Artificial intelligence-based tools applied to pathological diagnosis of microbiological diseases

VENet: Variational energy network for gland segmentation of pathological images and early gastric cancer diagnosis of whole slide images

BACKGROUND AND OBJECTIVE

Gland segmentation of pathological images is an essential but challenging step for adenocarcinoma diagnosis. Although deep learning methods have recently made tremendous progress in gland segmentation, they have not given satisfactory boundary and region segmentation results of adjacent glands. These glands usually have a large difference in glandular appearance, and the statistical distribution between the training and test sets in deep learning is inconsistent. These problems make networks not generalize well in the test dataset, bringing difficulties to gland segmentation and early cancer diagnosis.

METHODS

To address these problems, we propose a Variational Energy Network named VENet with a traditional variational energy Lv loss for gland segmentation of pathological images and early gastric cancer detection in whole slide images (WSIs). It effectively integrates the variational mathematical model and the data-adaptability of deep learning methods to balance boundary and region segmentation. Furthermore, it can effectively segment and classify glands in large-size WSIs with reliable nucleus width and nucleus-to-cytoplasm ratio features.

RESULTS

The VENet was evaluated on the 2015 MICCAI Gland Segmentation challenge (GlaS) dataset, the Colorectal Adenocarcinoma Glands (CRAG) dataset, and the self-collected Nanfang Hospital dataset. Compared with state-of-the-art methods, our method achieved excellent performance for GlaS Test A (object dice 0.9562, object F1 0.9271, object Hausdorff distance 73.13), GlaS Test B (object dice 94.95, object F1 95.60, object Hausdorff distance 59.63), and CRAG (object dice 95.08, object F1 92.94, object Hausdorff distance 28.01). For the Nanfang Hospital dataset, our method achieved a kappa of 0.78, an accuracy of 0.9, a sensitivity of 0.98, and a specificity of 0.80 on the classification task of test 69 WSIs.

CONCLUSIONS

The experimental results show that the proposed model accurately predicts boundaries and outperforms state-of-the-art methods. It can be applied to the early diagnosis of gastric cancer by detecting regions of high-grade gastric intraepithelial neoplasia in WSI, which can assist pathologists in analyzing large WSI and making accurate diagnostic decisions.

The Laboratory Diagnosis of Malaria: A Focus on the Diagnostic Assays in Non-Endemic Areas

Even if malaria is rare in Europe, it is a medical emergency and programs for its control should ensure both an early diagnosis and a prompt treatment within 24-48 h from the onset of the symptoms. The increasing number of imported malaria cases as well as the risk of the reintroduction of autochthonous cases encouraged laboratories in non-endemic countries to adopt diagnostic methods/algorithms. Microscopy remains the gold standard, but with limitations. Rapid diagnostic tests have greatly expanded the ability to diagnose malaria for rapid results due to simplicity and low cost, but they lack sensitivity and specificity. PCR-based assays provide more relevant information but need well-trained technicians. As reported in the World Health Organization Global Technical Strategy for Malaria 2016-2030, the development of point-of-care testing is important for the improvement of diagnosis with beneficial consequences for prompt/accurate treatment and for preventing the spread of the disease. Despite their limitations, diagnostic methods contribute to the decline of malaria mortality. Recently, evidence suggested that artificial intelligence could be utilized for assisting pathologists in malaria diagnosis.

Automated Identification of Cutaneous Leishmaniasis Lesions Using Deep-Learning-Based Artificial Intelligence

The polymorphism of cutaneous leishmaniasis (CL) complicates diagnosis in health care services because lesions may be confused with other dermatoses such as sporotrichosis, paracocidiocomycosis, and venous insufficiency. Automated identification of skin diseases based on deep learning (DL) has been applied to assist diagnosis. In this study, we evaluated the performance of AlexNet, a DL algorithm, to identify pictures of CL lesions in patients from Midwest Brazil. We used a set of 2458 pictures (up to 10 of each lesion) obtained from patients treated between 2015 and 2022 in the Leishmaniasis Clinic at the University Hospital of Brasilia. We divided the picture database into training (80%), internal validation (10%), and testing sets (10%), and trained and tested AlexNet to identify pictures of CL lesions. We performed three simulations and trained AlexNet to differentiate CL from 26 other dermatoses (e.g., chromomycosis, ecthyma, venous insufficiency). We obtained an average accuracy of 95.04% (Confidence Interval 95%: 93.81-96.04), indicating an excellent performance of AlexNet in identifying pictures of CL lesions. We conclude that automated CL identification using AlexNet has the potential to assist clinicians in diagnosing skin lesions. These results contribute to the development of a mobile application to assist in the diagnosis of CL in health care services.

Natural Language Processing to extract SNOMED-CT codes from pathological reports

Objective

The use of standardized structured reports (SSR) and suitable terminologies like SNOMED-CT can enhance data retrieval and analysis, fostering large-scale studies and collaboration. However, the still large prevalence of narrative reports in our laboratories warrants alternative and automated labeling approaches. In this project, natural language processing (NLP) methods were used to associate SNOMED-CT codes to structured and unstructured reports from an Italian Digital Pathology Department.

Methods

Two NLP-based automatic coding systems (support vector machine, SVM, and long-short term memory, LSTM) were trained and applied to a series of narrative reports.

Results

The 1163 cases were tested with both algorithms, showing good performances in terms of accuracy, precision, recall, and F1 score, with SVM showing slightly better performances as compared to LSTM (0.84, 0.87, 0.83, 0.82 vs 0.83, 0.85, 0.83, 0.82, respectively). The integration of an explainability allowed identification of terms and groups of words of importance, enabling fine-tuning, balancing semantic meaning and model performance.

Conclusions

AI tools allow the automatic SNOMED-CT labeling of the pathology archives, providing a retrospective fix to the large lack of organization of narrative reports.

Phenotyping the virulence of SARS-CoV-2 variants in hamsters by digital pathology and machine learning

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has continued to evolve throughout the coronavirus disease-19 (COVID-19) pandemic, giving rise to multiple variants of concern (VOCs) with different biological properties. As the pandemic progresses, it will be essential to test in near real time the potential of any new emerging variant to cause severe disease. BA.1 (Omicron) was shown to be attenuated compared to the previous VOCs like Delta, but it is possible that newly emerging variants may regain a virulent phenotype. Hamsters have been proven to be an exceedingly good model for SARS-CoV-2 pathogenesis. Here, we aimed to develop robust quantitative pipelines to assess the virulence of SARS-CoV-2 variants in hamsters. We used various approaches including RNAseq, RNA in situ hybridization, immunohistochemistry, and digital pathology, including software assisted whole section imaging and downstream automatic analyses enhanced by machine learning, to develop methods to assess and quantify virus-induced pulmonary lesions in an unbiased manner. Initially, we used Delta and Omicron to develop our experimental pipelines. We then assessed the virulence of recent Omicron sub-lineages including BA.5, XBB, BQ.1.18, BA.2, BA.2.75 and EG.5.1. We show that in experimentally infected hamsters, accurate quantification of alveolar epithelial hyperplasia and macrophage infiltrates represent robust markers for assessing the extent of virus-induced pulmonary pathology, and hence virus virulence. In addition, using these pipelines, we could reveal how some Omicron sub-lineages (e.g., BA.2.75 and EG.5.1) have regained virulence compared to the original BA.1. Finally, to maximise the utility of the digital pathology pipelines reported in our study, we developed an online repository containing representative whole organ histopathology sections that can be visualised at variable magnifications (https://covid-atlas.cvr.gla.ac.uk). Overall, this pipeline can provide unbiased and invaluable data for rapidly assessing newly emerging variants and their potential to cause severe disease.

Advancing the PD-L1 CPS test in metastatic TNBC: Insights from pathologists and findings from a nationwide survey

Pembrolizumab has received approval as a first-line treatment for unresectable/metastatic triple-negative breast cancer (mTNBC) with a PD-L1 combined positive score (CPS) of ≥ 10. However, assessing CPS in mTNBC poses challenges. Firstly, it represents a novel analysis for breast pathologists. Secondly, the heterogeneity of PD-L1 expression in mTNBC further complicates the assessment. Lastly, the lack of standardized assays and staining platforms adds to the complexity. In KEYNOTE trials, PD-L1 expression was evaluated using the IHC 22C3 pharmDx kit as a companion diagnostic test. However, both the 22C3 pharmDx and VENTANA PD-L1 (SP263) assays are validated for CPS assessment. Consequently, assay-platform choice, staining conditions, and scoring methods can significantly impact the testing outcomes. This consensus paper aims to discuss the intricacies of PD-L1 CPS testing in mTNBC and provide practical recommendations for pathologists. Additionally, we present findings from a nationwide Italian survey elucidating the state-of-the-art in PD-L1 CPS testing in mTNBC.