Interobserver variability in breast carcinoma grading results in prognostic stage differences

Validation of prostate and breast cancer detection artificial intelligence algorithms for accurate histopathological diagnosis and grading: a retrospective study with a Japanese cohort

Prostate and breast cancer incidence rates have been on the rise in Japan, emphasising the need for precise histopathological diagnosis to determine patient prognosis and guide treatment decisions. However, existing diagnostic methods face numerous challenges and are susceptible to inconsistencies between observers. To tackle these issues, artificial intelligence (AI) algorithms have been developed to aid in the diagnosis of prostate and breast cancer. This study focuses on validating the performance of two such algorithms, Galen Prostate and Galen Breast, in a Japanese cohort, with a particular focus on the grading accuracy and the ability to differentiate between invasive and non-invasive tumours. The research entailed a retrospective examination of 100 consecutive prostate and 100 consecutive breast biopsy cases obtained from a Japanese institution. Our findings demonstrated that the AI algorithms showed accurate cancer detection, with AUCs of 0.969 and 0.997 for the Galen Prostate and Galen Breast, respectively. The Galen Prostate was able to detect a higher Gleason score in four adenocarcinoma cases and detect a previously unreported cancer. The two algorithms successfully identified relevant pathological features, such as perineural invasions and lymphovascular invasions. Although further improvements are required to accurately differentiate rare cancer subtypes, these findings highlight the potential of these algorithms to enhance the precision and efficiency of prostate and breast cancer diagnosis in Japan. Furthermore, this validation paves the way for broader adoption of these algorithms as decision support tools within the Asian population.

DIEANet: an attention model for histopathological image grading of lung adenocarcinoma based on dimensional information embedding

Efficient and rapid auxiliary diagnosis of different grades of lung adenocarcinoma is conducive to helping doctors accelerate individualized diagnosis and treatment processes, thus improving patient prognosis. Currently, there is often a problem of large intra-class differences and small inter-class differences between pathological images of lung adenocarcinoma tissues under different grades. If attention mechanisms such as Coordinate Attention (CA) are directly used for lung adenocarcinoma grading tasks, it is prone to excessive compression of feature information and overlooking the issue of information dependency within the same dimension. Therefore, we propose a Dimension Information Embedding Attention Network (DIEANet) for the task of lung adenocarcinoma grading. Specifically, we combine different pooling methods to automatically select local regions of key growth patterns such as lung adenocarcinoma cells, enhancing the model's focus on local information. Additionally, we employ an interactive fusion approach to concentrate feature information within the same dimension and across dimensions, thereby improving model performance. Extensive experiments have shown that under the condition of maintaining equal computational expenses, the accuracy of DIEANet with ResNet34 as the backbone reaches 88.19%, with an AUC of 96.61%, MCC of 81.71%, and Kappa of 81.16%. Compared to seven other attention mechanisms, it achieves state-of-the-art objective metrics. Additionally, it aligns more closely with the visual attention of pathology experts under subjective visual assessment.

Changes in breast cancer grade from biopsy to excision following surgery or primary chemotherapy

Objective

To compare histological grade (G) of breast cancer and its components (scores for tubule formation - T, nuclear pleomorphism - P and mitotic counts - M) in core needle biopsies (CNBs) and surgical excision specimens (EXC) in patients treated with primary surgery (CHIR) or primary chemotherapy (PST).

Methods

Grade of matched pairs of carcinomas in CNB and EXC was assessed according to the Nottingham grading system.

Results

PST cases tended to have higher pretreatment G. Concordance rates in the CHIR (n = 760) and PST (n = 148) groups for T, P, M and G were 79%, 70%, 75%, 71% and 77%, 70%, 50%, 62%, respectively; differences in concordance rates were significant in M (p < 0.0001) and G (p = 0.024). For discordant cases in the CHIR group, CNBs tended to overestimate T and underestimate P, M and G, whereas in the PST group, the same trends were identified for T and P, but there was a significant tendency for M and G to be lower in EXC specimens.

Conclusions

The reversal of M and G underestimation in CNB to "overestimation" in the PST group can only be explained with the effect of mitosis reduction following chemotherapy. Whether the posttreatment decrease in G reflects any prognostic value remains to be elucidated.

PET/CT-based deep learning grading signature to optimize surgical decisions for clinical stage I invasive lung adenocarcinoma and biologic basis under its prediction: a multicenter study

PURPOSE

No consensus on a grading system for invasive lung adenocarcinoma had been built over a long period of time. Until October 2020, a novel grading system was proposed to quantify the whole landscape of histologic subtypes and proportions of pulmonary adenocarcinomas. This study aims to develop a deep learning grading signature (DLGS) based on positron emission tomography/computed tomography (PET/CT) to personalize surgical treatments for clinical stage I invasive lung adenocarcinoma and explore the biologic basis under its prediction.

METHODS

A total of 2638 patients with clinical stage I invasive lung adenocarcinoma from 4 medical centers were retrospectively included to construct and validate the DLGS. The predictive performance of the DLGS was evaluated by the area under the receiver operating characteristic curve (AUC), its potential to optimize surgical treatments was investigated via survival analyses in risk groups defined by the DLGS, and its biological basis was explored by comparing histologic patterns, genotypic alternations, genetic pathways, and infiltration of immune cells in microenvironments between risk groups.

RESULTS

The DLGS to predict grade 3 achieved AUCs of 0.862, 0.844, and 0.851 in the validation set (n = 497), external cohort (n = 382), and prospective cohort (n = 600), respectively, which were significantly better than 0.814, 0.810, and 0.806 of the PET model, 0.813, 0.795, and 0.824 of the CT model, and 0.762, 0.734, and 0.751 of the clinical model. Additionally, for DLGS-defined high-risk population, lobectomy yielded an improved prognosis compared to sublobectomy p = 0.085 for overall survival [OS] and p = 0.038 for recurrence-free survival [RFS]) and systematic nodal dissection conferred a superior prognosis to limited nodal dissection (p = 0.001 for OS and p = 0.041 for RFS).

CONCLUSION

The DLGS harbors the potential to predict the histologic grade and personalize the surgical treatments for clinical stage I invasive lung adenocarcinoma. Its applicability to other territories should be further validated by a larger international study.

Agreement on risk assessment and chemotherapy recommendations among breast cancer specialists: A survey within the MINDACT cohort

PURPOSE

Tailored recommendation for adjuvant chemotherapy in breast cancer patients is of great importance. This survey assessed agreement among oncologists on risk assessment and chemotherapy recommendation, the impact of adding the 70-gene signature to clinical-pathological characteristics, and changes over time.

METHODS

A survey consisting of 37 discordant patient cases from the MINDACT trial (T1-3N0-1M0) was sent to European breast cancer specialists for assessment of risk (high or low) and chemotherapy administration (yes or no). In 2015 the survey was sent twice (survey 1 and 2), several weeks apart, and in 2021 a third time (survey 3). Only the second and third surveys included the 70-gene signature result.

RESULTS

41 breast cancer specialists participated in all three surveys. Overall agreement between respondents decreased slightly between survey 1 and 2, but increased again in survey 3. Over time there was an increase in agreement with the 70-gene signature result on risk assessment, 23% in survey 2 versus 1 and 11% in survey 3 versus 2. With information available indicating a low risk 70-gene signature (n = 25 cases), 20% of risk assessments changed from high to low and 19% of recommendations changed from yes to no chemotherapy in survey 2 versus 1, further increasing with 18% and 21%, respectively, in survey 3 versus 2.

CONCLUSION

There is a variability in risk assessment of early breast cancer patients among breast cancer specialists. The 70-gene signature provided valuable information, resulting in fewer patients being assessed as high risk and fewer recommendations for chemotherapy, increasing over time.

A CT-Based Radiomics Nomogram Combined with Clinic-Radiological Characteristics for Preoperative Prediction of the Novel IASLC Grading of Invasive Pulmonary Adenocarcinoma

RATIONALE AND OBJECTIVES

The novel International Association for the Study of Lung Cancer (IASLC) grading system of invasive lung adenocarcinoma (ADC) demonstrated a remarkable prognostic effect and enabled numerous patients to benefit from adjuvant chemotherapy. We sought to build a CT-based nomogram for preoperative prediction of the IASLC grading.

MATERIALS AND METHODS

This work retrospectively analyzed the CT images and clinical data of 303 patients with pathologically confirmed invasive ADC. The histological subtypes and radiological characteristics of the patients were re-evaluated. Radiomics features were extracted, and the optimal subset of features was established by ANOVA, spearman correlation analysis, and the least absolute shrinkage and selection operator (LASSO). Univariate and multivariate analyses identified the independent clinical and radiological variables. Finally, multivariate logistic regression analysis incorporated clinical, radiological, and optimal radiomics features into the nomogram. Receiver operating characteristic (ROC) curve, and accuracy were applied to assess the model's performance. Decision curve analysis (DCA), and calibration curve were applied to assess the clinical usefulness.

RESULTS

Nine selected CT image features were used to develop the radiomics model. The accuracy, precision, sensitivity, and specificity of the radiomics model outperformed the clinic-radiological model in the training and testing sets. Integrating Radscore with independent radiological characteristics showed higher prediction performance than clinic-radiological characteristics alone in the training (AUC, 0.915 vs. 0.882; DeLong, p < 0.05) and testing (AUC, 0.838 vs. 0.782; DeLong, p < 0.05) sets. Good calibration and decision curve analysis demonstrated the clinical usefulness of the nomogram.

CONCLUSION

Radiomics features effectively predict high-grade ADC. The combined nomogram may facilitate selecting patients who benefit from adjuvant treatment.

Comparative Evaluation of Breast Ductal Carcinoma Grading: A Deep-Learning Model and General Pathologists' Assessment Approach

Breast cancer is the most prevalent neoplasia among women, with early and accurate diagnosis critical for effective treatment. In clinical practice, however, the subjective nature of histological grading of infiltrating ductal adenocarcinoma of the breast (DAC-NOS) often leads to inconsistencies among pathologists, posing a significant challenge to achieving optimal patient outcomes. Our study aimed to address this reproducibility problem by leveraging artificial intelligence (AI). We trained a deep-learning model using a convolutional neural network-based algorithm (CNN-bA) on 100 whole slide images (WSIs) of DAC-NOS from the Cancer Genome Atlas Breast Invasive Carcinoma (TCGA-BRCA) dataset. Our model demonstrated high precision, sensitivity, and F1 score across different grading components in about 17.5 h with 19,000 iterations. However, the agreement between the model's grading and that of general pathologists varied, showing the highest agreement for the mitotic count score. These findings suggest that AI has the potential to enhance the accuracy and reproducibility of breast cancer grading, warranting further refinement and validation of this approach.

Ensemble deep learning enhanced with self-attention for predicting immunotherapeutic responses to cancers

Introduction

Despite the many benefits immunotherapy has brought to patients with different cancers, its clinical applications and improvements are still hindered by drug resistance. Fostering a reliable approach to identifying sufferers who are sensitive to certain immunotherapeutic agents is of great clinical relevance.

Methods

We propose an ELISE (Ensemble Learning for Immunotherapeutic Response Evaluation) pipeline to generate a robust and highly accurate approach to predicting individual responses to immunotherapies. ELISE employed iterative univariable logistic regression to select genetic features of patients, using Monte Carlo Tree Search (MCTS) to tune hyperparameters. In each trial, ELISE selected multiple models for integration based on add or concatenate stacking strategies, including deep neural network, automatic feature interaction learning via self-attentive neural networks, deep factorization machine, compressed interaction network, and linear neural network, then adopted the best trial to generate a final approach. SHapley Additive exPlanations (SHAP) algorithm was applied to interpret ELISE, which was then validated in an independent test set.

Result

Regarding prediction of responses to atezolizumab within esophageal adenocarcinoma (EAC) patients, ELISE demonstrated a superior accuracy (Area Under Curve [AUC] = 100.00%). AC005786.3 (Mean [|SHAP value|] = 0.0097) was distinguished as the most valuable contributor to ELISE output, followed by SNORD3D (0.0092), RN7SKP72 (0.0081), EREG (0.0069), IGHV4-80 (0.0063), and MIR4526 (0.0063). Mechanistically, immunoglobulin complex, immunoglobulin production, adaptive immune response, antigen binding and others, were downregulated in ELISE-neg EAC subtypes and resulted in unfavorable responses. More encouragingly, ELISE could be extended to accurately estimate the responsiveness of various immunotherapeutic agents against other cancers, including PD1/PD-L1 suppressor against metastatic urothelial cancer (AUC = 88.86%), and MAGE-A3 immunotherapy against metastatic melanoma (AUC = 100.00%).

Discussion

This study presented deep insights into integrating ensemble deep learning with self-attention as a mechanism for predicting immunotherapy responses to human cancers, highlighting ELISE as a potential tool to generate reliable approaches to individualized treatment.

Machine learning with in silico analysis markedly improves survival prediction modeling in colon cancer patients

BACKGROUND

Predicting the survival of cancer patients provides prognostic information and therapeutic guidance. However, improved prediction models are needed for use in diagnosis and treatment.

OBJECTIVE

This study aimed to identify genomic prognostic biomarkers related to colon cancer (CC) based on computational data and to develop survival prediction models.

METHODS

We performed machine-learning (ML) analysis to screen pathogenic survival-related driver genes related to patient prognosis by integrating copy number variation and gene expression data. Moreover, in silico system analysis was performed to clinically assess data from ML analysis, and we identified RABGAP1L, MYH9, and DRD4 as candidate genes. These three genes and tumor stages were used to generate survival prediction models. Moreover, the genes were validated by experimental and clinical analyses, and the theranostic application of the survival prediction models was assessed.

RESULTS

RABGAP1L, MYH9, and DRD4 were identified as survival-related candidate genes by ML and in silico system analysis. The survival prediction model using the expression of the three genes showed higher predictive performance when applied to predict the prognosis of CC patients. A series of functional analyses revealed that each knockdown of three genes reduced the protumor activity of CC cells. In particular, validation with an independent cohort of CC patients confirmed that the coexpression of MYH9 and DRD4 gene expression reflected poorer clinical outcomes in terms of overall survival and disease-free survival.

CONCLUSIONS

Our survival prediction approach will contribute to providing information on patients and developing a therapeutic strategy for CC patients.

The prognostic impact of lung adenocarcinoma predominance classification relating to pathological factors in lobectomy, the Japanese Joint Committee of Lung Cancer Registry Database in 2010

Objective

We studied the prognosis and clinicopathological background of lung adenocarcinoma predominance among patients who underwent lobectomy using data from the Japanese Joint Committee of Lung Cancer Registry.

Methods

Two thousand eight hundred sixty-three cases were extracted. Recurrence free survival (RFS) rates, overall survival (OS) rates and clinicopathological factors and epidermal growth factor receptor (EGFR) mutation status were examined.

Results

Median follow-up period was 65.5 months. Adenocarcinoma predominance was sub-grouped according to OS and RFS rate. In pathological stage I, 5-year RFS and OS rates were respectively 92.2% and 95.8% in group A (adenocarcinoma-in-situ + minimally invasive adenocarcinoma), 89.3% and 92.1% in group B (lepidic), 79.2% and 89.7% in group C (papillary + acinar + variants) and 69.0% and 79.0% in group D (solid + micropapillary). In pathological stage II + IIIA, they were, 43.6% and 72.4% in B, 39.5% and 66.9% in C and 31.0% and 53.7% in D. Group D showed significant worst outcome both in stage I and II + IIIA. Up stage rate from clinical stage I to pathological stage II + IIIA was 0.0%, 3.7%, 15.9% and 33.3%. The frequency of lymph-vessel, vascular, pleura invasion and positive EGFR mutation were 0.0%, 0.0%, 0.0% and 57.1% in group A, 15.6%, 10.0%, 12.1% and 55.1% in B, 36.6%, 31.8%, 29.7% and 44.9% in C, 50.2%, 57.8%, 38.9% and 21.3% in D. In group D, lymph-vessel, vascular and pleura invasion were most, EGFR mutation was least frequent not only in pathological stage I but also stage II + IIIA. In multivariate analysis, age, pathological stage, vascular invasion, and group D were independent factors affected RFS and OS.

Conclusion

Limited to lobectomy cases, solid + micropapillary was independent prognostic factor both in early and locally advanced stage. Its malignant degree was related to the frequency of pathological invasive factors and EGFR mutation status.

Pan-cancer integrative histology-genomic analysis via multimodal deep learning

The rapidly emerging field of computational pathology has demonstrated promise in developing objective prognostic models from histology images. However, most prognostic models are either based on histology or genomics alone and do not address how these data sources can be integrated to develop joint image-omic prognostic models. Additionally, identifying explainable morphological and molecular descriptors from these models that govern such prognosis is of interest. We use multimodal deep learning to jointly examine pathology whole-slide images and molecular profile data from 14 cancer types. Our weakly supervised, multimodal deep-learning algorithm is able to fuse these heterogeneous modalities to predict outcomes and discover prognostic features that correlate with poor and favorable outcomes. We present all analyses for morphological and molecular correlates of patient prognosis across the 14 cancer types at both a disease and a patient level in an interactive open-access database to allow for further exploration, biomarker discovery, and feature assessment.

Infiltration pattern predicts metastasis and progression better than the T-stage and grade in pancreatic neuroendocrine tumors: a proposal for a novel infiltration-based morphologic grading

The advancing edge profile is a powerful determinant of tumor behavior in many organs. In this study, a grading system assessing the tumor-host interface was developed and tested in 181 pancreatic neuroendocrine tumors (PanNETs), 63 of which were <=2 cm. Three tumor slides representative of the spectrum (least, medium, and most) of invasiveness at the advancing edge of the tumor were selected, and then each slide was scored as follows. Well-demarcated/encapsulated, 1 point; Mildly irregular borders and/or minimal infiltration into adjacent tissue, 2 points; Infiltrative edges with several clusters beyond the main tumor but still relatively close, and/or satellite demarcated nodules, 3 points; No demarcation, several cellular clusters away from the tumor, 4 points; Exuberantly infiltrative pattern, scirrhous growth, dissecting the normal parenchymal elements, 5 points. The sum of the rankings on the three slides was obtained. Cases with scores of 3-6 were defined as "non/minimally infiltrative" (NI; n = 77), 7-9 as "moderately infiltrative" (MI; n = 68), and 10-15 as "highly infiltrative" (HI; n = 36). In addition to showing a statistically significant correlation with all the established signs of aggressiveness (grade, size, T-stage), this grading system was found to be the most significant predictor of adverse outcomes (metastasis, progression, and death) on multivariate analysis, more strongly than T-stage, while Ki-67 index did not stand the multivariate test. As importantly, cases <=2 cm were also stratified by this grading system rendering it applicable also to this group that is currently placed in "watchful waiting" protocols. In conclusion, the proposed grading system has a strong, independent prognostic value and therefore should be considered for integration into routine pathology practice after being evaluated in validation studies with larger series.

Quantitative analysis of breast cancer tissue composition and associations with tumor subtype

The tumor microenvironment is an important determinant of breast cancer progression, but standard methods for describing the tumor microenvironment are lacking. Measures of microenvironment composition such as stromal area and immune infiltrate are labor-intensive and few large studies have systematically collected this data. However, digital histologic approaches are becoming more widely available, allowing high-throughput, quantitative estimation. We applied such methods to tissue microarrays of tumors from 1687 women (mean 4 cores per case) in the Carolina Breast Cancer Study Phase 3. Tumor composition was quantified as percentage of epithelium, stroma, adipose, and lymphocytic infiltrate (with the latter as presence/absence using a ≥1% cutoff). Composition proportions and presence/absence were evaluated in association with clinical and molecular features of breast cancer (intrinsic subtype and RNA-based risk of recurrence [ROR] scores) using multivariable linear and logistic regression. Lower stromal content was associated with aggressive tumor phenotypes, including triple-negative (31.1% vs. 41.6% in HR+/HER2-; RFD [95% CI]: -10.5%, [-13.1, -7.9]), Basal-like subtypes (29.0% vs. 44.0% in Luminal A; RFD [95% CI]: -14.9%, [-17.8, -12.0]), and high RNA-based PAM50 ROR scores (27.6% vs. 48.1% in ROR low; RFD [95% CI]: -20.5%, [24.3, 16.7]), after adjusting for age and race. HER2+ tumors also had lower stromal content, particularly among RNA-based HER2-enriched (35.2% vs. 44.0% in Luminal A; RFD [95% CI]: -8.8%, [-13.8, -3.8]). Similar associations were observed between immune infiltrate and tumor phenotypes. Quantitative digital image analysis of the breast cancer microenvironment showed significant associations with demographic characteristics and biological indicators of aggressive behavior.

Deep Learning Approaches to Colorectal Cancer Diagnosis: A Review

Unprecedented breakthroughs in the development of graphical processing systems have led to great potential for deep learning (DL) algorithms in analyzing visual anatomy from high-resolution medical images. Recently, in digital pathology, the use of DL technologies has drawn a substantial amount of attention for use in the effective diagnosis of various cancer types, especially colorectal cancer (CRC), which is regarded as one of the dominant causes of cancer-related deaths worldwide. This review provides an in-depth perspective on recently published research articles on DL-based CRC diagnosis and prognosis. Overall, we provide a retrospective synopsis of simple image-processing-based and machine learning (ML)-based computer-aided diagnosis (CAD) systems, followed by a comprehensive appraisal of use cases with different types of state-of-the-art DL algorithms for detecting malignancies. We first list multiple standardized and publicly available CRC datasets from two imaging types: colonoscopy and histopathology. Secondly, we categorize the studies based on the different types of CRC detected (tumor tissue, microsatellite instability, and polyps), and we assess the data preprocessing steps and the adopted DL architectures before presenting the optimum diagnostic results. CRC diagnosis with DL algorithms is still in the preclinical phase, and therefore, we point out some open issues and provide some insights into the practicability and development of robust diagnostic systems in future health care and oncology.

Utility of Newly Proposed Grading System From International Association for the Study of Lung Cancer for Invasive Lung Adenocarcinoma

Introduction

The International Association for the Study of Lung Cancer proposed a new grading criteria for invasive adenocarcinoma. However, its utility has not been validated.

Methods

Patients who underwent complete resection of lung adenocarcinoma were included in this study. Then, they were divided into the following three groups on the basis of the criteria recently proposed by the International Association for the Study of Lung Cancer: grade 1, lepidic predominant tumor, with less than 20% of high-grade patterns; grade 2, acinar or papillary predominant tumor, with less than 20% of high-grade patterns; and grade 3, any tumor with greater than or equal to 20% of high-grade patterns.

Results

Recurrence-free survival (RFS) was significantly different among the proposed grades (p < 0.001). The RFS of patients upgrading from current grade 2 (papillary or acinar predominant tumor) to proposed grade 3 (5-y RFS, 65.2%) was significantly worse than that of patients with proposed grade 2 (77.1%, hazard ratio = 1.882, 95% confidence interval: 1.236–2.866) but not significantly different from that of patients with grade 3 in both the current (micropapillary or solid predominant tumor) and proposed criteria (53.2%, hazard ratio = 0.761, 95% confidence interval: 0.456–1.269). Among patients with pathologic stage 0 or I, RFS was well stratified by the new grading system (p < 0.001) but not among patients with stage II or III (p = 0.334). In the multivariable analysis, the new grading was not a predictive factor of RFS.

Conclusions

Although the proposed grading system well stratified RFS in patients with pathologic stage 0 or I lung adenocarcinoma, there is room for improvement.

Deep learning in cancer diagnosis, prognosis and treatment selection

Deep learning is a subdiscipline of artificial intelligence that uses a machine learning technique called artificial neural networks to extract patterns and make predictions from large data sets. The increasing adoption of deep learning across healthcare domains together with the availability of highly characterised cancer datasets has accelerated research into the utility of deep learning in the analysis of the complex biology of cancer. While early results are promising, this is a rapidly evolving field with new knowledge emerging in both cancer biology and deep learning. In this review, we provide an overview of emerging deep learning techniques and how they are being applied to oncology. We focus on the deep learning applications for omics data types, including genomic, methylation and transcriptomic data, as well as histopathology-based genomic inference, and provide perspectives on how the different data types can be integrated to develop decision support tools. We provide specific examples of how deep learning may be applied in cancer diagnosis, prognosis and treatment management. We also assess the current limitations and challenges for the application of deep learning in precision oncology, including the lack of phenotypically rich data and the need for more explainable deep learning models. Finally, we conclude with a discussion of how current obstacles can be overcome to enable future clinical utilisation of deep learning.

Grading of invasive breast carcinoma: the way forward

Histologic grading has been a simple and inexpensive method to assess tumor behavior and prognosis of invasive breast cancer grading, thereby identifying patients at risk for adverse outcomes, who may be eligible for (neo)adjuvant therapies. Histologic grading needs to be performed accurately, on properly fixed specimens, and by adequately trained dedicated pathologists that take the time to diligently follow the protocol methodology. In this paper, we review the history of histologic grading, describe the basics of grading, review prognostic value and reproducibility issues, compare performance of grading to gene expression profiles, and discuss how to move forward to improve reproducibility of grading by training, feedback and artificial intelligence algorithms, and special stains to better recognize mitoses. We conclude that histologic grading, when adequately carried out, remains to be of important prognostic value in breast cancer patients.

Fractal dimension analysis as an easy computational approach to improve breast cancer histopathological diagnosis

Histopathology is a well-established standard diagnosis employed for the majority of malignancies, including breast cancer. Nevertheless, despite training and standardization, it is considered operator-dependent and errors are still a concern. Fractal dimension analysis is a computational image processing technique that allows assessing the degree of complexity in patterns. We aimed here at providing a robust and easily attainable method for introducing computer-assisted techniques to histopathology laboratories. Slides from two databases were used: A) Breast Cancer Histopathological; and B) Grand Challenge on Breast Cancer Histology. Set A contained 2480 images from 24 patients with benign alterations, and 5429 images from 58 patients with breast cancer. Set B comprised 100 images of each type: normal tissue, benign alterations, in situ carcinoma, and invasive carcinoma. All images were analyzed with the FracLac algorithm in the ImageJ computational environment to yield the box count fractal dimension (Db) results. Images on set A on 40x magnification were statistically different (p = 0.0003), whereas images on 400x did not present differences in their means. On set B, the mean Db values presented promissing statistical differences when comparing. Normal and/or benign images to in situ and/or invasive carcinoma (all p < 0.0001). Interestingly, there was no difference when comparing normal tissue to benign alterations. These data corroborate with previous work in which fractal analysis allowed differentiating malignancies. Computer-aided diagnosis algorithms may beneficiate from using Db data; specific Db cut-off values may yield ~ 99% specificity in diagnosing breast cancer. Furthermore, the fact that it allows assessing tissue complexity, this tool may be used to understand the progression of the histological alterations in cancer.

Histologic grading of breast carcinoma: a multi-institution study of interobserver variation using virtual microscopy

Breast carcinoma grading is an important prognostic feature recently incorporated into the AJCC Cancer Staging Manual. There is increased interest in applying virtual microscopy (VM) using digital whole slide imaging (WSI) more broadly. Little is known regarding concordance in grading using VM and how such variability might affect AJCC prognostic staging (PS). We evaluated interobserver variability amongst a multi-institutional group of breast pathologists using digital WSI and how discrepancies in grading would affect PS. A digitally scanned slide from 143 invasive carcinomas was independently reviewed by 6 pathologists and assigned grades based on established criteria for tubule formation (TF), nuclear pleomorphism (NP), and mitotic count (MC). Statistical analysis was performed. Interobserver agreement for grade was moderate (κ = 0.497). Agreement was fair (κ = 0.375), moderate (κ = 0.491), and good (κ = 0.705) for grades 2, 3, and 1, respectively. Observer pair concordance ranged from fair to good (κ = 0.354-0.684) Perfect agreement was observed in 43 cases (30%). Interobserver agreement for the individual components was best for TF (κ = 0.503) and worst for MC (κ = 0.281). Seventeen of 86 (19.8%) discrepant cases would have resulted in changes in PS and discrepancies most frequently resulted in a PS change from IA to IB (n = 9). For two of these nine cases, Oncotype DX results would have led to a PS of 1A regardless of grade. Using VM, a multi-institutional cohort of pathologists showed moderate concordance for breast cancer grading, similar to studies using light microscopy. Agreement was the best at the extremes of grade and for evaluation of TF. Whether the higher variability noted for MC is a consequence of VM grading warrants further investigation. Discordance in grading infrequently leads to clinically meaningful changes in the prognostic stage.

Breast cancer tumor heterogeneity has only little impact on the estimation of the Oncotype DX® recurrence score using Magee Equations and Magee Decision Algorithm™

Oncotype DX® assay is used to guide therapeutic decisions in early-stage invasive breast carcinoma but remains expensive. Magee Equations (MEs) and Magee Decision Algorithm (MDA) predict the Oncotype DX® recurrence score (RS) on the basis of histopathological parameters. The influence of intratumor heterogeneity on MEs and MDA remains uncertain. We compared Ki-67, estrogen and progesterone receptors, and human erb-b2 receptor tyrosine kinase 2 (HER2) status on tissue microarray cores with the corresponding findings on the whole slides to calculate MEs scores and to decide if Oncotype DX® testing was required as per MDA in two sets of 175 and 59 tumors, without and with Oncotype DX® results, respectively. Agreements in the interpretation of Ki-67, estrogen and progesterone receptors, and HER2 status were very good between limited areas and whole-slide analyses. This resulted also in very good agreements about the results of MEs and MDA. For 7 of 175 (4%) and 3 of 59 (5.1%) cases, MEs and MDA results in different tumor areas would have changed the indication to perform or not perform Oncotype DX® assays. Oncotype DX® RSs were significantly correlated with MEs and MDA results, but among cases initially predicted to have an RS ≤25 using MDA, 3 of 34 cases (8.8%) had in fact an RS >25. Tumor heterogeneity appears to have little impact on the estimation of the Oncotype DX® RS using MEs and MDA and would have permitted to avoid half of Oncotype DX® assays in our series. Caution is nevertheless required in discarding Oncotype DX® assay in cases with ME scores >18 associated with low mitotic activity.

A Grading System for Invasive Pulmonary Adenocarcinoma: A Proposal From the International Association for the Study of Lung Cancer Pathology Committee

INTRODUCTION

A grading system for pulmonary adenocarcinoma has not been established. The International Association for the Study of Lung Cancer pathology panel evaluated a set of histologic criteria associated with prognosis aimed at establishing a grading system for invasive pulmonary adenocarcinoma.

METHODS

A multi-institutional study involving multiple cohorts of invasive pulmonary adenocarcinomas was conducted. A cohort of 284 stage I pulmonary adenocarcinomas was used as a training set to identify histologic features associated with patient outcomes (recurrence-free survival [RFS] and overall survival [OS]). Receiver operating characteristic curve analysis was used to select the best model, which was validated (n = 212) and tested (n = 300, including stage I-III) in independent cohorts. Reproducibility of the model was assessed using kappa statistics.

RESULTS

The best model (area under the receiver operating characteristic curve [AUC] = 0.749 for RFS and 0.787 for OS) was composed of a combination of predominant plus high-grade histologic pattern with a cutoff of 20% for the latter. The model consists of the following: grade 1, lepidic predominant tumor; grade 2, acinar or papillary predominant tumor, both with no or less than 20% of high-grade patterns; and grade 3, any tumor with 20% or more of high-grade patterns (solid, micropapillary, or complex gland). Similar results were seen in the validation (AUC = 0.732 for RFS and 0.787 for OS) and test cohorts (AUC = 0.690 for RFS and 0.743 for OS), confirming the predictive value of the model. Interobserver reproducibility revealed good agreement (k = 0.617).

CONCLUSIONS

A grading system based on the predominant and high-grade patterns is practical and prognostic for invasive pulmonary adenocarcinoma.