Martin C. Stumpe

Qiyuan Hu, Abbas A. Rizvi, Geoffery Schau et al.

Microsatellite instability-high (MSI-H) is a tumor agnostic biomarker for immune checkpoint inhibitor therapy. However, MSI status is not routinely tested in prostate cancer, in part due to low prevalence and assay cost. As such, prediction of MSI status from hematoxylin and eosin (H&E) stained whole-slide images (WSIs) could identify prostate cancer patients most likely to benefit from confirmatory testing and becoming eligible for immunotherapy. Prostate biopsies and surgical resections from de-identified records of consecutive prostate cancer patients referred to our institution were analyzed. Their MSI status was determined by next generation sequencing. Patients before a cutoff date were split into an algorithm development set (n=4015, MSI-H 1.8%) and a paired validation set (n=173, MSI-H 19.7%) that consisted of two serial sections from each sample, one stained and scanned internally and the other at an external site. Patients after the cutoff date formed the temporal validation set (n=1350, MSI-H 2.3%). Attention-based multiple instance learning models were trained to predict MSI-H from H&E WSIs. The MSI-H predictor achieved area under the receiver operating characteristic curve values of 0.78 (95% CI [0.69-0.86]), 0.72 (95% CI [0.63-0.81]), and 0.72 (95% CI [0.62-0.82]) on the internally prepared, externally prepared, and temporal validation sets, respectively. While MSI-H status is significantly correlated with Gleason score, the model remained predictive within each Gleason score subgroup. In summary, we developed and validated an AI-based MSI-H diagnostic model on a large real-world cohort of routine H&E slides, which effectively generalized to externally stained and scanned samples and a temporally independent validation cohort. This algorithm has the potential to direct prostate cancer patients toward immunotherapy and to identify MSI-H cases secondary to Lynch syndrome.

Rohan P. Joshi, Bolesław L. Osinski, Niha Beig et al.

MET is a proto-oncogene whose somatic activation in non-small cell lung cancer leads to increased cell growth and tumor progression. The two major classes of MET alterations are gene amplification and exon 14 deletion, both of which are therapeutic targets and detectable using existing molecular assays. However, existing tests are limited by their consumption of valuable tissue, cost and complexity that prevent widespread use. MET alterations could have an effect on cell morphology, and quantifying these associations could open new avenues for research and development of morphology-based screening tools. Using H&E-stained whole slide images (WSIs), we investigated the association of distinct cell-morphological features with MET amplifications and MET exon 14 deletions. We found that cell shape, color, grayscale intensity and texture-based features from both tumor infiltrating lymphocytes and tumor cells distinguished MET wild-type from MET amplified or MET exon 14 deletion cases. The association of individual cell features with MET alterations suggested a predictive model could distinguish MET wild-type from MET amplification or MET exon 14 deletion. We therefore developed an L1-penalized logistic regression model, achieving a mean Area Under the Receiver Operating Characteristic Curve (ROC-AUC) of 0.77 +/- 0.05sd in cross-validation and 0.77 on an independent holdout test set. A sparse set of 43 features differentiated these classes, which included features similar to what was found in the univariate analysis as well as the percent of tumor cells in the tissue. Our study demonstrates that MET alterations result in a detectable morphological signal in tumor cells and lymphocytes. These results suggest that development of low-cost predictive models based on H&E-stained WSIs may improve screening for MET altered tumors.

Bolesław L. Osinski, Aïcha BenTaieb, Irvin Ho et al.

To achieve minimum DNA input and tumor purity requirements for next-generation sequencing (NGS), pathologists visually estimate macrodissection and slide count decisions. Misestimation may cause tissue waste and increased laboratory costs. We developed an AI-augmented smart pathology review system (SmartPath) to empower pathologists with quantitative metrics for determining tissue extraction parameters. Using digitized H&E-stained FFPE slides as inputs, SmartPath segments tumors, extracts cell-based features, and suggests macrodissection areas. To predict DNA yield per slide, the extracted features are correlated with known DNA yields. Then, a pathologist-defined target yield divided by the predicted DNA yield/slide gives the number of slides to scrape. Following model development, an internal validation trial was conducted within the Tempus Labs molecular sequencing laboratory. We evaluated our system on 501 clinical colorectal cancer slides, where half received SmartPath-augmented review and half traditional pathologist review. The SmartPath cohort had 25% more DNA yields within a desired target range of 100-2000ng. The SmartPath system recommended fewer slides to scrape for large tissue sections, saving tissue in these cases. Conversely, SmartPath recommended more slides to scrape for samples with scant tissue sections, helping prevent costly re-extraction due to insufficient extraction yield. A statistical analysis was performed to measure the impact of covariates on the results, offering insights on how to improve future applications of SmartPath. Overall, the study demonstrated that AI-augmented histopathologic review using SmartPath could decrease tissue waste, sequencing time, and laboratory costs by optimizing DNA yields and tumor purity.

Kshitij Ingale, Sun Hae Hong, Qiyuan Hu et al.

Molecular testing of tumor samples for targetable biomarkers is restricted by a lack of standardization, turnaround-time, cost, and tissue availability across cancer types. Additionally, targetable alterations of low prevalence may not be tested in routine workflows. Algorithms that predict DNA alterations from routinely generated hematoxylin and eosin (H&E)-stained images could prioritize samples for confirmatory molecular testing. Costs and the necessity of a large number of samples containing mutations limit approaches that train individual algorithms for each alteration. In this work, models were trained for simultaneous prediction of multiple DNA alterations from H&E images using a multi-task approach. Compared to biomarker-specific models, this approach performed better on average, with pronounced gains for rare mutations. The models reasonably generalized to independent temporal-holdout, externally-stained, and multi-site TCGA test sets. Additionally, whole slide image embeddings derived using multi-task models demonstrated strong performance in downstream tasks that were not a part of training. Overall, this is a promising approach to develop clinically useful algorithms that provide multiple actionable predictions from a single slide.

Will E. Thompson, David M. Vidmar, Jessica K. De Freitas et al.

Identifying disease phenotypes from electronic health records (EHRs) is critical for numerous secondary uses. Manually encoding physician knowledge into rules is particularly challenging for rare diseases due to inadequate EHR coding, necessitating review of clinical notes. Large language models (LLMs) offer promise in text understanding but may not efficiently handle real-world clinical documentation. We propose a zero-shot LLM-based method enriched by retrieval-augmented generation and MapReduce, which pre-identifies disease-related text snippets to be used in parallel as queries for the LLM to establish diagnosis. We show that this method as applied to pulmonary hypertension (PH), a rare disease characterized by elevated arterial pressures in the lungs, significantly outperforms physician logic rules ($F_1$ score of 0.62 vs. 0.75). This method has the potential to enhance rare disease cohort identification, expanding the scope of robust clinical research and care gap identification.

Nathaniel Braman, Jacob W. H. Gordon, Emery T. Goossens et al.

Clinical decision-making in oncology involves multimodal data such as radiology scans, molecular profiling, histopathology slides, and clinical factors. Despite the importance of these modalities individually, no deep learning framework to date has combined them all to predict patient prognosis. Here, we predict the overall survival (OS) of glioma patients from diverse multimodal data with a Deep Orthogonal Fusion (DOF) model. The model learns to combine information from multiparametric MRI exams, biopsy-based modalities (such as H&E slide images and/or DNA sequencing), and clinical variables into a comprehensive multimodal risk score. Prognostic embeddings from each modality are learned and combined via attention-gated tensor fusion. To maximize the information gleaned from each modality, we introduce a multimodal orthogonalization (MMO) loss term that increases model performance by incentivizing constituent embeddings to be more complementary. DOF predicts OS in glioma patients with a median C-index of 0.788 +/- 0.067, significantly outperforming (p=0.023) the best performing unimodal model with a median C-index of 0.718 +/- 0.064. The prognostic model significantly stratifies glioma patients by OS within clinical subsets, adding further granularity to prognostic clinical grading and molecular subtyping.

Ellery Wulczyn, Kunal Nagpal, Matthew Symonds et al.

Gleason grading of prostate cancer is an important prognostic factor but suffers from poor reproducibility, particularly among non-subspecialist pathologists. Although artificial intelligence (A.I.) tools have demonstrated Gleason grading on-par with expert pathologists, it remains an open question whether A.I. grading translates to better prognostication. In this study, we developed a system to predict prostate-cancer specific mortality via A.I.-based Gleason grading and subsequently evaluated its ability to risk-stratify patients on an independent retrospective cohort of 2,807 prostatectomy cases from a single European center with 5-25 years of follow-up (median: 13, interquartile range 9-17). The A.I.'s risk scores produced a C-index of 0.84 (95%CI 0.80-0.87) for prostate cancer-specific mortality. Upon discretizing these risk scores into risk groups analogous to pathologist Grade Groups (GG), the A.I. had a C-index of 0.82 (95%CI 0.78-0.85). On the subset of cases with a GG in the original pathology report (n=1,517), the A.I.'s C-indices were 0.87 and 0.85 for continuous and discrete grading, respectively, compared to 0.79 (95%CI 0.71-0.86) for GG obtained from the reports. These represent improvements of 0.08 (95%CI 0.01-0.15) and 0.07 (95%CI 0.00-0.14) respectively. Our results suggest that A.I.-based Gleason grading can lead to effective risk-stratification and warrants further evaluation for improving disease management.

Ellery Wulczyn, David F. Steiner, Melissa Moran et al.

Deriving interpretable prognostic features from deep-learning-based prognostic histopathology models remains a challenge. In this study, we developed a deep learning system (DLS) for predicting disease specific survival for stage II and III colorectal cancer using 3,652 cases (27,300 slides). When evaluated on two validation datasets containing 1,239 cases (9,340 slides) and 738 cases (7,140 slides) respectively, the DLS achieved a 5-year disease-specific survival AUC of 0.70 (95%CI 0.66-0.73) and 0.69 (95%CI 0.64-0.72), and added significant predictive value to a set of 9 clinicopathologic features. To interpret the DLS, we explored the ability of different human-interpretable features to explain the variance in DLS scores. We observed that clinicopathologic features such as T-category, N-category, and grade explained a small fraction of the variance in DLS scores (R2=18% in both validation sets). Next, we generated human-interpretable histologic features by clustering embeddings from a deep-learning based image-similarity model and showed that they explain the majority of the variance (R2 of 73% to 80%). Furthermore, the clustering-derived feature most strongly associated with high DLS scores was also highly prognostic in isolation. With a distinct visual appearance (poorly differentiated tumor cell clusters adjacent to adipose tissue), this feature was identified by annotators with 87.0-95.5% accuracy. Our approach can be used to explain predictions from a prognostic deep learning model and uncover potentially-novel prognostic features that can be reliably identified by people for future validation studies.

Ellery Wulczyn, David F. Steiner, Zhaoyang Xu et al.

Prognostic information at diagnosis has important implications for cancer treatment and monitoring. Although cancer staging, histopathological assessment, molecular features, and clinical variables can provide useful prognostic insights, improving risk stratification remains an active research area. We developed a deep learning system (DLS) to predict disease specific survival across 10 cancer types from The Cancer Genome Atlas (TCGA). We used a weakly-supervised approach without pixel-level annotations, and tested three different survival loss functions. The DLS was developed using 9,086 slides from 3,664 cases and evaluated using 3,009 slides from 1,216 cases. In multivariable Cox regression analysis of the combined cohort including all 10 cancers, the DLS was significantly associated with disease specific survival (hazard ratio of 1.58, 95% CI 1.28-1.70, p<0.0001) after adjusting for cancer type, stage, age, and sex. In a per-cancer adjusted subanalysis, the DLS remained a significant predictor of survival in 5 of 10 cancer types. Compared to a baseline model including stage, age, and sex, the c-index of the model demonstrated an absolute 3.7% improvement (95% CI 1.0-6.5) in the combined cohort. Additionally, our models stratified patients within individual cancer stages, particularly stage II (p=0.025) and stage III (p<0.001). By developing and evaluating prognostic models across multiple cancer types, this work represents one of the most comprehensive studies exploring the direct prediction of clinical outcomes using deep learning and histopathology images. Our analysis demonstrates the potential for this approach to provide prognostic information in multiple cancer types, and even within specific pathologic stages. However, given the relatively small number of clinical events, we observed wide confidence intervals, suggesting that future work will benefit from larger datasets.

Carrie J. Cai, Emily Reif, Narayan Hegde et al.

Machine learning (ML) is increasingly being used in image retrieval systems for medical decision making. One application of ML is to retrieve visually similar medical images from past patients (e.g. tissue from biopsies) to reference when making a medical decision with a new patient. However, no algorithm can perfectly capture an expert's ideal notion of similarity for every case: an image that is algorithmically determined to be similar may not be medically relevant to a doctor's specific diagnostic needs. In this paper, we identified the needs of pathologists when searching for similar images retrieved using a deep learning algorithm, and developed tools that empower users to cope with the search algorithm on-the-fly, communicating what types of similarity are most important at different moments in time. In two evaluations with pathologists, we found that these refinement tools increased the diagnostic utility of images found and increased user trust in the algorithm. The tools were preferred over a traditional interface, without a loss in diagnostic accuracy. We also observed that users adopted new strategies when using refinement tools, re-purposing them to test and understand the underlying algorithm and to disambiguate ML errors from their own errors. Taken together, these findings inform future human-ML collaborative systems for expert decision-making.

Narayan Hegde, Jason D. Hipp, Yun Liu et al.

The increasing availability of large institutional and public histopathology image datasets is enabling the searching of these datasets for diagnosis, research, and education. Though these datasets typically have associated metadata such as diagnosis or clinical notes, even carefully curated datasets rarely contain annotations of the location of regions of interest on each image. Because pathology images are extremely large (up to 100,000 pixels in each dimension), further laborious visual search of each image may be needed to find the feature of interest. In this paper, we introduce a deep learning based reverse image search tool for histopathology images: Similar Medical Images Like Yours (SMILY). We assessed SMILY's ability to retrieve search results in two ways: using pathologist-provided annotations, and via prospective studies where pathologists evaluated the quality of SMILY search results. As a negative control in the second evaluation, pathologists were blinded to whether search results were retrieved by SMILY or randomly. In both types of assessments, SMILY was able to retrieve search results with similar histologic features, organ site, and prostate cancer Gleason grade compared with the original query. SMILY may be a useful general-purpose tool in the pathologist's arsenal, to improve the efficiency of searching large archives of histopathology images, without the need to develop and implement specific tools for each application.

Timo Kohlberger, Yun Liu, Melissa Moran et al.

Digital pathology enables remote access or consults and powerful image analysis algorithms. However, the slide digitization process can create artifacts such as out-of-focus (OOF). OOF is often only detected upon careful review, potentially causing rescanning and workflow delays. Although scan-time operator screening for whole-slide OOF is feasible, manual screening for OOF affecting only parts of a slide is impractical. We developed a convolutional neural network (ConvFocus) to exhaustively localize and quantify the severity of OOF regions on digitized slides. ConvFocus was developed using our refined semi-synthetic OOF data generation process, and evaluated using real whole-slide images spanning 3 different tissue types and 3 different stain types that were digitized by two different scanners. ConvFocus's predictions were compared with pathologist-annotated focus quality grades across 514 distinct regions representing 37,700 35x35 $μ$m image patches, and 21 digitized "z-stack" whole-slide images that contain known OOF patterns. When compared to pathologist-graded focus quality, ConvFocus achieved Spearman rank coefficients of 0.81 and 0.94 on two scanners, and reproduced the expected OOF patterns from z-stack scanning. We also evaluated the impact of OOF on the accuracy of a state-of-the-art metastatic breast cancer detector and saw a consistent decrease in performance with increasing OOF. Comprehensive whole-slide OOF categorization could enable rescans prior to pathologist review, potentially reducing the impact of digitization focus issues on the clinical workflow. We show that the algorithm trained on our semi-synthetic OOF data generalizes well to real OOF regions across tissue types, stains, and scanners. Finally, quantitative OOF maps can flag regions that might otherwise be misclassified by image analysis algorithms, preventing OOF-induced errors.

Po-Hsuan Cameron Chen, Krishna Gadepalli, Robert MacDonald et al.

The brightfield microscope is instrumental in the visual examination of both biological and physical samples at sub-millimeter scales. One key clinical application has been in cancer histopathology, where the microscopic assessment of the tissue samples is used for the diagnosis and staging of cancer and thus guides clinical therapy. However, the interpretation of these samples is inherently subjective, resulting in significant diagnostic variability. Moreover, in many regions of the world, access to pathologists is severely limited due to lack of trained personnel. In this regard, Artificial Intelligence (AI) based tools promise to improve the access and quality of healthcare. However, despite significant advances in AI research, integration of these tools into real-world cancer diagnosis workflows remains challenging because of the costs of image digitization and difficulties in deploying AI solutions. Here we propose a cost-effective solution to the integration of AI: the Augmented Reality Microscope (ARM). The ARM overlays AI-based information onto the current view of the sample through the optical pathway in real-time, enabling seamless integration of AI into the regular microscopy workflow. We demonstrate the utility of ARM in the detection of lymph node metastases in breast cancer and the identification of prostate cancer with a latency that supports real-time workflows. We anticipate that ARM will remove barriers towards the use of AI in microscopic analysis and thus improve the accuracy and efficiency of cancer diagnosis. This approach is applicable to other microscopy tasks and AI algorithms in the life sciences and beyond.

Kunal Nagpal, Davis Foote, Yun Liu et al.

For prostate cancer patients, the Gleason score is one of the most important prognostic factors, potentially determining treatment independent of the stage. However, Gleason scoring is based on subjective microscopic examination of tumor morphology and suffers from poor reproducibility. Here we present a deep learning system (DLS) for Gleason scoring whole-slide images of prostatectomies. Our system was developed using 112 million pathologist-annotated image patches from 1,226 slides, and evaluated on an independent validation dataset of 331 slides, where the reference standard was established by genitourinary specialist pathologists. On the validation dataset, the mean accuracy among 29 general pathologists was 0.61. The DLS achieved a significantly higher diagnostic accuracy of 0.70 (p=0.002) and trended towards better patient risk stratification in correlations to clinical follow-up data. Our approach could improve the accuracy of Gleason scoring and subsequent therapy decisions, particularly where specialist expertise is unavailable. The DLS also goes beyond the current Gleason system to more finely characterize and quantitate tumor morphology, providing opportunities for refinement of the Gleason system itself.

Yun Liu, Krishna Gadepalli, Mohammad Norouzi et al.

Each year, the treatment decisions for more than 230,000 breast cancer patients in the U.S. hinge on whether the cancer has metastasized away from the breast. Metastasis detection is currently performed by pathologists reviewing large expanses of biological tissues. This process is labor intensive and error-prone. We present a framework to automatically detect and localize tumors as small as 100 x 100 pixels in gigapixel microscopy images sized 100,000 x 100,000 pixels. Our method leverages a convolutional neural network (CNN) architecture and obtains state-of-the-art results on the Camelyon16 dataset in the challenging lesion-level tumor detection task. At 8 false positives per image, we detect 92.4% of the tumors, relative to 82.7% by the previous best automated approach. For comparison, a human pathologist attempting exhaustive search achieved 73.2% sensitivity. We achieve image-level AUC scores above 97% on both the Camelyon16 test set and an independent set of 110 slides. In addition, we discover that two slides in the Camelyon16 training set were erroneously labeled normal. Our approach could considerably reduce false negative rates in metastasis detection.

Qian Yu, Christian Szegedy, Martin C. Stumpe et al.

Search with local intent is becoming increasingly useful due to the popularity of the mobile device. The creation and maintenance of accurate listings of local businesses worldwide is time consuming and expensive. In this paper, we propose an approach to automatically discover businesses that are visible on street level imagery. Precise business store front detection enables accurate geo-location of businesses, and further provides input for business categorization, listing generation, etc. The large variety of business categories in different countries makes this a very challenging problem. Moreover, manual annotation is prohibitive due to the scale of this problem. We propose the use of a MultiBox based approach that takes input image pixels and directly outputs store front bounding boxes. This end-to-end learning approach instead preempts the need for hand modeling either the proposal generation phase or the post-processing phase, leveraging large labelled training datasets. We demonstrate our approach outperforms the state of the art detection techniques with a large margin in terms of performance and run-time efficiency. In the evaluation, we show this approach achieves human accuracy in the low-recall settings. We also provide an end-to-end evaluation of business discovery in the real world.