PathoNet introduced as a deep neural network backend for evaluation of Ki-67 and tumor-infiltrating lymphocytes in breast cancer

Assessment of the current and emerging criteria for the histopathological classification of lung neuroendocrine tumours in the lungNENomics project

BACKGROUND

Six thoracic pathologists reviewed 259 lung neuroendocrine tumours (LNETs) from the lungNENomics project, with 171 of them having associated survival data. This cohort presents a unique opportunity to assess the strengths and limitations of current World Health Organization (WHO) classification criteria and to evaluate the utility of emerging markers.

PATIENTS AND METHODS

Patients were diagnosed based on the 2021 WHO criteria, with atypical carcinoids (ACs) defined by the presence of focal necrosis and/or 2-10 mitoses per 2 mm2. We investigated two markers of tumour proliferation: the Ki-67 index and phospho-histone H3 (PHH3) protein expression, quantified by pathologists and automatically via deep learning. Additionally, an unsupervised deep learning algorithm was trained to uncover previously unnoticed morphological features with diagnostic value.

RESULTS

The accuracy in distinguishing typical from ACs is hampered by interobserver variability in mitotic counting and the limitations of morphological criteria in identifying aggressive cases. Our study reveals that different Ki-67 cut-offs can categorise LNETs similarly to current WHO criteria. Counting mitoses in PHH3+ areas does not improve diagnosis, while providing a similar prognostic value to the current criteria. With the advantage of being time efficient, automated assessment of these markers leads to similar conclusions. Lastly, state-of-the-art deep learning modelling does not uncover undisclosed morphological features with diagnostic value.

CONCLUSIONS

This study suggests that the mitotic criteria can be complemented by manual or automated assessment of Ki-67 or PHH3 protein expression, but these markers do not significantly improve the prognostic value of the current classification, as the AC group remains highly unspecific for aggressive cases. Therefore, we may have exhausted the potential of morphological features in classifying and prognosticating LNETs. Our study suggests that it might be time to shift the research focus towards investigating molecular markers that could contribute to a more clinically relevant morpho-molecular classification.

Comparison between a deep-learning and a pixel-based approach for the automated quantification of HIV target cells in foreskin tissue

The availability of target cells expressing the HIV receptors CD4 and CCR5 in genital tissue is a critical determinant of HIV susceptibility during sexual transmission. Quantification of immune cells in genital tissue is therefore an important outcome for studies on HIV susceptibility and prevention. Immunofluorescence microscopy allows for precise visualization of immune cells in mucosal tissues; however, this technique is limited in clinical studies by the lack of an accurate, unbiased, high-throughput image analysis method. Current pixel-based thresholding methods for cell counting struggle in tissue regions with high cell density and autofluorescence, both of which are common features in genital tissue. We describe a deep-learning approach using the publicly available StarDist method to count cells in immunofluorescence microscopy images of foreskin stained for nuclei, CD3, CD4, and CCR5. The accuracy of the model was comparable to manual counting (gold standard) and surpassed the capability of a previously described pixel-based cell counting method. We show that the performance of our deep-learning model is robust in tissue regions with high cell density and high autofluorescence. Moreover, we show that this deep-learning analysis method is both easy to implement and to adapt for the identification of other cell types in genital mucosal tissue.

AI-Powered Segmentation of Invasive Carcinoma Regions in Breast Cancer Immunohistochemical Whole-Slide Images

AIMS

The automation of quantitative evaluation for breast immunohistochemistry (IHC) plays a crucial role in reducing the workload of pathologists and enhancing the objectivity of diagnoses. However, current methods face challenges in achieving fully automated immunohistochemistry quantification due to the complexity of segmenting the tumor area into distinct ductal carcinoma in situ (DCIS) and invasive carcinoma (IC) regions. Moreover, the quantitative analysis of immunohistochemistry requires a specific focus on invasive carcinoma regions.

METHODS AND RESULTS

In this study, we propose an innovative approach to automatically identify invasive carcinoma regions in breast cancer immunohistochemistry whole-slide images (WSIs). Our method leverages a neural network that combines multi-scale morphological features with boundary features, enabling precise segmentation of invasive carcinoma regions without the need for additional H&E and P63 staining slides. In addition, we introduced an advanced semi-supervised learning algorithm, allowing efficient training of the model using unlabeled data. To evaluate the effectiveness of our approach, we constructed a dataset consisting of 618 IHC-stained WSIs from 170 cases, including four types of staining (ER, PR, HER2, and Ki-67). Notably, the model demonstrated an impressive intersection over union (IoU) score exceeding 80% on the test set. Furthermore, to ascertain the practical utility of our model in IHC quantitative evaluation, we constructed a fully automated Ki-67 scoring system based on the model's predictions. Comparative experiments convincingly demonstrated that our system exhibited high consistency with the scores given by experienced pathologists.

CONCLUSIONS

Our developed model excels in accurately distinguishing between DCIS and invasive carcinoma regions in breast cancer immunohistochemistry WSIs. This method paves the way for a clinically available, fully automated immunohistochemistry quantitative scoring system.

Lymphocyte detection for cancer analysis using a novel fusion block based channel boosted CNN

Tumor-infiltrating lymphocytes, specialized immune cells, are considered an important biomarker in cancer analysis. Automated lymphocyte detection is challenging due to its heterogeneous morphology, variable distribution, and presence of artifacts. In this work, we propose a novel Boosted Channels Fusion-based CNN "BCF-Lym-Detector" for lymphocyte detection in multiple cancer histology images. The proposed network initially selects candidate lymphocytic regions at the tissue level and then detects lymphocytes at the cellular level. The proposed "BCF-Lym-Detector" generates diverse boosted channels by utilizing the feature learning capability of different CNN architectures. In this connection, a new adaptive fusion block is developed to combine and select the most relevant lymphocyte-specific features from the generated enriched feature space. Multi-level feature learning is used to retain lymphocytic spatial information and detect lymphocytes with variable appearances. The assessment of the proposed "BCF-Lym-Detector" show substantial improvement in terms of F-score (0.93 and 0.84 on LYSTO and NuClick, respectively), which suggests that the diverse feature extraction and dynamic feature selection enhanced the feature learning capacity of the proposed network. Moreover, the proposed technique's generalization on unseen test sets with a good recall (0.75) and F-score (0.73) shows its potential use for pathologists' assistance.

SPARTIN: a Bayesian method for the quantification and characterization of cell type interactions in spatial pathology data

Introduction: The acquisition of high-resolution digital pathology imaging data has sparked the development of methods to extract context-specific features from such complex data. In the context of cancer, this has led to increased exploration of the tumor microenvironment with respect to the presence and spatial composition of immune cells. Spatial statistical modeling of the immune microenvironment may yield insights into the role played by the immune system in the natural development of cancer as well as downstream therapeutic interventions. Methods: In this paper, we present SPatial Analysis of paRtitioned Tumor-Immune imagiNg (SPARTIN), a Bayesian method for the spatial quantification of immune cell infiltration from pathology images. SPARTIN uses Bayesian point processes to characterize a novel measure of local tumor-immune cell interaction, Cell Type Interaction Probability (CTIP). CTIP allows rigorous incorporation of uncertainty and is highly interpretable, both within and across biopsies, and can be used to assess associations with genomic and clinical features. Results: Through simulations, we show SPARTIN can accurately distinguish various patterns of cellular interactions as compared to existing methods. Using SPARTIN, we characterized the local spatial immune cell infiltration within and across 335 melanoma biopsies and evaluated their association with genomic, phenotypic, and clinical outcomes. We found that CTIP was significantly (negatively) associated with deconvolved immune cell prevalence scores including CD8+ T-Cells and Natural Killer cells. Furthermore, average CTIP scores differed significantly across previously established transcriptomic classes and significantly associated with survival outcomes. Discussion: SPARTIN provides a general framework for investigating spatial cellular interactions in high-resolution digital histopathology imaging data and its associations with patient level characteristics. The results of our analysis have potential implications relevant to both treatment and prognosis in the context of Skin Cutaneous Melanoma. The R-package for SPARTIN is available at https://github.com/bayesrx/SPARTIN along with a visualization tool for the images and results at: https://nateosher.github.io/SPARTIN.

Unsupervised domain adaptive tumor region recognition for Ki67 automated assisted quantification

PURPOSE

Ki67 is a protein associated with tumor proliferation and metastasis in breast cancer and acts as an essential prognostic factor. Clinical work requires recognizing tumor regions on Ki67-stained whole-slide images (WSIs) before quantitation. Deep learning has the potential to provide assistance but largely relies on massive annotations and consumes a huge amount of time and energy. Hence, a novel tumor region recognition approach is proposed for more precise Ki67 quantification.

METHODS

An unsupervised domain adaptive method is proposed, which combines adversarial and self-training. The model trained on labeled hematoxylin and eosin (H&E) data and unlabeled Ki67 data can recognize tumor regions in Ki67 WSIs. Based on the UDA method, a Ki67 automated assisted quantification system is developed, which contains foreground segmentation, tumor region recognition, cell counting, and WSI-level score calculation.

RESULTS

The proposed UDA method achieves high performance in tumor region recognition and Ki67 quantification. The AUC reached 0.9915, 0.9352, and 0.9689 on the validation set and internal and external test sets, respectively, substantially exceeding baseline (0.9334, 0.9167, 0.9408) and rivaling the fully supervised method (0.9950, 0.9284, 0.9652). The evaluation of automated quantification on 148 WSIs illustrated statistical agreement with pathological reports.

CONCLUSION

The model trained by the proposed method is capable of accurately recognizing Ki67 tumor regions. The proposed UDA method can be readily extended to other types of immunohistochemical staining images. The results of automated assisted quantification are accurate and interpretable to provide assistance to both junior and senior pathologists in their interpretation.

The Approach of Artificial Intelligence in Neuroendocrine Carcinomas of the Breast: A Next Step towards Precision Pathology?—A Case Report and Review of the Literature

Primary neuroendocrine tumors (NETs) of the breast are considered a rare and undervalued subtype of breast carcinoma that occur mainly in postmenopausal women and are graded as G1 or G2 NETs or an invasive neuroendocrine carcinoma (NEC) (small cell or large cell). To establish a final diagnosis of breast carcinoma with neuroendocrine differentiation, it is essential to perform an immunohistochemical profile of the tumor, using antibodies against synaptophysin or chromogranin, as well as the MIB-1 proliferation index, one of the most controversial markers in breast pathology regarding its methodology in current clinical practice. A standardization error between institutions and pathologists regarding the evaluation of the MIB-1 proliferation index is present. Another challenge refers to the counting process of MIB-1′s expressiveness, which is known as a time-consuming process. The involvement of AI (artificial intelligence) automated systems could be a solution for diagnosing early stages, as well. We present the case of a post-menopausal 79-year-old woman diagnosed with primary neuroendocrine carcinoma of the breast (NECB). The purpose of this paper is to expose the interpretation of MIB-1 expression in our patient’ s case of breast neuroendocrine carcinoma, assisted by artificial intelligence (AI) software (HALO—IndicaLabs), and to analyze the associations between MIB-1 and common histopathological parameters.

Artificial intelligence in breast cancer histopathology

This is a review on the use of artificial intelligence for digital breast pathology. A systematic search on PubMed was conducted, identifying 17,324 research papers related to breast cancer pathology. Following a semimanual screening, 664 papers were retrieved and pursued. The papers are grouped into six major tasks performed by pathologists-namely, molecular and hormonal analysis, grading, mitotic figure counting, ki-67 indexing, tumour-infiltrating lymphocyte assessment, and lymph node metastases identification. Under each task, open-source datasets for research to build artificial intelligence (AI) tools are also listed. Many AI tools showed promise and demonstrated feasibility in the automation of routine pathology investigations. We expect continued growth of AI in this field as new algorithms mature.

An integrative non-invasive malignant brain tumors classification and Ki-67 labeling index prediction pipeline with radiomics approach

BACKGROUND

The histological sub-classes of brain tumors and the Ki-67 labeling index (LI) of tumor cells are major factors in the diagnosis, prognosis, and treatment management of patients. Many existing studies primarily focused on the classification of two classes of brain tumors and the Ki-67LI of gliomas. This study aimed to develop a preoperative non-invasive radiomics pipeline based on multiparametric-MRI to classify-three types of brain tumors, glioblastoma (GBM), metastasis (MET) and primary central nervous system lymphoma (PCNSL), and to predict their corresponding Ki-67LI.

METHODS

In this retrospective study, 153 patients with malignant brain tumors were involved. The radiomics features were extracted from three types of MRI (T1-weighted imaging (T1WI), fluid-attenuated inversion recovery (FLAIR), and contrast-enhanced T1-weighted imaging (CE-T1WI)) with three masks (tumor core, edema, and whole tumor masks) and selected by a combination of Pearson correlation coefficient (CORR), LASSO, and Max-Relevance and Min-Redundancy (mRMR) filters. The performance of six classifiers was compared and the top three performing classifiers were used to construct the ensemble learning model (ELM). The proposed ELM was evaluated in the training dataset (108 patients) by 5-fold cross-validation and in the test dataset (45 patients) by hold-out. The accuracy (ACC), sensitivity (SEN), specificity (SPE), F1-Score, and the area under the receiver operating characteristic curve (AUC) indicators evaluated the performance of the models.

RESULTS

The best feature sets and ELM with the optimal performance were selected to construct the tri-categorized brain tumor aided diagnosis model (training dataset AUC: 0.96 (95% CI: 0.93, 0.99); test dataset AUC: 0.93) and Ki-67LI prediction model (training dataset AUC: 0.96 (95% CI: 0.94, 0.98); test dataset AUC: 0.91). The CE-T1WI was the best single modality for all classifiers. Meanwhile, the whole tumor was the most vital mask for the tumor classification and the tumor core was the most vital mask for the Ki-67LI prediction.

CONCLUSION

The developed radiomics models led to the precise preoperative classification of GBM, MET, and PCNSL and the prediction of Ki-67LI, which could be utilized in clinical practice for the treatment planning for brain tumors.

Comparison Between Manual and Automated Assessment of Ki-67 in Breast Carcinoma: Test of a Simple Method in Daily Practice

BACKGROUND

In the era of "precision medicine," the availability of high-quality tumor biomarker tests is critical and tumor proliferation evaluated by Ki-67 antibody is one of the most important prognostic factors in breast cancer. But the evaluation of Ki-67 index has been shown to suffer from some interobserver variability. The goal of the study is to develop an easy, automated, and reliable Ki-67 assessment approach for invasive breast carcinoma in routine practice.

PATIENTS AND METHODS

A total of 151 biopsies of invasive breast carcinoma were analyzed. The Ki-67 index was evaluated by 2 pathologists with MIB-1 antibody as a global tumor index and also in a hotspot. These 2 areas were also analyzed by digital image analysis (DIA).

RESULTS

For Ki-67 index assessment, in the global and hotspot tumor area, the concordances were very good between DIA and pathologists when DIA focused on the annotations made by pathologist (0.73 and 0.83, respectively). However, this was definitely not the case when DIA was not constrained within the pathologist's annotations and automatically established its global or hotspot area in the whole tissue sample (concordance correlation coefficients between 0.28 and 0.58).

CONCLUSIONS

The DIA technique demonstrated a meaningful concordance with the indices evaluated by pathologists when the tumor area is previously identified by a pathologist. In contrast, basing Ki-67 assessment on automatic tissue detection was not satisfactory and provided bad concordance results. A representative tumoral zone must therefore be manually selected prior to the measurement made by the DIA.

Construction of machine learning-based models for cancer outcomes in low and lower-middle income countries: A scoping review

Background

The impact and utility of machine learning (ML)-based prediction tools for cancer outcomes including assistive diagnosis, risk stratification, and adjunctive decision-making have been largely described and realized in the high income and upper-middle-income countries. However, statistical projections have estimated higher cancer incidence and mortality risks in low and lower-middle-income countries (LLMICs). Therefore, this review aimed to evaluate the utilization, model construction methods, and degree of implementation of ML-based models for cancer outcomes in LLMICs.

Methods

PubMed/Medline, Scopus, and Web of Science databases were searched and articles describing the use of ML-based models for cancer among local populations in LLMICs between 2002 and 2022 were included. A total of 140 articles from 22,516 citations that met the eligibility criteria were included in this study.

Results

ML-based models from LLMICs were often based on traditional ML algorithms than deep or deep hybrid learning. We found that the construction of ML-based models was skewed to particular LLMICs such as India, Iran, Pakistan, and Egypt with a paucity of applications in sub-Saharan Africa. Moreover, models for breast, head and neck, and brain cancer outcomes were frequently explored. Many models were deemed suboptimal according to the Prediction model Risk of Bias Assessment tool (PROBAST) due to sample size constraints and technical flaws in ML modeling even though their performance accuracy ranged from 0.65 to 1.00. While the development and internal validation were described for all models included (n=137), only 4.4% (6/137) have been validated in independent cohorts and 0.7% (1/137) have been assessed for clinical impact and efficacy.

Conclusion

Overall, the application of ML for modeling cancer outcomes in LLMICs is increasing. However, model development is largely unsatisfactory. We recommend model retraining using larger sample sizes, intensified external validation practices, and increased impact assessment studies using randomized controlled trial designs

Systematic review registration

https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=308345, identifier CRD42022308345.

Automatic CD30 scoring method for whole slide images of primary cutaneous CD30+ lymphoproliferative diseases

AIMS

Deep-learning methods for scoring biomarkers are an active research topic. However, the superior performance of many studies relies on large datasets collected from clinical samples. In addition, there are fewer studies on immunohistochemical marker assessment for dermatological diseases. Accordingly, we developed a method for scoring CD30 based on convolutional neural networks for a few primary cutaneous CD30+ lymphoproliferative disorders and used this method to evaluate other biomarkers.

METHODS

A multipatch spatial attention mechanism and conditional random field algorithm were used to fully fuse tumour tissue characteristics on immunohistochemical slides and alleviate the few sample feature deficits. We trained and tested 28 CD30+ immunohistochemical whole slide images (WSIs), evaluated them with a performance index, and compared them with the diagnoses of senior dermatologists. Finally, the model's performance was further demonstrated on the publicly available Yale HER2 cohort.

RESULTS

Compared with the diagnoses by senior dermatologists, this method can better locate the tumour area and reduce the misdiagnosis rate. The prediction of CD3 and Ki-67 validated the model's ability to identify other biomarkers.

CONCLUSIONS

In this study, using a few immunohistochemical WSIs, our model can accurately identify CD30, CD3 and Ki-67 markers. In addition, the model could be applied to additional tumour identification tasks to aid pathologists in diagnosis and benefit clinical evaluation.

A Review of Watershed Implementations for Segmentation of Volumetric Images

Watershed is a widely used image segmentation algorithm. Most researchers understand just an idea of this method: a grayscale image is considered as topographic relief, which is flooded from initial basins. However, frequently they are not aware of the options of the algorithm and the peculiarities of its realizations. There are many watershed implementations in software packages and products. Even if these packages are based on the identical algorithm-watershed, by flooding their outcomes, processing speed, and consumed memory, vary greatly. In particular, the difference among various implementations is noticeable for huge volumetric images; for instance, tomographic 3D images, for which low performance and high memory requirements of watershed might be bottlenecks. In our review, we discuss the peculiarities of algorithms with and without waterline generation, the impact of connectivity type and relief quantization level on the result, approaches for parallelization, as well as other method options. We present detailed benchmarking of seven open-source and three commercial software implementations of marker-controlled watershed for semantic or instance segmentation. We compare those software packages for one synthetic and two natural volumetric images. The aim of the review is to provide information and advice for practitioners to select the appropriate version of watershed for their problem solving. In addition, we forecast future directions of software development for 3D image segmentation by watershed.

Deep Learning-Inferred Multiplex ImmunoFluorescence for Immunohistochemical Image Quantification

Reporting biomarkers assessed by routine immunohistochemical (IHC) staining of tissue is broadly used in diagnostic pathology laboratories for patient care. To date, clinical reporting is predominantly qualitative or semi-quantitative. By creating a multitask deep learning framework referred to as DeepLIIF, we present a single-step solution to stain deconvolution/separation, cell segmentation, and quantitative single-cell IHC scoring. Leveraging a unique de novo dataset of co-registered IHC and multiplex immunofluorescence (mpIF) staining of the same slides, we segment and translate low-cost and prevalent IHC slides to more expensive-yet-informative mpIF images, while simultaneously providing the essential ground truth for the superimposed brightfield IHC channels. Moreover, a new nuclear-envelop stain, LAP2beta, with high (>95%) cell coverage is introduced to improve cell delineation/segmentation and protein expression quantification on IHC slides. By simultaneously translating input IHC images to clean/separated mpIF channels and performing cell segmentation/classification, we show that our model trained on clean IHC Ki67 data can generalize to more noisy and artifact-ridden images as well as other nuclear and non-nuclear markers such as CD3, CD8, BCL2, BCL6, MYC, MUM1, CD10, and TP53. We thoroughly evaluate our method on publicly available benchmark datasets as well as against pathologists' semi-quantitative scoring. The code, the pre-trained models, along with easy-to-run containerized docker files as well as Google CoLab project are available at https://github.com/nadeemlab/deepliif.

Assessment of Ki-67 proliferation index with deep learning in DCIS (ductal carcinoma in situ)

The proliferation index (PI) is crucial in histopathologic diagnostics, in particular tumors. It is calculated based on Ki-67 protein expression by immunohistochemistry. PI is routinely evaluated by a visual assessment of the sample by a pathologist. However, this approach is far from ideal due to its poor intra- and interobserver variability and time-consuming. These factors force the community to seek out more precise solutions. Virtual pathology as being increasingly popular in diagnostics, armed with artificial intelligence, may potentially address this issue. The proposed solution calculates the Ki-67 proliferation index by utilizing a deep learning model and fuzzy-set interpretations for hot-spots detection. The obtained region-of-interest is then used to segment relevant cells via classical methods of image processing. The index value is approximated by relating the total surface area occupied by immunopositive cells to the total surface area of relevant cells. The achieved results are compared to the manual calculation of the Ki-67 index made by a domain expert. To increase results reliability, we trained several models in a threefold manner and compared the impact of different hyper-parameters. Our best-proposed method estimates PI with 0.024 mean absolute error, which gives a significant advantage over the current state-of-the-art solution.

Deep Learning-Based Mapping of Tumor Infiltrating Lymphocytes in Whole Slide Images of 23 Types of Cancer

The role of tumor infiltrating lymphocytes (TILs) as a biomarker to predict disease progression and clinical outcomes has generated tremendous interest in translational cancer research. We present an updated and enhanced deep learning workflow to classify 50x50 um tiled image patches (100x100 pixels at 20x magnification) as TIL positive or negative based on the presence of 2 or more TILs in gigapixel whole slide images (WSIs) from the Cancer Genome Atlas (TCGA). This workflow generates TIL maps to study the abundance and spatial distribution of TILs in 23 different types of cancer. We trained three state-of-the-art, popular convolutional neural network (CNN) architectures (namely VGG16, Inception-V4, and ResNet-34) with a large volume of training data, which combined manual annotations from pathologists (strong annotations) and computer-generated labels from our previously reported first-generation TIL model for 13 cancer types (model-generated annotations). Specifically, this training dataset contains TIL positive and negative patches from cancers in additional organ sites and curated data to help improve algorithmic performance by decreasing known false positives and false negatives. Our new TIL workflow also incorporates automated thresholding to convert model predictions into binary classifications to generate TIL maps. The new TIL models all achieve better performance with improvements of up to 13% in accuracy and 15% in F-score. We report these new TIL models and a curated dataset of TIL maps, referred to as TIL-Maps-23, for 7983 WSIs spanning 23 types of cancer with complex and diverse visual appearances, which will be publicly available along with the code to evaluate performance. Code Available at: https://github.com/ShahiraAbousamra/til_classification.

Compound computer vision workflow for efficient and automated immunohistochemical analysis of whole slide images

AIMS

Immunohistochemistry (IHC) assessment of tissue is a central component of the modern pathology workflow, but quantification is challenged by subjective estimates by pathologists or manual steps in semi-automated digital tools. This study integrates various computer vision tools to develop a fully automated workflow for quantifying Ki-67, a standard IHC test used to assess cell proliferation on digital whole slide images (WSIs).

METHODS

We create an automated nuclear segmentation strategy by deploying a Mask R-CNN classifier to recognise and count 3,3'-diaminobenzidine positive and negative nuclei. To further improve automation, we replaced manual selection of regions of interest (ROIs) by aligning Ki-67 WSIs with corresponding H&E-stained sections, using scale-invariant feature transform (SIFT) and a conventional histomorphological convolutional neural networks to define tumour-rich areas for quantification.

RESULTS

The Mask R-CNN was tested on 147 images generated from 34 brain tumour Ki-67 WSIs and showed a high concordance with aggregate pathologists' estimates ([Formula: see text] assessors; [Formula: see text] r=0.9750). Concordance of each assessor's Ki-67 estimates was higher when compared with the Mask R-CNN than between individual assessors (r_avg=0.9322 vs 0.8703; p=0.0213). Coupling the Mask R-CNN with SIFT-CNN workflow demonstrated ROIs can be automatically chosen and partially sampled to improve automation and dramatically decrease computational time (average: 88.55-19.28 min; p<0.0001).

CONCLUSIONS

We show how innovations in computer vision can be serially compounded to automate and improve implementation in clinical workflows. Generalisation of this approach to other ancillary studies has significant implications for computational pathology.

Deep Neural Network-Aided Histopathological Analysis of Myocardial Injury

Deep neural networks have become the mainstream approach for analyzing and interpreting histology images. In this study, we established and validated an interpretable DNN model to assess endomyocardial biopsy (EMB) data of patients with myocardial injury. Deep learning models were used to extract features and classify EMB histopathological images of heart failure cases diagnosed with either ischemic cardiomyopathy or idiopathic dilated cardiomyopathy and non-failing cases (organ donors without a history of heart failure). We utilized the gradient-weighted class activation mapping (Grad-CAM) technique to emphasize injured regions, providing an entry point to assess the dominant morphology in the process of a comprehensive evaluation. To visualize clustered regions of interest (ROI), we utilized uniform manifold approximation and projection (UMAP) embedding for dimension reduction. We further implemented a multi-model ensemble mechanism to improve the quantitative metric (area under the receiver operating characteristic curve, AUC) to 0.985 and 0.992 on ROI-level and case-level, respectively, outperforming the achievement of 0.971 ± 0.017 and 0.981 ± 0.020 based on the sub-models. Collectively, this new methodology provides a robust and interpretive framework to explore local histopathological patterns, facilitating the automatic and high-throughput quantification of cardiac EMB analysis.

CDKN1C as a prognostic biomarker correlated with immune infiltrates and therapeutic responses in breast cancer patients

Breast cancer (BC) prognosis and therapeutic sensitivity could not be predicted efficiently. Previous evidence have shown the vital roles of CDKN1C in BC. Therefore, we aimed to construct a CDKN1C‐based model to accurately predicting overall survival (OS) and treatment responses in BC patients. In this study, 995 BC patients from The Cancer Genome Atlas database were selected. Kaplan‐Meier curve, Gene set enrichment and immune infiltrates analyses were executed. We developed a novel CDKN1C‐based nomogram to predict the OS, verified by the time‐dependent receiver operating characteristic curve, calibration curve and decision curve. Therapeutic response prediction was followed based on the low‐ and high‐nomogram score groups. Our results indicated that low‐CDKN1C expression was associated with shorter OS and lower proportion of naïve B cells, CD8 T cells, activated NK cells. The predictive accuracy of the nomogram for 5‐year OS was superior to the tumour‐node‐metastasis stage (area under the curve: 0.746 vs. 0.634, p < 0.001). The nomogram exhibited excellent predictive performance, calibration ability and clinical utility. Moreover, low‐risk patients were identified with stronger sensitivity to therapeutic agents. This tool can improve BC prognosis and therapeutic responses prediction, thus guiding individualized treatment decisions.

Quantification of the Immune Content in Neuroblastoma: Deep Learning and Topological Data Analysis in Digital Pathology

We introduce here a novel machine learning (ML) framework to address the issue of the quantitative assessment of the immune content in neuroblastoma (NB) specimens. First, the EUNet, a U-Net with an EfficientNet encoder, is trained to detect lymphocytes on tissue digital slides stained with the CD3 T-cell marker. The training set consists of 3782 images extracted from an original collection of 54 whole slide images (WSIs), manually annotated for a total of 73,751 lymphocytes. Resampling strategies, data augmentation, and transfer learning approaches are adopted to warrant reproducibility and to reduce the risk of overfitting and selection bias. Topological data analysis (TDA) is then used to define activation maps from different layers of the neural network at different stages of the training process, described by persistence diagrams (PD) and Betti curves. TDA is further integrated with the uniform manifold approximation and projection (UMAP) dimensionality reduction and the hierarchical density-based spatial clustering of applications with noise (HDBSCAN) algorithm for clustering, by the deep features, the relevant subgroups and structures, across different levels of the neural network. Finally, the recent TwoNN approach is leveraged to study the variation of the intrinsic dimensionality of the U-Net model. As the main task, the proposed pipeline is employed to evaluate the density of lymphocytes over the whole tissue area of the WSIs. The model achieves good results with mean absolute error 3.1 on test set, showing significant agreement between densities estimated by our EUNet model and by trained pathologists, thus indicating the potentialities of a promising new strategy in the quantification of the immune content in NB specimens. Moreover, the UMAP algorithm unveiled interesting patterns compatible with pathological characteristics, also highlighting novel insights into the dynamics of the intrinsic dataset dimensionality at different stages of the training process. All the experiments were run on the Microsoft Azure cloud platform.