Self-Learning for Weakly Supervised Gleason Grading of Local Patterns

Probabilistic Attention Based on Gaussian Processes for Deep Multiple Instance Learning

Multiple instance learning (MIL) is a weakly supervised learning paradigm that is becoming increasingly popular because it requires less labeling effort than fully supervised methods. This is especially interesting for areas where the creation of large annotated datasets remains challenging, as in medicine. Although recent deep learning MIL approaches have obtained state-of-the-art results, they are fully deterministic and do not provide uncertainty estimations for the predictions. In this work, we introduce the attention Gaussian process (AGP) model, a novel probabilistic attention mechanism based on Gaussian processes (GPs) for deep MIL. AGP provides accurate bag-level predictions as well as instance-level explainability and can be trained end-to-end. Moreover, its probabilistic nature guarantees robustness to overfit on small datasets and uncertainty estimations for the predictions. The latter is especially important in medical applications, where decisions have a direct impact on the patient's health. The proposed model is validated experimentally as follows. First, its behavior is illustrated in two synthetic MIL experiments based on the well-known MNIST and CIFAR-10 datasets, respectively. Then, it is evaluated in three different real-world cancer detection experiments. AGP outperforms state-of-the-art MIL approaches, including deterministic deep learning ones. It shows a strong performance even on a small dataset with less than 100 labels and generalizes better than competing methods on an external test set. Moreover, we experimentally show that predictive uncertainty correlates with the risk of wrong predictions, and therefore it is a good indicator of reliability in practice. Our code is publicly available.

MICIL: Multiple-Instance Class-Incremental Learning for skin cancer whole slide images

Artificial intelligence (AI) agents encounter the problem of catastrophic forgetting when they are trained in sequentially with new data batches. This issue poses a barrier to the implementation of AI-based models in tasks that involve ongoing evolution, such as cancer prediction. Moreover, whole slide images (WSI) play a crucial role in cancer management, and their automated analysis has become increasingly popular in assisting pathologists during the diagnosis process. Incremental learning (IL) techniques aim to develop algorithms capable of retaining previously acquired information while also acquiring new insights to predict future data. Deep IL techniques need to address the challenges posed by the gigapixel scale of WSIs, which often necessitates the use of multiple instance learning (MIL) frameworks. In this paper, we introduce an IL algorithm tailored for analyzing WSIs within a MIL paradigm. The proposed Multiple Instance Class-Incremental Learning (MICIL) algorithm combines MIL with class-IL for the first time, allowing for the incremental prediction of multiple skin cancer subtypes from WSIs within a class-IL scenario. Our framework incorporates knowledge distillation and data rehearsal, along with a novel embedding-level distillation, aiming to preserve the latent space at the aggregated WSI level. Results demonstrate the algorithm's effectiveness in addressing the challenge of balancing IL-specific metrics, such as intransigence and forgetting, and solving the plasticity-stability dilemma.

Don't Fear the Artificial Intelligence: A Systematic Review of Machine Learning for Prostate Cancer Detection in Pathology

CONTEXT

Automated prostate cancer detection using machine learning technology has led to speculation that pathologists will soon be replaced by algorithms. This review covers the development of machine learning algorithms and their reported effectiveness specific to prostate cancer detection and Gleason grading.

OBJECTIVE

To examine current algorithms regarding their accuracy and classification abilities. We provide a general explanation of the technology and how it is being used in clinical practice. The challenges to the application of machine learning algorithms in clinical practice are also discussed.

DATA SOURCES

The literature for this review was identified and collected using a systematic search. Criteria were established prior to the sorting process to effectively direct the selection of studies. A 4-point system was implemented to rank the papers according to their relevancy. For papers accepted as relevant to our metrics, all cited and citing studies were also reviewed. Studies were then categorized based on whether they implemented binary or multi-class classification methods. Data were extracted from papers that contained accuracy, area under the curve (AUC), or κ values in the context of prostate cancer detection. The results were visually summarized to present accuracy trends between classification abilities.

CONCLUSIONS

It is more difficult to achieve high accuracy metrics for multiclassification tasks than for binary tasks. The clinical implementation of an algorithm that can assign a Gleason grade to clinical whole slide images (WSIs) remains elusive. Machine learning technology is currently not able to replace pathologists but can serve as an important safeguard against misdiagnosis.

Harnessing artificial intelligence for prostate cancer management

Prostate cancer (PCa) is a common malignancy in males. The pathology review of PCa is crucial for clinical decision-making, but traditional pathology review is labor intensive and subjective to some extent. Digital pathology and whole-slide imaging enable the application of artificial intelligence (AI) in pathology. This review highlights the success of AI in detecting and grading PCa, predicting patient outcomes, and identifying molecular subtypes. We propose that AI-based methods could collaborate with pathologists to reduce workload and assist clinicians in formulating treatment recommendations. We also introduce the general process and challenges in developing AI pathology models for PCa. Importantly, we summarize publicly available datasets and open-source codes to facilitate the utilization of existing data and the comparison of the performance of different models to improve future studies.

Weakly Supervised MRI Slice-Level Deep Learning Classification of Prostate Cancer Approximates Full Voxel- and Slice-Level Annotation: Effect of Increasing Training Set Size

BACKGROUND

Weakly supervised learning promises reduced annotation effort while maintaining performance.

PURPOSE

To compare weakly supervised training with full slice-wise annotated training of a deep convolutional classification network (CNN) for prostate cancer (PC).

STUDY TYPE

Retrospective.

SUBJECTS

One thousand four hundred eighty-nine consecutive institutional prostate MRI examinations from men with suspicion for PC (65 ± 8 years) between January 2015 and November 2020 were split into training (N = 794, enriched with 204 PROSTATEx examinations) and test set (N = 695).

FIELD STRENGTH/SEQUENCE

1.5 and 3T, T2-weighted turbo-spin-echo and diffusion-weighted echo-planar imaging.

ASSESSMENT

Histopathological ground truth was provided by targeted and extended systematic biopsy. Reference training was performed using slice-level annotation (SLA) and compared to iterative training utilizing patient-level annotations (PLAs) with supervised feedback of CNN estimates into the next training iteration at three incremental training set sizes (N = 200, 500, 998). Model performance was assessed by comparing specificity at fixed sensitivity of 0.97 [254/262] emulating PI-RADS ≥ 3, and 0.88-0.90 [231-236/262] emulating PI-RADS ≥ 4 decisions.

STATISTICAL TESTS

Receiver operating characteristic (ROC) and area under the curve (AUC) was compared using DeLong and Obuchowski test. Sensitivity and specificity were compared using McNemar test. Statistical significance threshold was P = 0.05.

RESULTS

Test set (N = 695) ROC-AUC performance of SLA (trained with 200/500/998 exams) was 0.75/0.80/0.83, respectively. PLA achieved lower ROC-AUC of 0.64/0.72/0.78. Both increased performance significantly with increasing training set size. ROC-AUC for SLA at 500 exams was comparable to PLA at 998 exams (P = 0.28). ROC-AUC was significantly different between SLA and PLA at same training set sizes, however the ROC-AUC difference decreased significantly from 200 to 998 training exams. Emulating PI-RADS ≥ 3 decisions, difference between PLA specificity of 0.12 [51/433] and SLA specificity of 0.13 [55/433] became undetectable (P = 1.0) at 998 exams. Emulating PI-RADS ≥ 4 decisions, at 998 exams, SLA specificity of 0.51 [221/433] remained higher than PLA specificity at 0.39 [170/433]. However, PLA specificity at 998 exams became comparable to SLA specificity of 0.37 [159/433] at 200 exams (P = 0.70).

DATA CONCLUSION

Weakly supervised training of a classification CNN using patient-level-only annotation had lower performance compared to training with slice-wise annotations, but improved significantly faster with additional training data.

EVIDENCE LEVEL

3 TECHNICAL EFFICACY: Stage 2.

A visual-language foundation model for computational pathology

The accelerated adoption of digital pathology and advances in deep learning have enabled the development of robust models for various pathology tasks across a diverse array of diseases and patient cohorts. However, model training is often difficult due to label scarcity in the medical domain, and a model's usage is limited by the specific task and disease for which it is trained. Additionally, most models in histopathology leverage only image data, a stark contrast to how humans teach each other and reason about histopathologic entities. We introduce CONtrastive learning from Captions for Histopathology (CONCH), a visual-language foundation model developed using diverse sources of histopathology images, biomedical text and, notably, over 1.17 million image-caption pairs through task-agnostic pretraining. Evaluated on a suite of 14 diverse benchmarks, CONCH can be transferred to a wide range of downstream tasks involving histopathology images and/or text, achieving state-of-the-art performance on histology image classification, segmentation, captioning, and text-to-image and image-to-text retrieval. CONCH represents a substantial leap over concurrent visual-language pretrained systems for histopathology, with the potential to directly facilitate a wide array of machine learning-based workflows requiring minimal or no further supervised fine-tuning.

One label is all you need: Interpretable AI-enhanced histopathology for oncology

Artificial Intelligence (AI)-enhanced histopathology presents unprecedented opportunities to benefit oncology through interpretable methods that require only one overall label per hematoxylin and eosin (H&E) slide with no tissue-level annotations. We present a structured review of these methods organized by their degree of verifiability and by commonly recurring application areas in oncological characterization. First, we discuss morphological markers (tumor presence/absence, metastases, subtypes, grades) in which AI-identified regions of interest (ROIs) within whole slide images (WSIs) verifiably overlap with pathologist-identified ROIs. Second, we discuss molecular markers (gene expression, molecular subtyping) that are not verified via H&E but rather based on overlap with positive regions on adjacent tissue. Third, we discuss genetic markers (mutations, mutational burden, microsatellite instability, chromosomal instability) that current technologies cannot verify if AI methods spatially resolve specific genetic alterations. Fourth, we discuss the direct prediction of survival to which AI-identified histopathological features quantitatively correlate but are nonetheless not mechanistically verifiable. Finally, we discuss in detail several opportunities and challenges for these one-label-per-slide methods within oncology. Opportunities include reducing the cost of research and clinical care, reducing the workload of clinicians, personalized medicine, and unlocking the full potential of histopathology through new imaging-based biomarkers. Current challenges include explainability and interpretability, validation via adjacent tissue sections, reproducibility, data availability, computational needs, data requirements, domain adaptability, external validation, dataset imbalances, and finally commercialization and clinical potential. Ultimately, the relative ease and minimum upfront cost with which relevant data can be collected in addition to the plethora of available AI methods for outcome-driven analysis will surmount these current limitations and achieve the innumerable opportunities associated with AI-driven histopathology for the benefit of oncology.

Annotation protocol and crowdsourcing multiple instance learning classification of skin histological images: The CR-AI4SkIN dataset

Digital Pathology (DP) has experienced a significant growth in recent years and has become an essential tool for diagnosing and prognosis of tumors. The availability of Whole Slide Images (WSIs) and the implementation of Deep Learning (DL) algorithms have paved the way for the appearance of Artificial Intelligence (AI) systems that support the diagnosis process. These systems require extensive and varied data for their training to be successful. However, creating labeled datasets in histopathology is laborious and time-consuming. We have developed a crowdsourcing-multiple instance labeling/learning protocol that is applied to the creation and use of the CR-AI4SkIN dataset.2 CR-AI4SkIN contains 271 WSIs of 7 Cutaneous Spindle Cell (CSC) neoplasms with expert and non-expert labels at region and WSI levels. It is the first dataset of these types of neoplasms made available. The regions selected by the experts are used to learn an automatic extractor of Regions of Interest (ROIs) from WSIs. To produce the embedding of each WSI, the representations of patches within the ROIs are obtained using a contrastive learning method, and then combined. Finally, they are fed to a Gaussian process-based crowdsourcing classifier, which utilizes the noisy non-expert WSI labels. We validate our crowdsourcing-multiple instance learning method in the CR-AI4SkIN dataset, addressing a binary classification problem (malign vs. benign). The proposed method obtains an F1 score of 0.7911 on the test set, outperforming three widely used aggregation methods for crowdsourcing tasks. Furthermore, our crowdsourcing method also outperforms the supervised model with expert labels on the test set (F1-score = 0.6035). The promising results support the proposed crowdsourcing multiple instance learning annotation protocol. It also validates the automatic extraction of interest regions and the use of contrastive embedding and Gaussian process classification to perform crowdsourcing classification tasks.

Learning how to detect: A deep reinforcement learning method for whole-slide melanoma histopathology images

Cutaneous melanoma represents one of the most life-threatening malignancies. Histopathological image analysis serves as a vital tool for early melanoma detection. Deep neural network (DNN) models are frequently employed to aid pathologists in enhancing the efficiency and accuracy of diagnoses. However, due to the paucity of well-annotated, high-resolution, whole-slide histopathology image (WSI) datasets, WSIs are typically fragmented into numerous patches during the model training and testing stages. This process disregards the inherent interconnectedness among patches, potentially impeding the models' performance. Additionally, the presence of excess, non-contributing patches extends processing times and introduces substantial computational burdens. To mitigate these issues, we draw inspiration from the clinical decision-making processes of dermatopathologists to propose an innovative, weakly supervised deep reinforcement learning framework, titled Fast medical decision-making in melanoma histopathology images (FastMDP-RL). This framework expedites model inference by reducing the number of irrelevant patches identified within WSIs. FastMDP-RL integrates two DNN-based agents: the search agent (SeAgent) and the decision agent (DeAgent). The SeAgent initiates actions, steered by the image features observed in the current viewing field at various magnifications. Simultaneously, the DeAgent provides labeling probabilities for each patch. We utilize multi-instance learning (MIL) to construct a teacher-guided model (MILTG), serving a dual purpose: rewarding the SeAgent and guiding the DeAgent. Our evaluations were conducted using two melanoma datasets: the publicly accessible TCIA-CM dataset and the proprietary MELSC dataset. Our experimental findings affirm FastMDP-RL's ability to expedite inference and accurately predict WSIs, even in the absence of pixel-level annotations. Moreover, our research investigates the WSI-based interactive environment, encompassing the design of agents, state and reward functions, and feature extractors suitable for melanoma tissue images. This investigation offers valuable insights and references for researchers engaged in related studies. The code is available at: https://github.com/titizheng/FastMDP-RL.

Contrastive Multiple Instance Learning: An Unsupervised Framework for Learning Slide-Level Representations of Whole Slide Histopathology Images without Labels

Recent methods in computational pathology have trended towards semi- and weakly-supervised methods requiring only slide-level labels. Yet, even slide-level labels may be absent or irrelevant to the application of interest, such as in clinical trials. Hence, we present a fully unsupervised method to learn meaningful, compact representations of WSIs. Our method initially trains a tile-wise encoder using SimCLR, from which subsets of tile-wise embeddings are extracted and fused via an attention-based multiple-instance learning framework to yield slide-level representations. The resulting set of intra-slide-level and inter-slide-level embeddings are attracted and repelled via contrastive loss, respectively. This resulted in slide-level representations with self-supervision. We applied our method to two tasks- (1) non-small cell lung cancer subtyping (NSCLC) as a classification prototype and (2) breast cancer proliferation scoring (TUPAC16) as a regression prototype-and achieved an AUC of 0.8641 ± 0.0115 and correlation (R²) of 0.5740 ± 0.0970, respectively. Ablation experiments demonstrate that the resulting unsupervised slide-level feature space can be fine-tuned with small datasets for both tasks. Overall, our method approaches computational pathology in a novel manner, where meaningful features can be learned from whole-slide images without the need for annotations of slide-level labels. The proposed method stands to benefit computational pathology, as it theoretically enables researchers to benefit from completely unlabeled whole-slide images.

Constrained multiple instance learning for ulcerative colitis prediction using histological images

BACKGROUND AND OBJECTIVE

Ulcerative colitis (UC) is an inflammatory bowel disease (IBD) affecting the colon and the rectum characterized by a remitting-relapsing course. To detect mucosal inflammation associated with UC, histology is considered the most stringent criteria. In turn, histologic remission (HR) correlates with improved clinical outcomes and has been recently recognized as a desirable treatment target. The leading biomarker for assessing histologic remission is the presence or absence of neutrophils. Therefore, the finding of this cell in specific colon structures indicates that the patient has UC activity. However, no previous studies based on deep learning have been developed to identify UC based on neutrophils detection using whole-slide images (WSI).

METHODS

The methodological core of this work is a novel multiple instance learning (MIL) framework with location constraints able to determine the presence of UC activity using WSI. In particular, we put forward an effective way to introduce constraints about positive instances to effectively explore additional weakly supervised information that is easy to obtain and enjoy a significant boost to the learning process. In addition, we propose a new weighted embedding to enlarge the relevance of the positive instances.

RESULTS

Extensive experiments on a multi-center dataset of colon and rectum WSIs, PICASSO-MIL, demonstrate that using the location information we can improve considerably the results at WSI-level. In comparison with prior MIL settings, our method allows for 10% improvements in bag-level accuracy.

CONCLUSION

Our model, which introduces a new form of constraints, surpass the results achieved from current state-of-the-art methods that focus on the MIL paradigm. Our method can be applied to other histological concerns where the morphological features determining a positive WSI are tiny and similar to others in the image.

Proportion constrained weakly supervised histopathology image classification

Multiple instance learning (MIL) deals with data grouped into bags of instances, of which only the global information is known. In recent years, this weakly supervised learning paradigm has become very popular in histological image analysis because it alleviates the burden of labeling all cancerous regions of large Whole Slide Images (WSIs) in detail. However, these methods require large datasets to perform properly, and many approaches only focus on simple binary classification. This often does not match the real-world problems where multi-label settings are frequent and possible constraints must be taken into account. In this work, we propose a novel multi-label MIL formulation based on inequality constraints that is able to incorporate prior knowledge about instance proportions. Our method has a theoretical foundation in optimization with log-barrier extensions, applied to bag-level class proportions. This encourages the model to respect the proportion ordering during training. Extensive experiments on a new public dataset of prostate cancer WSIs analysis, SICAP-MIL, demonstrate that using the prior proportion information we can achieve instance-level results similar to supervised methods on datasets of similar size. In comparison with prior MIL settings, our method allows for ∼13% improvements in instance-level accuracy, and ∼3% in the multi-label mean area under the ROC curve at the bag-level.

A Deep Learning Model for Prostate Adenocarcinoma Classification in Needle Biopsy Whole-Slide Images Using Transfer Learning

The histopathological diagnosis of prostate adenocarcinoma in needle biopsy specimens is of pivotal importance for determining optimum prostate cancer treatment. Since diagnosing a large number of cases containing 12 core biopsy specimens by pathologists using a microscope is time-consuming manual system and limited in terms of human resources, it is necessary to develop new techniques that can rapidly and accurately screen large numbers of histopathological prostate needle biopsy specimens. Computational pathology applications that can assist pathologists in detecting and classifying prostate adenocarcinoma from whole-slide images (WSIs) would be of great benefit for routine pathological practice. In this paper, we trained deep learning models capable of classifying needle biopsy WSIs into adenocarcinoma and benign (non-neoplastic) lesions. We evaluated the models on needle biopsy, transurethral resection of the prostate (TUR-P), and The Cancer Genome Atlas (TCGA) public dataset test sets, achieving an ROC-AUC up to 0.978 in needle biopsy test sets and up to 0.9873 in TCGA test sets for adenocarcinoma.

An attention-based weakly supervised framework for spitzoid melanocytic lesion diagnosis in whole slide images

Melanoma is an aggressive neoplasm responsible for the majority of deaths from skin cancer. Specifically, spitzoid melanocytic tumors are one of the most challenging melanocytic lesions due to their ambiguous morphological features. The gold standard for its diagnosis and prognosis is the analysis of skin biopsies. In this process, dermatopathologists visualize skin histology slides under a microscope, in a highly time-consuming and subjective task. In the last years, computer-aided diagnosis (CAD) systems have emerged as a promising tool that could support pathologists in daily clinical practice. Nevertheless, no automatic CAD systems have yet been proposed for the analysis of spitzoid lesions. Regarding common melanoma, no system allows both the selection of the tumor region and the prediction of the benign or malignant form in the diagnosis. Motivated by this, we propose a novel end-to-end weakly supervised deep learning model, based on inductive transfer learning with an improved convolutional neural network (CNN) to refine the embedding features of the latent space. The framework is composed of a source model in charge of finding the tumor patch-level patterns, and a target model focuses on the specific diagnosis of a biopsy. The latter retrains the backbone of the source model through a multiple instance learning workflow to obtain the biopsy-level scoring. To evaluate the performance of the proposed methods, we performed extensive experiments on a private skin database with spitzoid lesions. Test results achieved an accuracy of 0.9231 and 0.80 for the source and the target models, respectively. In addition, the heat map findings are directly in line with the clinicians' medical decision and even highlight, in some cases, patterns of interest that were overlooked by the pathologist.

A novel self-learning framework for bladder cancer grading using histopathological images

In recent times, bladder cancer has increased significantly in terms of incidence and mortality. Currently, two subtypes are known based on tumour growth: non-muscle invasive (NMIBC) and muscle-invasive bladder cancer (MIBC). In this work, we focus on the MIBC subtype because it has the worst prognosis and can spread to adjacent organs. We present a self-learning framework to grade bladder cancer from histological images stained by immunohistochemical techniques. Specifically, we propose a novel Deep Convolutional Embedded Attention Clustering (DCEAC) which allows for the classification of histological patches into different levels of disease severity, according to established patterns in the literature. The proposed DCEAC model follows a fully unsupervised two-step learning methodology to discern between non-tumour, mild and infiltrative patterns from high-resolution 512 × 512 pixel samples. Our system outperforms previous clustering-based methods by including a convolutional attention module, which enables the refinement of the features of the latent space prior to the classification stage. The proposed network surpasses state-of-the-art approaches by 2-3% across different metrics, reaching a final average accuracy of 0.9034 in a multi-class scenario. Furthermore, the reported class activation maps evidence that our model is able to learn by itself the same patterns that clinicians consider relevant, without requiring previous annotation steps. This represents a breakthrough in MIBC grading that bridges the gap with respect to training the model on labelled data.