Diagnostic concordance among pathologists interpreting breast biopsy specimens

Variability in grading of ductal carcinoma in situ among an international group of pathologists

The prognostic value of cytonuclear grade in ductal carcinoma in situ (DCIS) is debated, partly due to high interobserver variability and the use of multiple guidelines. The aim of this study was to evaluate interobserver agreement in grading DCIS between Dutch, British, and American pathologists. Haematoxylin and eosin-stained slides of 425 women with primary DCIS were independently reviewed by nine breast pathologists based in the Netherlands, the UK, and the USA. Chance-corrected kappa (κma ) for association between pathologists was calculated based on a generalised linear mixed model using the ordinal package in R. Overall κma for grade of DCIS (low, intermediate, or high) was estimated to be 0.50 (95% confidence interval [CI] 0.44-0.56), indicating a moderate association between pathologists. When the model was adjusted for national guidelines, the association for grade did not change (κma = 0.53; 95% CI 0.48-0.57); subgroup analysis for pathologists using the UK pathology guidelines only had significantly higher association (κma = 0.58; 95% CI 0.56-0.61). To assess if concordance of grading relates to the expression of the oestrogen receptor (ER) and HER2, archived immunohistochemistry was analysed on a subgroup (n = 106). This showed that non-high grade according to the majority opinion was associated with ER positivity and HER2 negativity (100 and 89% of non-high grade cases, respectively). In conclusion, DCIS grade showed only moderate association using whole slide images scored by nine breast pathologists. As therapeutic decisions and inclusion in ongoing clinical trials are guided by DCIS grade, there is a pressing need to reduce interobserver variability in grading. ER and HER2 might be supportive to prevent the accidental and unwanted inclusion of high-grade DCIS in such trials.

Improving DCIS diagnosis and predictive outcome by applying artificial intelligence

Breast ductal carcinoma in situ (DCIS) is a preinvasive lesion that is considered to be a precursor to invasive breast cancer. Nevertheless, not all DCIS will progress to invasion. Current histopathological classification systems are unable to predict which cases will or will not progress, and therefore many women with DCIS may be overtreated. Artificial intelligence (AI) image-based analysis methods have potential to identify and analyze novel features that may facilitate tumor identification, prediction of disease outcome and response to treatment. Indeed, these methods prove promising for accurately identifying DCIS lesions, and show potential clinical utility in the therapeutic stratification of DCIS patients. Here, we review how AI techniques in histopathology may aid diagnosis and clinical decisions in regards to DCIS, and how such techniques could be incorporated into clinical practice.

Classification of Hematoxylin and Eosin-Stained Breast Cancer Histology Microscopy Images Using Transfer Learning with EfficientNets

Breast cancer is a fatal disease and is a leading cause of death in women worldwide. The process of diagnosis based on biopsy tissue is nontrivial, time-consuming, and prone to human error, and there may be conflict about the final diagnosis due to interobserver variability. Computer-aided diagnosis systems have been designed and implemented to combat these issues. These systems contribute significantly to increasing the efficiency and accuracy and reducing the cost of diagnosis. Moreover, these systems must perform better so that their determined diagnosis can be more reliable. This research investigates the application of the EfficientNet architecture for the classification of hematoxylin and eosin-stained breast cancer histology images provided by the ICIAR2018 dataset. Specifically, seven EfficientNets were fine-tuned and evaluated on their ability to classify images into four classes: normal, benign, in situ carcinoma, and invasive carcinoma. Moreover, two standard stain normalization techniques, Reinhard and Macenko, were observed to measure the impact of stain normalization on performance. The outcome of this approach reveals that the EfficientNet-B2 model yielded an accuracy and sensitivity of 98.33% using Reinhard stain normalization method on the training images and an accuracy and sensitivity of 96.67% using the Macenko stain normalization method. These satisfactory results indicate that transferring generic features from natural images to medical images through fine-tuning on EfficientNets can achieve satisfactory results.

Selection of Representative Histologic Slides in Interobserver Reproducibility Studies: Insights from Expert Review for Ovarian Carcinoma Subtype Classification

BACKGROUND

Observer studies in pathology often utilize a limited number of representative slides per case, selected and reported in a nonstandardized manner. Reference diagnoses are commonly assumed to be generalizable to all slides of a case. We examined these issues in the context of pathologist concordance for histologic subtype classification of ovarian carcinomas (OCs).

MATERIALS AND METHODS

A cohort of 114 OCs consisting of 72 cases with a single representative slide (Group 1) and 42 cases with multiple representative slides (148 slides, 2-6 sections per case, Group 2) was independently reviewed by three experts in gynecologic pathology (case-based review). In a follow-up study, each individual slide was independently reviewed in a randomized order by the same pathologists (section-based review).

RESULTS

Average interobserver concordance varied from 100% for Group 1 to 64.3% for Group 2 (86.8% across all cases). Across Group 2, 19 cases (45.2%) had at least one slide classified as a different subtype than the subtype assigned from case-based review, demonstrating the impact of intratumoral heterogeneity. Section-based concordance across individual sections from Group 2 was comparable to case-based concordance for those cases indicating diagnostic challenges at the individual section level. Findings demonstrate the increased diagnostic complexity of heterogeneous tumors that require multiple section sampling and its impact on pathologist performance.

CONCLUSIONS

The proportion of cases with multiple representative slides in cohorts used in validation studies, such as those conducted to evaluate artificial intelligence/machine learning tools, can influence diagnostic performance, and if not accounted for, can cause disparities between research and real-world observations and between research studies. Case selection in validation studies should account for tumor heterogeneity to create balanced datasets in terms of diagnostic complexity.

A multi-phase deep CNN based mitosis detection framework for breast cancer histopathological images

The mitotic activity index is a key prognostic measure in tumour grading. Microscopy based detection of mitotic nuclei is a significant overhead and necessitates automation. This work proposes deep CNN based multi-phase mitosis detection framework "MP-MitDet" for mitotic nuclei identification in breast cancer histopathological images. The workflow constitutes: (1) label-refiner, (2) tissue-level mitotic region selection, (3) blob analysis, and (4) cell-level refinement. We developed an automatic label-refiner to represent weak labels with semi-sematic information for training of deep CNNs. A deep instance-based detection and segmentation model is used to explore probable mitotic regions on tissue patches. More probable regions are screened based on blob area and then analysed at cell-level by developing a custom CNN classifier "MitosRes-CNN" to filter false mitoses. The performance of the proposed "MitosRes-CNN" is compared with the state-of-the-art CNNs that are adapted to cell-level discrimination through cross-domain transfer learning and by adding task-specific layers. The performance of the proposed framework shows good discrimination ability in terms of F-score (0.75), recall (0.76), precision (0.71) and area under the precision-recall curve (0.78) on challenging TUPAC16 dataset. Promising results suggest good generalization of the proposed framework that can learn characteristic features from heterogenous mitotic nuclei.

Artificial Intelligence Improves the Accuracy in Histologic Classification of Breast Lesions

OBJECTIVES

This study evaluated the usefulness of artificial intelligence (AI) algorithms as tools in improving the accuracy of histologic classification of breast tissue.

METHODS

Overall, 100 microscopic photographs (test A) and 152 regions of interest in whole-slide images (test B) of breast tissue were classified into 4 classes: normal, benign, carcinoma in situ (CIS), and invasive carcinoma. The accuracy of 4 pathologists and 3 pathology residents were evaluated without and with the assistance of algorithms.

RESULTS

In test A, algorithm A had accuracy of 0.87, with the lowest accuracy in the benign class (0.72). The observers had average accuracy of 0.80, and most clinically relevant discordances occurred in distinguishing benign from CIS (7.1% of classifications). With the assistance of algorithm A, the observers significantly increased their average accuracy to 0.88. In test B, algorithm B had accuracy of 0.49, with the lowest accuracy in the CIS class (0.06). The observers had average accuracy of 0.86, and most clinically relevant discordances occurred in distinguishing benign from CIS (6.3% of classifications). With the assistance of algorithm B, the observers maintained their average accuracy.

CONCLUSIONS

AI tools can increase the classification accuracy of pathologists in the setting of breast lesions.

Disease diagnostic coding to facilitate evidence-based medicine: current and future perspectives

Technologic advances in information management have rapidly changed laboratory testing and the practice of veterinary medicine. Timely and strategic sampling, same-day assays, and 24-h access to laboratory results allow for rapid implementation of intervention and treatment protocols. Although agent detection and monitoring systems have progressed, and wider tracking of diseases across veterinary diagnostic laboratories exists, such as by the National Animal Health Laboratory Network (NAHLN), the distinction between detection of agent and manifestation of disease is critical to improved disease management. The implementation of a consistent, intuitive, and useful disease diagnosis coding system, specific for veterinary medicine and applicable to multiple animal species within and between veterinary diagnostic laboratories, is the first phase of disease data aggregation. Feedback loops for continuous improvement that could aggregate existing clinical and laboratory databases to improve the value and applications of diagnostic processes and clinical interventions, with interactive capabilities between clinicians and diagnosticians, and that differentiate disease causation from mere agent detection, remain incomplete. Creating an interface that allows aggregation of existing data from clinicians, including final diagnosis, interventions, or treatments applied, and measures of outcomes, is the second phase. Prototypes for stakeholder cooperation, collaboration, and beta testing of this vision are in development and becoming a reality. We focus here on how such a system is being developed and utilized at the Iowa State University Veterinary Diagnostic Laboratory to facilitate evidence-based medicine and utilize diagnostic coding for continuous improvement of animal health and welfare.

Unmasking the immune microecology of ductal carcinoma in situ with deep learning

Despite increasing evidence supporting the clinical relevance of tumour infiltrating lymphocytes (TILs) in invasive breast cancer, TIL spatial variability within ductal carcinoma in situ (DCIS) samples and its association with progression are not well understood. To characterise tissue spatial architecture and the microenvironment of DCIS, we designed and validated a new deep learning pipeline, UNMaSk. Following automated detection of individual DCIS ducts using a new method IM-Net, we applied spatial tessellation to create virtual boundaries for each duct. To study local TIL infiltration for each duct, DRDIN was developed for mapping the distribution of TILs. In a dataset comprising grade 2-3 pure DCIS and DCIS adjacent to invasive cancer (adjacent DCIS), we found that pure DCIS cases had more TILs compared to adjacent DCIS. However, the colocalisation of TILs with DCIS ducts was significantly lower in pure DCIS compared to adjacent DCIS, which may suggest a more inflamed tissue ecology local to DCIS ducts in adjacent DCIS cases. Our study demonstrates that technological developments in deep convolutional neural networks and digital pathology can enable an automated morphological and microenvironmental analysis of DCIS, providing a new way to study differential immune ecology for individual ducts and identify new markers of progression.

A tree-based multiclassification of breast tumor histopathology images through deep learning

Worldwide, the burden of cancer is drastically increasing over the past few years. Among all types of cancers in women, breast cancer (BrC) is the main cause of unnatural deaths. For early diagnosis, histopathology (Hp) imaging is a gold standard for positive and detailed (at tissue level) diagnosis of breast tumor (BrT) compared to mammogram images. A large number of studies used BrT Hp images to solve binary or multiclassification problems using high computational resources. However, classification models' performance may be compromised due to the high correlation among various types of BrT in Hp images, which raises the misclassification rate. Thus, this paper aims to develop a tree-based BrT multiclassification model via deep learning (DL) to extract discriminative features to solve the multiclassification problem with better performance using less computational resources. The main contributions of this work are to create an ensemble, tree-based DL model that is pre-trained on the BreakHis dataset, and implementation of a misclassification reduction algorithm. The ensemble, tree-based DL model, extracts discriminative BrT features from Hp images. The target dataset (i.e., Bioimaging challenge 2015 breast histology) is small in size; thus, to avoid overfitting of the proposed model, pretraining is performed on the BreakHis dataset. Whereas, misclassification reduction algorithm is implemented to enhance the performance of the classification model. The experimental results show that the proposed model outperformed the existing state-of-the-art baseline studies. The achieved classification accuracy is ranging from 87.50 % to 100 % for four subtypes of BrT. Thus, the proposed model can assist doctors as the second opinion in any healthcare centre.

High throughput assessment of biomarkers in tissue microarrays using artificial intelligence: PTEN loss as a proof-of-principle in multi-center prostate cancer cohorts

Phosphatase and tensin homolog (PTEN) loss is associated with adverse outcomes in prostate cancer and has clinical potential as a prognostic biomarker. The objective of this work was to develop an artificial intelligence (AI) system for automated detection and localization of PTEN loss on immunohistochemically (IHC) stained sections. PTEN loss was assessed using IHC in two prostate tissue microarrays (TMA) (internal cohort, n = 272 and external cohort, n = 129 patients). TMA cores were visually scored for PTEN loss by pathologists and, if present, spatially annotated. Cores from each patient within the internal TMA cohort were split into 90% cross-validation (N = 2048) and 10% hold-out testing (N = 224) sets. ResNet-101 architecture was used to train core-based classification using a multi-resolution ensemble approach (×5, ×10, and ×20). For spatial annotations, single resolution pixel-based classification was trained from patches extracted at ×20 resolution, interpolated to ×40 resolution, and applied in a sliding-window fashion. A final AI-based prediction model was created from combining multi-resolution and pixel-based models. Performance was evaluated in 428 cores of external cohort. From both cohorts, a total of 2700 cores were studied, with a frequency of PTEN loss of 14.5% in internal (180/1239) and external 13.5% (43/319) cancer cores. The final AI-based prediction of PTEN status demonstrated 98.1% accuracy (95.0% sensitivity, 98.4% specificity; median dice score = 0.811) in internal cohort cross-validation set and 99.1% accuracy (100% sensitivity, 99.0% specificity; median dice score = 0.804) in internal cohort test set. Overall core-based classification in the external cohort was significantly improved in the external cohort (area under the curve = 0.964, 90.6% sensitivity, 95.7% specificity) when further trained (fine-tuned) using 15% of cohort data (19/124 patients). These results demonstrate a robust and fully automated method for detection and localization of PTEN loss in prostate cancer tissue samples. AI-based algorithms have potential to streamline sample assessment in research and clinical laboratories.

Infiltrating immune cells in benign breast disease and risk of subsequent invasive breast cancer

Background

It is well established that tumors are antigenic and can induce an immune response by the host, entailing lymphocytic infiltration of the tumor and surrounding stroma. The extent and composition of the immune response to the tumor, assessed through evaluation of tumor-infiltrating lymphocyte counts, has been shown in many studies to have prognostic and predictive value for invasive breast cancer, but currently, there is little evidence regarding the association between infiltrating immune cell counts (IICCs) in women with benign breast disease (BBD) and risk of subsequent invasive breast cancer.

Methods

Using a cohort of 15,395 women biopsied for BBD at Kaiser Permanente Northwest, we conducted a nested case-control study in which cases were women who developed a subsequent invasive breast cancer during follow-up and controls were individually matched to cases on age at BBD diagnosis. We assessed IICCs in normal tissue and in the BBD lesions, and we used unconditional logistic regression to estimate the multivariable odds ratios (OR) and 95% confidence intervals (CI) for the associations between IICCs and breast cancer risk.

Results

There was no association between the IICC in normal tissue (multivariable OR per 5% increase in IICC = 1.05, 95% CI = 0.96–1.16) or in the BBD lesion (OR per 5% increase in IICC = 1.06, 95% CI = 0.96–1.18) and risk of subsequent invasive breast cancer. Also, there were no associations within subgroups defined by menopausal status, BBD histology, BMI, and history of smoking.

Conclusion

The results of this study suggest that IICCs in BBD tissue are not associated with altered risk of subsequent invasive breast cancer.

Deep learning-enabled medical computer vision

A decade of unprecedented progress in artificial intelligence (AI) has demonstrated the potential for many fields-including medicine-to benefit from the insights that AI techniques can extract from data. Here we survey recent progress in the development of modern computer vision techniques-powered by deep learning-for medical applications, focusing on medical imaging, medical video, and clinical deployment. We start by briefly summarizing a decade of progress in convolutional neural networks, including the vision tasks they enable, in the context of healthcare. Next, we discuss several example medical imaging applications that stand to benefit-including cardiology, pathology, dermatology, ophthalmology-and propose new avenues for continued work. We then expand into general medical video, highlighting ways in which clinical workflows can integrate computer vision to enhance care. Finally, we discuss the challenges and hurdles required for real-world clinical deployment of these technologies.

Style transfer strategy for developing a generalizable deep learning application in digital pathology

BACKGROUND AND OBJECTIVES

Despite recent advances in artificial intelligence for medical images, the development of a robust deep learning model for identifying malignancy on pathology slides has been limited by problems related to substantial inter- and intra-institutional heterogeneity attributable to tissue preparation. The paucity of available data aggravates this limitation for relatively rare cancers. Here, using ovarian cancer pathology images, we explored the effect of image-to-image style transfer approaches on diagnostic performance.

METHODS

We leveraged a relatively large public image set for 142 patients with ovarian cancer from The Cancer Image Archive (TCIA) to fine-tune the renowned deep learning model Inception V3 for identifying malignancy on tissue slides. As an external validation, the performance of the developed classifier was tested using a relatively small institutional pathology image set for 32 patients. To reduce deterioration of the performance associated with the inter-institutional heterogeneity of pathology slides, we translated the style of the small image set of the local institution into the large image set style of the TCIA using cycle-consistent generative adversarial networks.

RESULTS

Without style transfer, the performance of the classifier was as follows: area under the receiver operating characteristic curve (AUROC) = 0.737 and area under the precision recall curve (AUPRC) = 0.710. After style transfer, AUROC and AUPRC improved to 0.916 and 0.898, respectively.

CONCLUSIONS

This study provides a case of the successful application of style transfer technology to generalize a deep learning model into small image sets in the field of digital pathology. Researchers at local institutions can select this collaborative system to make their small image sets acceptable to the deep learning model.

Machine learning techniques for mitoses classification

BACKGROUND

Pathologists analyze biopsy material at both the cellular and structural level to determine diagnosis and cancer stage. Mitotic figures are surrogate biomarkers of cellular proliferation that can provide prognostic information; thus, their precise detection is an important factor for clinical care. Convolutional Neural Networks (CNNs) have shown remarkable performance on several recognition tasks. Utilizing CNNs for mitosis classification may aid pathologists to improve the detection accuracy.

METHODS

We studied two state-of-the-art CNN-based models, ESPNet and DenseNet, for mitosis classification on six whole slide images of skin biopsies and compared their quantitative performance in terms of sensitivity, specificity, and F-score. We used raw RGB images of mitosis and non-mitosis samples with their corresponding labels as training input. In order to compare with other work, we studied the performance of these classifiers and two other architectures, ResNet and ShuffleNet, on the publicly available MITOS breast biopsy dataset and compared the performance of all four in terms of precision, recall, and F-score (which are standard for this data set), architecture, training time and inference time.

RESULTS

The ESPNet and DenseNet results on our primary melanoma dataset had a sensitivity of 0.976 and 0.968, and a specificity of 0.987 and 0.995, respectively, with F-scores of .968 and .976, respectively. On the MITOS dataset, ESPNet and DenseNet showed a sensitivity of 0.866 and 0.916, and a specificity of 0.973 and 0.980, respectively. The MITOS results using DenseNet had a precision of 0.939, recall of 0.916, and F-score of 0.927. The best published result on MITOS (Saha et al. 2018) reported precision of 0.92, recall of 0.88, and F-score of 0.90. In our architecture comparisons on MITOS, we found that DenseNet beats the others in terms of F-Score (DenseNet 0.927, ESPNet 0.890, ResNet 0.865, ShuffleNet 0.847) and especially Recall (DenseNet 0.916, ESPNet 0.866, ResNet 0.807, ShuffleNet 0.753), while ResNet and ESPNet have much faster inference times (ResNet 6 s, ESPNet 8 s, DenseNet 31 s). ResNet is faster than ESPNet, but ESPNet has a higher F-Score and Recall than ResNet, making it a good compromise solution.

CONCLUSION

We studied several state-of-the-art CNNs for detecting mitotic figures in whole slide biopsy images. We evaluated two CNNs on a melanoma cancer dataset and then compared four CNNs on a public breast cancer data set, using the same methodology on both. Our methodology and architecture for mitosis finding in both melanoma and breast cancer whole slide images has been thoroughly tested and is likely to be useful for finding mitoses in any whole slide biopsy images.

Automatic Pancreatic Ductal Adenocarcinoma Detection in Whole Slide Images Using Deep Convolutional Neural Networks

Pancreatic ductal adenocarcinoma (PDAC) is one of the deadliest cancer types worldwide, with the lowest 5-year survival rate among all kinds of cancers. Histopathology image analysis is considered a gold standard for PDAC detection and diagnosis. However, the manual diagnosis used in current clinical practice is a tedious and time-consuming task and diagnosis concordance can be low. With the development of digital imaging and machine learning, several scholars have proposed PDAC analysis approaches based on feature extraction methods that rely on field knowledge. However, feature-based classification methods are applicable only to a specific problem and lack versatility, so that the deep-learning method is becoming a vital alternative to feature extraction. This paper proposes the first deep convolutional neural network architecture for classifying and segmenting pancreatic histopathological images on a relatively large WSI dataset. Our automatic patch-level approach achieved 95.3% classification accuracy and the WSI-level approach achieved 100%. Additionally, we visualized the classification and segmentation outcomes of histopathological images to determine which areas of an image are more important for PDAC identification. Experimental results demonstrate that our proposed model can effectively diagnose PDAC using histopathological images, which illustrates the potential of this practical application.

Classifying Breast Histopathology Images with a Ductal Instance-Oriented Pipeline

In this study, we propose the Ductal Instance-Oriented Pipeline (DIOP) that contains a duct-level instance segmentation model, a tissue-level semantic segmentation model, and three-levels of features for diagnostic classification. Based on recent advancements in instance segmentation and the Mask RCNN model, our duct-level segmenter tries to identify each ductal individual inside a microscopic image; then, it extracts tissue-level information from the identified ductal instances. Leveraging three levels of information obtained from these ductal instances and also the histopathology image, the proposed DIOP outperforms previous approaches (both feature-based and CNN-based) in all diagnostic tasks; for the four-way classification task, the DIOP achieves comparable performance to general pathologists in this unique dataset. The proposed DIOP only takes a few seconds to run in the inference time, which could be used interactively on most modern computers. More clinical explorations are needed to study the robustness and generalizability of this system in the future.

Scale-Aware Transformers for Diagnosing Melanocytic Lesions

Diagnosing melanocytic lesions is one of the most challenging areas of pathology with extensive intra- and inter-observer variability. The gold standard for a diagnosis of invasive melanoma is the examination of histopathological whole slide skin biopsy images by an experienced dermatopathologist. Digitized whole slide images offer novel opportunities for computer programs to improve the diagnostic performance of pathologists. In order to automatically classify such images, representations that reflect the content and context of the input images are needed. In this paper, we introduce a novel self-attention-based network to learn representations from digital whole slide images of melanocytic skin lesions at multiple scales. Our model softly weighs representations from multiple scales, allowing it to discriminate between diagnosis-relevant and -irrelevant information automatically. Our experiments show that our method outperforms five other state-of-the-art whole slide image classification methods by a significant margin. Our method also achieves comparable performance to 187 practicing U.S. pathologists who interpreted the same cases in an independent study. To facilitate relevant research, full training and inference code is made publicly available at https://github.com/meredith-wenjunwu/ScATNet.

Deep Learning in Toxicologic Pathology: A New Approach to Evaluate Rodent Retinal Atrophy

Quantification of retinal atrophy, caused by therapeutics and/or light, by manual measurement of retinal layers is labor intensive and time-consuming. In this study, we explored the role of deep learning (DL) in automating the assessment of retinal atrophy, particularly of the outer and inner nuclear layers, in rats. Herein, we report our experience creating and employing a hybrid approach, which combines conventional image processing and DL to quantify rodent retinal atrophy. Utilizing a DL approach based upon the VGG16 model architecture, models were trained, tested, and validated using 10,746 image patches scanned from whole slide images (WSIs) of hematoxylin-eosin stained rodent retina. The accuracy of this computational method was validated using pathologist annotated WSIs throughout and used to separately quantify the thickness of the outer and inner nuclear layers of the retina. Our results show that DL can facilitate the evaluation of therapeutic and/or light-induced atrophy, particularly of the outer retina, efficiently in rodents. In addition, this study provides a template which can be used to train, validate, and analyze the results of toxicologic pathology DL models across different animal species used in preclinical efficacy and safety studies.

Artificial Intelligence in Hematology: Current Challenges and Opportunities

PURPOSE OF REVIEW

Artificial intelligence (AI), and in particular its subcategory machine learning, is finding an increasing number of applications in medicine, driven in large part by an abundance of data and powerful, accessible tools that have made AI accessible to a larger circle of investigators.

RECENT FINDINGS

AI has been employed in the analysis of hematopathological, radiographic, laboratory, genomic, pharmacological, and chemical data to better inform diagnosis, prognosis, treatment planning, and foundational knowledge related to benign and malignant hematology. As more widespread implementation of clinical AI nears, attention has also turned to the effects this will have on other areas in medicine. AI offers many promising tools to clinicians broadly, and specifically in the practice of hematology. Ongoing research into its various applications will likely result in an increasing utilization of AI by a broader swath of clinicians.

A Semisupervised Learning Scheme with Self-Paced Learning for Classifying Breast Cancer Histopathological Images

The unavailability of large amounts of well-labeled data poses a significant challenge in many medical imaging tasks. Even in the likelihood of having access to sufficient data, the process of accurately labeling the data is an arduous and time-consuming one, requiring expertise skills. Again, the issue of unbalanced data further compounds the abovementioned problems and presents a considerable challenge for many machine learning algorithms. In lieu of this, the ability to develop algorithms that can exploit large amounts of unlabeled data together with a small amount of labeled data, while demonstrating robustness to data imbalance, can offer promising prospects in building highly efficient classifiers. This work proposes a semisupervised learning method that integrates self-training and self-paced learning to generate and select pseudolabeled samples for classifying breast cancer histopathological images. A novel pseudolabel generation and selection algorithm is introduced in the learning scheme to generate and select highly confident pseudolabeled samples from both well-represented classes to less-represented classes. Such a learning approach improves the performance by jointly learning a model and optimizing the generation of pseudolabels on unlabeled-target data to augment the training data and retraining the model with the generated labels. A class balancing framework that normalizes the class-wise confidence scores is also proposed to prevent the model from ignoring samples from less represented classes (hard-to-learn samples), hence effectively handling the issue of data imbalance. Extensive experimental evaluation of the proposed method on the BreakHis dataset demonstrates the effectiveness of the proposed method.

A clinically applicable deep-learning model for detecting intracranial aneurysm in computed tomography angiography images

Intracranial aneurysm is a common life-threatening disease. Computed tomography angiography is recommended as the standard diagnosis tool; yet, interpretation can be time-consuming and challenging. We present a specific deep-learning-based model trained on 1,177 digital subtraction angiography verified bone-removal computed tomography angiography cases. The model has good tolerance to image quality and is tested with different manufacturers. Simulated real-world studies are conducted in consecutive internal and external cohorts, in which it achieves an improved patient-level sensitivity and lesion-level sensitivity compared to that of radiologists and expert neurosurgeons. A specific cohort of suspected acute ischemic stroke is employed and it is found that 99.0% predicted-negative cases can be trusted with high confidence, leading to a potential reduction in human workload. A prospective study is warranted to determine whether the algorithm could improve patients' care in comparison to clinicians' assessment.

Breast cancer, screening and diagnostic tools: All you need to know

Breast cancer is one of the most frequent malignancies among women worldwide. Methods for screening and diagnosis allow health care professionals to provide personalized treatments that improve the outcome and survival. Scientists and physicians are working side-by-side to develop evidence-based guidelines and equipment to detect cancer earlier. However, the lack of comprehensive interdisciplinary information and understanding between biomedical, medical, and technology professionals makes innovation of new screening and diagnosis tools difficult. This critical review gathers, for the first time, information concerning normal breast and cancer biology, established and emerging methods for screening and diagnosis, staging and grading, molecular and genetic biomarkers. Our purpose is to address key interdisciplinary information about these methods for physicians and scientists. Only the multidisciplinary interaction and communication between scientists, health care professionals, technical experts and patients will lead to the development of better detection tools and methods for an improved screening and early diagnosis.

Variant Classification Concordance using the ACMG-AMP Variant Interpretation Guidelines across Nine Genomic Implementation Research Studies

Harmonization of variant pathogenicity classification across laboratories is important for advancing clinical genomics. The two CLIA-accredited Electronic Medical Record and Genomics Network sequencing centers and the six CLIA-accredited laboratories and one research laboratory performing genome or exome sequencing in the Clinical Sequencing Evidence-Generating Research Consortium collaborated to explore current sources of discordance in classification. Eight laboratories each submitted 20 classified variants in the ACMG secondary finding v.2.0 genes. After removing duplicates, each of the 158 variants was annotated and independently classified by two additional laboratories using the ACMG-AMP guidelines. Overall concordance across three laboratories was assessed and discordant variants were reviewed via teleconference and email. The submitted variant set included 28 P/LP variants, 96 VUS, and 34 LB/B variants, mostly in cancer (40%) and cardiac (27%) risk genes. Eighty-six (54%) variants reached complete five-category (i.e., P, LP, VUS, LB, B) concordance, and 17 (11%) had a discordance that could affect clinical recommendations (P/LP versus VUS/LB/B). 21% and 63% of variants submitted as P and LP, respectively, were discordant with VUS. Of the 54 originally discordant variants that underwent further review, 32 reached agreement, for a post-review concordance rate of 84% (118/140 variants). This project provides an updated estimate of variant concordance, identifies considerations for LP classified variants, and highlights ongoing sources of discordance. Continued and increased sharing of variant classifications and evidence across laboratories, and the ongoing work of ClinGen to provide general as well as gene- and disease-specific guidance, will lead to continued increases in concordance.

Interobserver Variability in Ductal Carcinoma In Situ of the Breast

OBJECTIVES

Since most patients with ductal carcinoma in situ (DCIS) of the breast are treated upon diagnosis, evidence on its natural progression to invasive carcinoma is limited. It is estimated that around half of the screen-detected DCIS lesions would have remained indolent if they had never been detected. Many patients with DCIS are therefore probably overtreated. Four ongoing randomized noninferiority trials explore active surveillance as a treatment option. Eligibility for these trials is mainly based on histopathologic features. Hence, the call for reproducible histopathologic assessment has never sounded louder.

METHODS

Here, the available classification systems for DCIS are discussed in depth.

RESULTS

This comprehensive review illustrates that histopathologic evaluation of DCIS is characterized by significant interobserver variability. Future digitalization of pathology, combined with development of deep learning algorithms or so-called artificial intelligence, may be an innovative solution to tackle this problem. However, implementation of digital pathology is not within reach for each laboratory worldwide. An alternative classification system could reduce the disagreement among histopathologists who use "conventional" light microscopy: the introduction of dichotomous histopathologic assessment is likely to increase interobserver concordance.

CONCLUSIONS

Reproducible histopathologic assessment is a prerequisite for robust risk stratification and adequate clinical decision-making. Two-tier histopathologic assessment might enhance the quality of care.

A Novel Pathological Images and Genomic Data Fusion Framework for Breast Cancer Survival Prediction

Survival analysis is a valid solution for cancer treatments and outcome evaluations. Due to the wide application of medical imaging and genome technology, computer-aided survival analysis has become a popular and promising area, from which we can get relatively satisfactory results. Although there are already some impressive technologies in this field, most of them make some recommendations using single-source medical data and have not combined multi-level and multi-source data efficiently. In this paper, we propose a novel pathological images and gene expression data fusion framework to perform the survival prediction. Different from previous methods, our framework can extract correlated multi-scale deep features from whole slide images (WSIs) and dimensionality reduced gene expression data respectively for jointly survival analysis. The experiment results demonstrate that the integrated multi-level image and genome features can achieve higher prediction accuracy compared with single-source features.

Artificial Intelligence-Based Multiclass Classification of Benign or Malignant Mucosal Lesions of the Stomach

Gastric cancer (GC) is one of the leading causes of cancer-related death worldwide. It takes some time from chronic gastritis to develop in GC. Early detection of GC will help patients obtain timely treatment. Understanding disease evolution is crucial for the prevention and treatment of GC. Here, we present a convolutional neural network (CNN)-based system to detect abnormalities in the gastric mucosa. We identified normal mucosa, chronic gastritis, and intestinal-type GC: this is the most common route of gastric carcinogenesis. We integrated digitalizing histopathology of whole-slide images (WSIs), stain normalization, a deep CNN, and a random forest classifier. The staining variability of WSIs was reduced significantly through stain normalization, and saved the cost and time of preparing new slides. Stain normalization improved the effect of the CNN model. The accuracy rate at the patch-level reached 98.4%, and 94.5% for discriminating normal → chronic gastritis → GC. The accuracy rate at the WSIs-level for discriminating normal tissue and cancerous tissue reached 96.0%, which is a state-of-the-art result. Survival analyses indicated that the features extracted from the CNN exerted a significant impact on predicting the survival of cancer patients. Our CNN model disclosed significant potential for adjuvant diagnosis of gastric diseases, especially GC, and usefulness for predicting the prognosis.

Performance indicators in breast cancer screening in the European Union: A comparison across countries of screen positivity and detection rates

Comparable performance indicators for breast cancer screening in the European Union (EU) have not been previously reported. We estimated adjusted breast cancer screening positivity rate (PR) and detection rates (DR) to investigate variation across EU countries. For the age 50-69 years, the adjusted EU-pooled PR for initial screening was 8.9% (cross-programme variation range 3.2-19.5%) while DR of invasive cancers was 5.3/1,000 (range 3.8-7.4/1,000) and DR of ductal carcinoma in situ (DCIS) was 1.3/1,000 (range 0.7-2.7/1,000). For subsequent screening, the adjusted EU-pooled PR was 3.6% (range 1.4-8.4%), the DR was 4.0/1,000 (range 2.2-5.8/1,000) and 0.8/1,000 (range 0.5-1.3/1,000) for invasive and DCIS, respectively. Adjusted performance indicators showed remarkable heterogeneity, likely due to different background breast cancer risk and awareness between target populations, and also different screening protocols and organisation. Periodic reporting of the screening indicators permits comparison and evaluation of the screening activities between and within countries aiming to improve the quality and the outcomes of screening programmes. Cancer Screening Registries would be a milestone in this direction and EU Screening Reports provide a fundamental contribution to building them.

Prognostic value of histopathological DCIS features in a large-scale international interrater reliability study

PURPOSE

For optimal management of ductal carcinoma in situ (DCIS), reproducible histopathological assessment is essential to distinguish low-risk from high-risk DCIS. Therefore, we analyzed interrater reliability of histopathological DCIS features and assessed their associations with subsequent ipsilateral invasive breast cancer (iIBC) risk.

METHODS

Using a case-cohort design, reliability was assessed in a population-based, nationwide cohort of 2767 women with screen-detected DCIS diagnosed between 1993 and 2004, treated by breast-conserving surgery with/without radiotherapy (BCS ± RT) using Krippendorff's alpha (KA) and Gwet's AC2 (GAC2). Thirty-eight raters scored histopathological DCIS features including grade (2-tiered and 3-tiered), growth pattern, mitotic activity, periductal fibrosis, and lymphocytic infiltrate in 342 women. Using majority opinion-based scores for each feature, their association with subsequent iIBC risk was assessed using Cox regression.

RESULTS

Interrater reliability of grade using various classifications was fair to moderate, and only substantial for grade 1 versus 2 + 3 when using GAC2 (0.78). Reliability for growth pattern (KA 0.44, GAC2 0.78), calcifications (KA 0.49, GAC2 0.70) and necrosis (KA 0.47, GAC2 0.70) was moderate using KA and substantial using GAC2; for (type of) periductal fibrosis and lymphocytic infiltrate fair to moderate estimates were found and for mitotic activity reliability was substantial using GAC2 (0.70). Only in patients treated with BCS-RT, high mitotic activity was associated with a higher iIBC risk in univariable analysis (Hazard Ratio (HR) 2.53, 95% Confidence Interval (95% CI) 1.05-6.11); grade 3 versus 1 + 2 (HR 2.64, 95% CI 1.35-5.14) and a cribriform/solid versus flat epithelial atypia/clinging/(micro)papillary growth pattern (HR 3.70, 95% CI 1.34-10.23) were independently associated with a higher iIBC risk.

CONCLUSIONS

Using majority opinion-based scores, DCIS grade, growth pattern, and mitotic activity are associated with iIBC risk in patients treated with BCS-RT, but interrater variability is substantial. Semi-quantitative grading, incorporating and separately evaluating nuclear pleomorphism, growth pattern, and mitotic activity, may improve the reliability and prognostic value of these features.

Prospective Multicenter Study Validate a Prediction Model for Surgery Uptake Among Women with Atypical Breast Lesions

BACKGROUND

Diagnosis of atypical breast lesions (ABLs) leads to unnecessary surgery in 75-90% of women. We have previously developed a model including age, complete radiological target excision after biopsy, and focus size that predicts the probability of cancer at surgery. The present study aimed to validate this model in a prospective multicenter setting.

- METHODS

Women with a recently diagnosed ABL on image-guided biopsy were recruited in 18 centers, before wire-guided localized excisional lumpectomy. Primary outcome was the negative predictive value (NPV) of the model.

RESULTS

The NOMAT model could be used in 287 of the 300 patients included (195 with ADH). At surgery, 12 invasive (all grade 1), and 43 in situ carcinomas were identified (all ABL: 55/287, 19%; ADH only: 49/195, 25%). The area under the receiving operating characteristics curve of the model was 0.64 (95% CI 0.58-0.69) for all ABL, and 0.63 for ADH only (95% CI 0.56-0.70). For the pre-specified threshold of 20% predicted probability of cancer, NPV was 82% (77-87%) for all ABL, and 77% (95% CI 71-83%) for patients with ADH. At a 10% threshold, NPV was 89% (84-94%) for all ABL, and 85% (95% CI 78--92%) for the ADH. At this threshold, 58% of the whole ABL population (and 54% of ADH patients) could have avoided surgery with only 2 missed invasive cancers.

CONCLUSION

The NOMAT model could be useful to avoid unnecessary surgery among women with ABL, including for patients with ADH.

CLINICAL TRIAL REGISTRATION

NCT02523612.

The Clinical Utility of a Negative Result at Molecular Breast Imaging: Initial Proof of Concept

Purpose

To calculate the negative predictive value (NPV) and false-negative rate (FNR) of molecular breast imaging (MBI) performed in patients who had low-suspicion index findings on mammograms and US images.

Materials and Methods

This retrospective study included patients who had undergone MBI between January 2015 and July 2017, who had index findings on screening mammograms and/or US images, and for whom either histopathologic results or a minimum of 1-year imaging follow-up results were available. A drawn dose of 8 mCi (296 MBq) of technetium 99m sestamibi was administered to all patients for MBI. The NPV and FNR of MBI was calculated for the cohort of 381 findings among 338 women (median age, 56 years; age range, 28-89 years) included in this study.

Results

Overall, 292 of the 381 (76.6%) MBI results were interpreted as negative. Of the 292, 27 patients underwent subsequent biopsies, results of which were negative for cancer; one patient underwent biopsy, and the result was positive for cancer; and 264 patients had true-negative findings based on follow-up imaging for a minimum of 1 year. Of the 89 MBI acquisitions interpreted as positive, there were 36 cancers. The NPV was calculated to be 99.7% (291 of 292, 95% confidence interval [CI]: 99.1%, 100%), and the FNR was 2.7% (one of 37, 95% CI: 0%, 7.9%). Interposing MBI reduced the number of biopsies by 67.5%.

Conclusion

The concept of the clinical utility of a negative MBI result may be valid but requires further testing.Keywords: Breast, Molecular Imaging-Cancer© RSNA, 2020.

Upgrade Rate of Flat Epithelial Atypia Diagnosed at Stereotactic Core Needle Biopsy of Microcalcifications: Is Excisional Biopsy Indicated?

OBJECTIVE

Whether the optimal management of pure flat epithelial atypia (FEA) found on core needle biopsy (CNB) specimens is surgical excision or imaging follow-up remains controversial. This study aimed to determine the upgrade rate to ductal carcinoma in situ (DCIS), invasive carcinoma or a high-risk lesion (atypical ductal hyperplasia, atypical lobular hyperplasia, or lobular carcinoma in situ), and it explored the relationship between a family history of breast cancer and the risk of upgrade.

METHODS

Cases with pure FEA found on stereotactic CNB of microcalcifications between March 2011 to December 2017 were followed by excisional biopsy or periodic imaging. The proportion of cases upgraded to a high-risk lesion and the odds of upgrade as related to a family history of breast cancer were determined with 95% confidence intervals (CIs).

RESULTS

We identified 622 cases of pure FEA; 101 (16.2%) underwent surgical excision and 269 (43.2%) had imaging follow-up of ≥ 24 months. There were no upgrades to DCIS or invasive cancer in any of these 370 individuals (0%), and 4.6% (17/370; 95% CI: 2.9%-7.2%) were upgraded to a high-risk lesion. There was a nonstatistically significant trend between family history and upgrade to high-risk lesion (odds ratio 1.72 [95% CI: 0.65%-4.57%]).

CONCLUSION

In our study, the upgrade rate of pure FEA to malignancy was 0%. We suggest that regular imaging follow-up is an appropriate alternative to surgery. Because of potential differences in biopsy techniques and pathologist interpretation of the primary biopsy, individual institutions should audit their own results prior to altering their management of FEA.

Breast Cancer Histopathology Image Classification Using an Ensemble of Deep Learning Models

Breast cancer is one of the major public health issues and is considered a leading cause of cancer-related deaths among women worldwide. Its early diagnosis can effectively help in increasing the chances of survival rate. To this end, biopsy is usually followed as a gold standard approach in which tissues are collected for microscopic analysis. However, the histopathological analysis of breast cancer is non-trivial, labor-intensive, and may lead to a high degree of disagreement among pathologists. Therefore, an automatic diagnostic system could assist pathologists to improve the effectiveness of diagnostic processes. This paper presents an ensemble deep learning approach for the definite classification of non-carcinoma and carcinoma breast cancer histopathology images using our collected dataset. We trained four different models based on pre-trained VGG16 and VGG19 architectures. Initially, we followed 5-fold cross-validation operations on all the individual models, namely, fully-trained VGG16, fine-tuned VGG16, fully-trained VGG19, and fine-tuned VGG19 models. Then, we followed an ensemble strategy by taking the average of predicted probabilities and found that the ensemble of fine-tuned VGG16 and fine-tuned VGG19 performed competitive classification performance, especially on the carcinoma class. The ensemble of fine-tuned VGG16 and VGG19 models offered sensitivity of 97.73% for carcinoma class and overall accuracy of 95.29%. Also, it offered an F1 score of 95.29%. These experimental results demonstrated that our proposed deep learning approach is effective for the automatic classification of complex-natured histopathology images of breast cancer, more specifically for carcinoma images.

Ipsilateral and Concurrent Breast Cancer and Atypical Ductal Hyperplasia: Does Atypia Also Need Surgical Excision?

BACKGROUND

Standard-of-care management of atypical ductal hyperplasia (ADH) is surgical excision. Multiple studies have identified features of ADH in patients at low risk for upgrade who may benefit from omission of surgical excision. Patients with an ipsilateral breast cancer have been excluded from studies investigating observation for the management of ADH.

METHODS

This was a retrospective review of women with both a breast cancer and an ipsilateral separate site of ADH diagnosed on percutaneous biopsy, who underwent excision of both sites from 2008 to 2018. Radiographic and pathologic features of ADH and cancer were analyzed, including imaging size, biopsy modality, distance between sites, cancer subtype, grade, prognostic markers, ADH foci, and presence of necrosis or micropapillary features. Final pathology at the ADH site was used to determine upgrade. Multivariable logistic regression was performed to identify variables significantly associated with ADH upgrade to malignancy.

RESULTS

Among 62 women meeting the inclusion criteria, 11 (17.7%) upgraded to malignancy [9 ductal carcinoma in situ (DCIS), 2 invasive cancer] at the site of ADH. Upgrade was significantly higher with ipsilateral DCIS (p = 0.03), ultrasound biopsy at the ADH site (p = 0.01), and ADH with necrosis (p = 0.04). The group at lowest risk for upgrade had stereotactic biopsy and ADH without necrosis (0% upgrade).

CONCLUSION

The presence of breast cancer does not significantly increase the likelihood for upgrade at a separate site of ipsilateral concurrent ADH above contemporary reported upgrade rates of ADH alone (10-30%). When considering breast conservation for breast cancer, omitting excision of the site of ADH can be considered when low-risk features are present.

Development and Validation of a Deep-learning Model to Assist With Renal Cell Carcinoma Histopathologic Interpretation

OBJECTIVE

To develop and test the ability of a convolutional neural network (CNN) to accurately identify the presence of renal cell carcinoma (RCC) on histopathology specimens, as well as differentiate RCC histologic subtype and grade.

MATERIALS AND METHODS

Digital hematoxylin and eosin stained biopsy images were downloaded from The Cancer Genome Atlas. A CNN model was trained on 100 um2 samples of either normal (3000 samples) or RCC (12,168 samples) tissue samples from 42 patients. RCC specimens included clear cell, chromophobe, and papillary histiotypes, as well as tissue of Fuhrman grades 1 through 4. Model testing was performed on an additional held-out cohort of benign and RCC specimens. Model performance was assessed on the basis of diagnostic accuracy, sensitivity, specificity, positive predictive value, and negative predictive value.

RESULTS

The CNN model achieved an overall accuracy of 99.1% in the testing cohort for distinguishing normal parenchyma from RCC (sensitivity 100%, specificity 97.1%). Accuracy for distinguishing between clear cell, papillary, and chromophobehistiotypes was 97.5%. Accuracy for predicting Fuhrman grade was 98.4%.

CONCLUSION

CNNs are able to rapidly and accurately identify the presence of RCC, distinguish RCC histologic subtypes, and identify tumor grade by analyzing histopathology specimens.

The Psychological Experience of Awaiting Breast Diagnosis

BACKGROUND

Each year, over 1 million women in the USA undergo diagnostic breast biopsies, many of which culminate in a benign outcome. However, for many patients, the experience of awaiting biopsy results is far from benign, instead provoking high levels of distress.

PURPOSE

To take a multifaceted approach to understanding the psychological experience of patients undergoing a breast biopsy.

METHOD

Female patients (N = 214) were interviewed at an appointment for a breast biopsy, just prior to undergoing the biopsy procedure. Pertinent to the current investigation, the interview assessed various patient characteristics, subjective health and cancer history, support availability, outcome expectations, distress, and coping strategies.

RESULTS

The findings revealed a complex set of interrelationships among patient characteristics, markers of distress, and use of coping strategies. Patients who were more distressed engaged in more avoidant coping strategies. Regarding the correlates of distress and coping, subjective health was more strongly associated with distress and coping than was cancer history; perceptions of support availability were also reliably associated with distress.

CONCLUSION

Taken together, the results suggest that patients focus on their immediate experience (e.g., subjective health, feelings of risk, perceptions of support) in the face of the acute moment of uncertainty prompted by a biopsy procedure, relative to more distal considerations such as cancer history and demographic characteristics. These findings can guide clinicians' interactions with patients at the biopsy appointment and can serve as a foundation for interventions designed to reduce distress in this context.

[Histological lesions of risk of breast carcinoma. Survival guide for the general pathologist]

Proliferative epithelial lesions are risk factors for breast cancer. They are a heterogeneous group of lesions in which the presence of atypia is related to varying degrees of risk. They should be considered in the differential diagnosis with benign lesions, in situ ductal carcinoma and infiltrating carcinoma. An accurate histopathological diagnosis is important in choosing the best therapeutic option, including vacuum assisted biopsy and surgery. We revise diagnostic criteria and the differential diagnosis of usual ductal hyperplasia, radial scar and complex sclerosing lesions, distinct types of adenosis, papillary lesions, atypical ductal hyperplasia, flat epithelial atypia and lobular neoplasia in situ. Furthermore, we summarize the degree of risk associated with the different conditions and management possibilities.

Explainable AI (xAI) for Anatomic Pathology

Pathologists are adopting whole slide images (WSIs) for diagnosis, thanks to recent FDA approval of WSI systems as class II medical devices. In response to new market forces and recent technology advances outside of pathology, a new field of computational pathology has emerged that applies artificial intelligence (AI) and machine learning algorithms to WSIs. Computational pathology has great potential for augmenting pathologists' accuracy and efficiency, but there are important concerns regarding trust of AI due to the opaque, black-box nature of most AI algorithms. In addition, there is a lack of consensus on how pathologists should incorporate computational pathology systems into their workflow. To address these concerns, building computational pathology systems with explainable AI (xAI) mechanisms is a powerful and transparent alternative to black-box AI models. xAI can reveal underlying causes for its decisions; this is intended to promote safety and reliability of AI for critical tasks such as pathology diagnosis. This article outlines xAI enabled applications in anatomic pathology workflow that improves efficiency and accuracy of the practice. In addition, we describe HistoMapr-Breast, an initial xAI enabled software application for breast core biopsies. HistoMapr-Breast automatically previews breast core WSIs and recognizes the regions of interest to rapidly present the key diagnostic areas in an interactive and explainable manner. We anticipate xAI will ultimately serve pathologists as an interactive computational guide for computer-assisted primary diagnosis.

Distinguishing metastatic triple-negative breast cancer from nonmetastatic breast cancer using second harmonic generation imaging and resonance Raman spectroscopy

Triple-negative breast cancer (TNBC) is an aggressive subset of breast cancer that is more common in African-American and Hispanic women. Early detection followed by intensive treatment is critical to improving poor survival rates. The current standard to diagnose TNBC from histopathology of biopsy samples is invasive and time-consuming. Imaging methods such as mammography and magnetic resonance (MR) imaging, while covering the entire breast, lack the spatial resolution and specificity to capture the molecular features that identify TNBC. Two nonlinear optical modalities of second harmonic generation (SHG) imaging of collagen, and resonance Raman spectroscopy (RRS) potentially offer novel rapid, label-free detection of molecular and morphological features that characterize cancerous breast tissue at subcellular resolution. In this study, we first applied MR methods to measure the whole-tumor characteristics of metastatic TNBC (4T1) and nonmetastatic estrogen receptor positive breast cancer (67NR) models, including tumor lactate concentration and vascularity. Subsequently, we employed for the first time in vivo SHG imaging of collagen and ex vivo RRS of biomolecules to detect different microenvironmental features of these two tumor models. We achieved high sensitivity and accuracy for discrimination between these two cancer types by quantitative morphometric analysis and nonnegative matrix factorization along with support vector machine. Our study proposes a new method to combine SHG and RRS together as a promising novel photonic and optical method for early detection of TNBC.

A New Deep Learning Model Selection Method for Colorectal Cancer Classification

Deep learning is one of the most commonly used techniques in computer-aided diagnosis systems. Their exploitation for histopathological image analysis is important because of the complex morphology of whole slide images. However, the main limitation of these methods is the restricted number of available medical images, which can lead to an overfitting problem. Many studies have suggested the use of static ensemble learning methods to address this issue. This article aims to propose a new dynamic ensemble deep learning method. First, it generates a set of models based on the transfer learning strategy from deep neural networks. Then, the relevant subset of models is selected by the particle swarm optimization algorithm and combined by voting or averaging methods. The proposed approach was tested on a histopathological dataset for colorectal cancer classification, based on seven types of CNNs. The method has achieved accurate results (94.52%) by the Resnet121 model and the voting strategy, which provides important insights into the efficiency of dynamic ensembling in deep learning.

Deployment of Artificial Intelligence in Real-World Practice: Opportunity and Challenge

Artificial intelligence has rapidly evolved from the experimental phase to the implementation phase in many image-driven clinical disciplines, including ophthalmology. A combination of the increasing availability of large datasets and computing power with revolutionary progress in deep learning has created unprecedented opportunities for major breakthrough improvements in the performance and accuracy of automated diagnoses that primarily focus on image recognition and feature detection. Such an automated disease classification would significantly improve the accessibility, efficiency, and cost-effectiveness of eye care systems where it is less dependent on human input, potentially enabling diagnosis to be cheaper, quicker, and more consistent. Although this technology will have a profound impact on clinical flow and practice patterns sooner or later, translating such a technology into clinical practice is challenging and requires similar levels of accountability and effectiveness as any new medication or medical device due to the potential problems of bias, and ethical, medical, and legal issues that might arise. The objective of this review is to summarize the opportunities and challenges of this transition and to facilitate the integration of artificial intelligence (AI) into routine clinical practice based on our best understanding and experience in this area.

Cellular resolution in clinical MALDI mass spectrometry imaging: the latest advancements and current challenges

Mass spectrometry (MS) is the workhorse of metabolomics, proteomics and lipidomics. Mass spectrometry imaging (MSI), its extension to spatially resolved analysis of tissues, is a powerful tool for visualizing molecular information within the histological context of tissue. This review summarizes recent developments in MSI and highlights current challenges that remain to achieve molecular imaging at the cellular level of clinical specimens. We focus on matrix-assisted laser desorption/ionization (MALDI)-MSI. We discuss the current status of each of the analysis steps and remaining challenges to reach the desired level of cellular imaging. Currently, analyte delocalization and degradation, matrix crystal size, laser focus restrictions and detector sensitivity are factors that are limiting spatial resolution. New sample preparation devices and laser optic systems are being developed to push the boundaries of these limitations. Furthermore, we review the processing of cellular MSI data and images, and the systematic integration of these data in the light of available algorithms and databases. We discuss roadblocks in the data analysis pipeline and show how technology from other fields can be used to overcome these. Finally, we conclude with curative and community efforts that are needed to enable contextualization of the information obtained.

Measuring intrarater association between correlated ordinal ratings

Variability between raters' ordinal scores is commonly observed in imaging tests, leading to uncertainty in the diagnostic process. In breast cancer screening, a radiologist visually interprets mammograms and MRIs, while skin diseases, Alzheimer's disease, and psychiatric conditions are graded based on clinical judgment. Consequently, studies are often conducted in clinical settings to investigate whether a new training tool can improve the interpretive performance of raters. In such studies, a large group of experts each classify a set of patients' test results on two separate occasions, before and after some form of training with the goal of assessing the impact of training on experts' paired ratings. However, due to the correlated nature of the ordinal ratings, few statistical approaches are available to measure association between raters' paired scores. Existing measures are restricted to assessing association at just one time point for a single screening test. We propose here a novel paired kappa to provide a summary measure of association between many raters' paired ordinal assessments of patients' test results before versus after rater training. Intrarater association also provides valuable insight into the consistency of ratings when raters view a patient's test results on two occasions with no intervention undertaken between viewings. In contrast to existing correlated measures, the proposed kappa is a measure that provides an overall evaluation of the association among multiple raters' scores from two time points and is robust to the underlying disease prevalence. We implement our proposed approach in two recent breast-imaging studies and conduct extensive simulation studies to evaluate properties and performance of our summary measure of association.

Intracystic papillary carcinoma: clinical presentation, patterns of practice, and oncological outcomes

BACKGROUND

Intracystic/encapsulated papillary carcinoma remains a poorly understood disease of the breast with a little amount of reports that describe it. It shares features with DCIS and IDC and predominantly affects postmenopausal women. This study aims to evaluate the clinical presentation, treatment, and outcomes in IPC patients managed at our institution.

METHODS

We retrospectively pooled twenty-eight IPC patients' medical records at our institution. Descriptive analysis of clinicopathological characteristics, approach, and outcomes was done along with a quantitative statistical analysis.

RESULTS

Cases were divided into three groups: isolated IPC, IPC associated with DCIS, and IPC associated with Invasive Carcinoma. Treatment modalities varied according to the IPC type and its associated components. All patients presented with a palpable mass. Immunohistochemical staining revealed that all isolated IPCs were ER and PR positive and HER2 negative. Lymph node dissection proved necessary only in IPC associated invasive carcinoma. Irregular borders and lobulations, among others, were found on non-invasive core biopsies that turned out to be associated with invasion on surgical pathology. All patients were alive after a median follow-up time of 23 months when the study was over with no reports of recurrence.

CONCLUSION

IPC cases and treatment approaches at our institution appear similar to the available literature and confirm the excellent prognosis among IPC. Even more, further studies into the key features such as BMI, family history, and radiological findings are necessary for a potential algorithm that could assess for risk of finding invasion in surgical pathology and subsequently the need for axillary/sentinel lymph node biopsy.