then entered into a WSI classifier. 46 AFB+ and 88 AFB- single-layer WSIs were used for the classifier training. 15 AFB+ (with rare microorganisms) and 5 AFB- multilayer WSIs comprised the test set. Parameters in the pipeline included: (a) a WSI representation: z-stack of image layers, middle image layer (a single image layer equivalent) or an extended focus image layer, (b) 4 methods of aggregating AFB probability scores across the z-stack, (c) 3 classifiers, (d) 3 AFB probability thresholds, and (e) 9 feature vector types extracted from the aggregated AFB probability heatmaps. Balanced accuracy (BACC) was used to measure the pipeline performance for all parameter combinations. Analysis of Covariance (ANCOVA) was used to statistically evaluate the effect of each parameter on the BACC. After adjusting for other factors, a significant effect of the WSI representation (p-value < 1.99E-76), classifier type (p-value < 1.73E-21), and AFB threshold (p-value = 0.003) was observed on the BACC. The feature type had no significant effect (p-value = 0.459) on the BACC. WSIs represented by the middle layer, extended focus layer and the z-stack followed by the weighted averaging of AFB probability scores were classified with the average BACC of 58.80%, 68.64%, and 77.28%, respectively. The multilayer WSIs represented by the z-stack with the weighted averaging of AFB probability scores were classified by a Random Forest classifier with the average BACC of 83.32%. Low classification accuracy of WSIs represented by the middle layer suggests that they contain fewer features permitting identification of AFB than the multilayer WSIs. Our results indicate that single-layer acquisition can introduce a bias (sampling error) into the WSI. This bias can be mitigated by the multilayer or the extended focus acquisitions.

Integrating and validating automated digital imaging analysis of estrogen receptor immunohistochemistry in a fully digital workflow for clinical use

Background

Design

Results

Conclusions