AI improves accuracy, agreement and efficiency of pathologists for Ki67 assessments in breast cancer.

The Ki-67 proliferation index (PI) guides treatment decisions in breast cancer but suffers from poor inter-rater reproducibility. Although AI tools have been designed for Ki-67 assessment, their impact on pathologists' work remains understudied. 90 international pathologists were recruited to assess the Ki-67 PI of ten breast cancer tissue microarrays with and without AI. Accuracy, agreement, and turnaround time with and without AI were compared. Pathologists' perspectives on AI were collected. Using AI led to a significant decrease in PI error (2.1% with AI vs. 5.9% without AI, p < 0.001), better inter-rater agreement (ICC: 0.70 vs. 0.92; Krippendorff's α: 0.63 vs. 0.89; Fleiss' Kappa: 0.40 vs. 0.86), and an 11.9% overall median reduction in turnaround time. Most pathologists (84%) found the AI reliable. For Ki-67 assessments, 76% of respondents believed AI enhances accuracy, 82% said it improves consistency, and 83% trust it will improve efficiency. This study highlights AI's potential to standardize Ki-67 scoring, especially between 5 and 30% PI-a range with low PI agreement. This could pave the way for a universally accepted PI score to guide treatment decisions, emphasizing the promising role of AI integration into pathologist workflows.

Reliability and Variability of Ki-67 Digital Image Analysis Methods for Clinical Diagnostics in Breast Cancer.

Ki-67 is a nuclear protein associated with proliferation, and a strong potential biomarker in breast cancer, but is not routinely measured in current clinical management owing to a lack of standardization. Digital image analysis (DIA) is a promising technology that could allow high-throughput analysis and standardization. There is a dearth of data on the clinical reliability as well as intra- and interalgorithmic variability of different DIA methods. In this study, we scored and compared a set of breast cancer cases in which manually counted Ki-67 has already been demonstrated to have prognostic value (n = 278) to 5 DIA methods, namely Aperio ePathology (Lieca Biosystems), Definiens Tissue Studio (Definiens AG), Qupath, an unsupervised immunohistochemical color histogram algorithm, and a deep-learning pipeline piNET. The piNET system achieved high agreement (interclass correlation coefficient: 0.850) and correlation (R = 0.85) with the reference score. The Qupath algorithm exhibited a high degree of reproducibility among all rater instances (interclass correlation coefficient: 0.889). Although piNET performed well against absolute manual counts, none of the tested DIA methods classified common Ki-67 cutoffs with high agreement or reached the clinically relevant Cohen's κ of at least 0.8. The highest agreement achieved was a Cohen's κ statistic of 0.73 for cutoffs 20% and 25% by the piNET system. The main contributors to interalgorithmic variation and poor cutoff characterization included heterogeneous tumor biology, varying algorithm implementation, and setting assignments. It appears that image segmentation is the primary explanation for semiautomated intra-algorithmic variation, which involves significant manual intervention to correct. Automated pipelines, such as piNET, may be crucial in developing robust and reproducible unbiased DIA approaches to accurately quantify Ki-67 for clinical diagnosis in the future.

Mitosis domain generalization in histopathology images - The MIDOG challenge.

The density of mitotic figures (MF) within tumor tissue is known to be highly correlated with tumor proliferation and thus is an important marker in tumor grading. Recognition of MF by pathologists is subject to a strong inter-rater bias, limiting its prognostic value. State-of-the-art deep learning methods can support experts but have been observed to strongly deteriorate when applied in a different clinical environment. The variability caused by using different whole slide scanners has been identified as one decisive component in the underlying domain shift. The goal of the MICCAI MIDOG 2021 challenge was the creation of scanner-agnostic MF detection algorithms. The challenge used a training set of 200 cases, split across four scanning systems. As test set, an additional 100 cases split across four scanning systems, including two previously unseen scanners, were provided. In this paper, we evaluate and compare the approaches that were submitted to the challenge and identify methodological factors contributing to better performance. The winning algorithm yielded an F1 score of 0.748 (CI95: 0.704-0.781), exceeding the performance of six experts on the same task.

Explainable Biomarkers for Automated Glomerular and Patient-Level Disease Classification.

Pathologists use multiple microscopy modalities to assess renal biopsy specimens. Besides usual diagnostic features, some changes are too subtle to be properly defined. Computational approaches have the potential to systematically quantitate subvisual clues, provide pathogenetic insight, and link to clinical outcomes. To this end, a proof-of-principle study is presented demonstrating that explainable biomarkers through machine learning can distinguish between glomerular disorders at the light-microscopy level. The proposed system used image analysis techniques and extracted 233 explainable biomarkers related to color, morphology, and microstructural texture. Traditional machine learning was then used to classify minimal change disease (MCD), membranous nephropathy (MN), and thin basement membrane nephropathy (TBMN) diseases on a glomerular and patient-level basis. The final model combined the Gini feature importance set and linear discriminant analysis classifier. Six morphologic (nuclei-to-glomerular tuft area, nuclei-to-glomerular area, glomerular tuft thickness greater than ten, glomerular tuft thickness greater than three, total glomerular tuft thickness, and glomerular circularity) and four microstructural texture features (luminal contrast using wavelets, nuclei energy using wavelets, nuclei variance using color vector LBP, and glomerular correlation using GLCM) were, together, the best performing biomarkers. Accuracies of 77% and 87% were obtained for glomerular and patient-level classification, respectively. Computational methods, using explainable glomerular biomarkers, have diagnostic value and are compatible with our existing knowledge of disease pathogenesis. Furthermore, this algorithm can be applied to clinical datasets for novel prognostic and mechanistic biomarker discovery.

MiNuGAN: Dual Segmentation of Mitoses and Nuclei Using Conditional GANs on Multi-center Breast H&E Images.

Breast cancer is the second most commonly diagnosed type of cancer among women as of 2021. Grading of histopathological images is used to guide breast cancer treatment decisions and a critical component of this is a mitotic score, which is related to tumor aggressiveness. Manual mitosis counting is an extremely tedious manual task, but automated approaches can be used to overcome inefficiency and subjectivity. In this paper, we propose an automatic mitosis and nuclear segmentation method for a diverse set of H&E breast cancer pathology images. The method is based on a conditional generative adversarial network to segment both mitoses and nuclei at the same time. Architecture optimizations are investigated, including hyper parameters and the addition of a focal loss. The accuracy of the proposed method is investigated using images from multiple centers and scanners, including TUPAC16, ICPR14 and ICPR12 datasets. In TUPAC16, we use 618 carefully annotated images of size 256×256 scanned at 40×. TUPAC16 is used to train the model, and segmentation performance is measured on the test set for both nuclei and mitoses. Results on 200 held-out testing images from the TUPAC16 dataset were mean DSC = 0.784 and 0.721 for nuclear and mitosis, respectively. On 202 ICPR12 images, mitosis segmentation accuracy had a mean DSC = 0.782, indicating the model generalizes well to unseen datasets. For datasets that had mitosis centroid annotations, which included 200 TUPAC16, 202 ICPR12 and 524 ICPR14, a mean F1-score of 0.854 was found indicating high mitosis detection accuracy.

Machine learning in renal pathology.

Introduction: When assessing kidney biopsies, pathologists use light microscopy, immunofluorescence, and electron microscopy to describe and diagnose glomerular lesions and diseases. These methods can be laborious, costly, fraught with inter-observer variability, and can have delays in turn-around time. Thus, computational approaches can be designed as screening and/or diagnostic tools, potentially relieving pathologist time, healthcare resources, while also having the ability to identify novel biomarkers, including subvisual features. Methods: Here, we implement our recently published biomarker feature extraction (BFE) model along with 3 pre-trained deep learning models (VGG16, VGG19, and InceptionV3) to diagnose 3 glomerular diseases using PAS-stained digital pathology images alone. The BFE model extracts a panel of 233 explainable features related to underlying pathology, which are subsequently narrowed down to 10 morphological and microstructural texture features for classification with a linear discriminant analysis machine learning classifier. 45 patient renal biopsies (371 glomeruli) from minimal change disease (MCD), membranous nephropathy (MN), and thin-basement membrane nephropathy (TBMN) were split into training/validation and held out sets. For the 3 deep learningmodels, data augmentation and Grad-CAM were used for better performance and interpretability. Results: The BFE model showed glomerular validation accuracy of 67.6% and testing accuracy of 76.8%. All deep learning approaches had higher validation accuracies (most for VGG16 at 78.5%) but lower testing accuracies. The highest testing accuracy at the glomerular level was VGG16 at 71.9%, while at the patient-level was InceptionV3 at 73.3%. Discussion: The results highlight the potential of both traditional machine learning and deep learning-based approaches for kidney biopsy evaluation.

piNET-An Automated Proliferation Index Calculator Framework for Ki67 Breast Cancer Images.

In this work, a novel proliferation index (PI) calculator for Ki67 images called piNET is proposed. It is successfully tested on four datasets, from three scanners comprised of patches, tissue microarrays (TMAs) and whole slide images (WSI), representing a diverse multi-centre dataset for evaluating Ki67 quantification. Compared to state-of-the-art methods, piNET consistently performs the best over all datasets with an average PI difference of 5.603%, PI accuracy rate of 86% and correlation coefficient R = 0.927. The success of the system can be attributed to several innovations. Firstly, this tool is built based on deep learning, which can adapt to wide variability of medical images-and it was posed as a detection problem to mimic pathologists' workflow which improves accuracy and efficiency. Secondly, the system is trained purely on tumor cells, which reduces false positives from non-tumor cells without needing the usual pre-requisite tumor segmentation step for Ki67 quantification. Thirdly, the concept of learning background regions through weak supervision is introduced, by providing the system with ideal and non-ideal (artifact) patches that further reduces false positives. Lastly, a novel hotspot analysis is proposed to allow automated methods to score patches from WSI that contain "significant" activity.

Assessing the Impact of Color Normalization in Convolutional Neural Network-Based Nuclei Segmentation Frameworks.

Image analysis tools for cancer, such as automatic nuclei segmentation, are impacted by the inherent variation contained in pathology image data. Convolutional neural networks (CNN), demonstrate success in generalizing to variable data, illustrating great potential as a solution to the problem of data variability. In some CNN-based segmentation works for digital pathology, authors apply color normalization (CN) to reduce color variability of data as a preprocessing step prior to prediction, while others do not. Both approaches achieve reasonable performance and yet, the reasoning for utilizing this step has not been justified. It is therefore important to evaluate the necessity and impact of CN for deep learning frameworks, and its effect on downstream processes. In this paper, we evaluate the effect of popular CN methods on CNN-based nuclei segmentation frameworks.

IHC Color Histograms for Unsupervised Ki67 Proliferation Index Calculation.

Automated image analysis tools for Ki67 breast cancer digital pathology images would have significant value if integrated into diagnostic pathology workflows. Such tools would reduce the workload of pathologists, while improving efficiency, and accuracy. Developing tools that are robust and reliable to multicentre data is challenging, however, differences in staining protocols, digitization equipment, staining compounds, and slide preparation can create variabilities in image quality and color across digital pathology datasets. In this work, a novel unsupervised color separation framework based on the IHC color histogram (IHCCH) is proposed for the robust analysis of Ki67 and hematoxylin stained images in multicentre datasets. An "overstaining" threshold is implemented to adjust for background overstaining, and an automated nuclei radius estimator is designed to improve nuclei detection. Proliferation index and F1 scores were compared between the proposed method and manually labeled ground truth data for 30 TMA cores that have ground truths for Ki67+ and Ki67- nuclei. The method accurately quantified the PI over the dataset, with an average proliferation index difference of 3.25%. To ensure the method generalizes to new, diverse datasets, 50 Ki67 TMAs from the Protein Atlas were used to test the validated approach. As the ground truth for this dataset is PI ranges, the automated result was compared to the PI range. The proposed method correctly classified 74 out of 80 TMA images, resulting in a 92.5% accuracy. In addition to these validations experiments, performance was compared to two color-deconvolution based methods, and to six machine learning classifiers. In all cases, the proposed work maintained more consistent (reproducible) results, and higher PI quantification accuracy.

Comparison of JPEG 2000 and Other Lossless Compression Schemes for Digital Mammograms.

In this study, we propose JPEG 2000 as an algorithm for the compression of digital mammograms and the proposed work is the first real-time implementation of JPEG 2000 on a mammogram image database. Only the lossless compression mode of JPEG 2000 was examined to ensure that the mammogram is delivered without distortion. The performance of JPEG 2000 was compared against several other lossless coders: JPEG-LS, lossless-JPEG, adaptive Huffman, arithmetic with a zero order and a first order probability model and Lempel-Ziv Welch (LZW) with 12 and 15 bit dictionaries. Each compressor was supplied the identical set of 50 mammograms, each having a resolution of 8bits/pixel and dimensions of 1024 × 1024. Experimental results indicate JPEG 2000 and JPEG-LS provide comparable compression performance since their compression ratios differed by 0.72% and both compressors also superseded the results of the other coders. Although JPEG 2000 suffered from a slightly longer encoding and decoding delay than JPEG-LS (0.8s on average), it is still preferred for mammogram images due to the wide variety of features that aid in reliable image transmission, provide an efficient mechanism for remote access to digital libraries and contribute to fast database access.