Developing a low-cost, open-source, locally manufactured workstation and computational pipeline for automated histopathology evaluation using deep learning.

BACKGROUND: Deployment and access to state-of-the-art precision medicine technologies remains a fundamental challenge in providing equitable global cancer care in low-resource settings. The expansion of digital pathology in recent years and its potential interface with diagnostic artificial intelligence algorithms provides an opportunity to democratize access to personalized medicine. Current digital pathology workstations, however, cost thousands to hundreds of thousands of dollars. As cancer incidence rises in many low- and middle-income countries, the validation and implementation of low-cost automated diagnostic tools will be crucial to helping healthcare providers manage the growing burden of cancer. METHODS: Here we describe a low-cost ($230) workstation for digital slide capture and computational analysis composed of open-source components. We analyze the predictive performance of deep learning models when they are used to evaluate pathology images captured using this open-source workstation versus images captured using common, significantly more expensive hardware. Validation studies assessed model performance on three distinct datasets and predictive models: head and neck squamous cell carcinoma (HPV positive versus HPV negative), lung cancer (adenocarcinoma versus squamous cell carcinoma), and breast cancer (invasive ductal carcinoma versus invasive lobular carcinoma). FINDINGS: When compared to traditional pathology image capture methods, low-cost digital slide capture and analysis with the open-source workstation, including the low-cost microscope device, was associated with model performance of comparable accuracy for breast, lung, and HNSCC classification. At the patient level of analysis, AUROC was 0.84 for HNSCC HPV status prediction, 1.0 for lung cancer subtype prediction, and 0.80 for breast cancer classification. INTERPRETATION: Our ability to maintain model performance despite decreased image quality and low-power computational hardware demonstrates that it is feasible to massively reduce costs associated with deploying deep learning models for digital pathology applications. Improving access to cutting-edge diagnostic tools may provide an avenue for reducing disparities in cancer care between high- and low-income regions. FUNDING: Funding for this project including personnel support was provided via grants from NIH/NCIR25-CA240134, NIH/NCIU01-CA243075, NIH/NIDCRR56-DE030958, NIH/NCIR01-CA276652, NIH/NCIK08-CA283261, NIH/NCI-SOAR25CA240134, SU2C (Stand Up to Cancer) Fanconi Anemia Research Fund - Farrah Fawcett Foundation Head and Neck Cancer Research Team Grant, and the European UnionHorizon Program (I3LUNG).

Carcinoma-associated mesenchymal stem cells promote ovarian cancer heterogeneity and metastasis through mitochondrial transfer.

Ovarian cancer is characterized by early metastatic spread. This study demonstrates that carcinoma-associated mesenchymal stromal cells (CA-MSCs) enhance metastasis by increasing tumor cell heterogeneity through mitochondrial donation. CA-MSC mitochondrial donation preferentially occurs in ovarian cancer cells with low levels of mitochondria ("mito poor"). CA-MSC mitochondrial donation rescues the phenotype of mito poor cells, restoring their proliferative capacity, resistance to chemotherapy, and cellular respiration. Receipt of CA-MSC-derived mitochondria induces tumor cell transcriptional changes leading to the secretion of ANGPTL3, which enhances the proliferation of tumor cells without CA-MSC mitochondria, thus amplifying the impact of mitochondrial transfer. Donated CA-MSC mitochondrial DNA persisted in recipient tumor cells for at least 14 days. CA-MSC mitochondrial donation occurs in vivo, enhancing tumor cell heterogeneity and decreasing mouse survival. Collectively, this work identifies CA-MSC mitochondrial transfer as a critical mediator of ovarian cancer cell survival, heterogeneity, and metastasis and presents a unique therapeutic target in ovarian cancer.

Artificial intelligence-based epigenomic, transcriptomic and histologic signatures of tobacco use in oral squamous cell carcinoma.

Oral squamous cell carcinoma (OSCC) biomarker studies rarely employ multi-omic biomarker strategies and pertinent clinicopathologic characteristics to predict mortality. In this study we determine for the first time a combined epigenetic, gene expression, and histology signature that differentiates between patients with different tobacco use history (heavy tobacco use with ≥10 pack years vs. no tobacco use). Using The Cancer Genome Atlas (TCGA) cohort (n = 257) and an internal cohort (n = 40), we identify 3 epigenetic markers (GPR15, GNG12, GDNF) and 13 expression markers (IGHA2, SCG5, RPL3L, NTRK1, CD96, BMP6, TFPI2, EFEMP2, RYR3, DMTN, GPD2, BAALC, and FMO3), which are dysregulated in OSCC patients who were never smokers vs. those who have a ≥ 10 pack year history. While mortality risk prediction based on smoking status and clinicopathologic covariates alone is inaccurate (c-statistic = 0.57), the combined epigenetic/expression and histologic signature has a c-statistic = 0.9409 in predicting 5-year mortality in OSCC patients.

Artificial intelligence-based morphologic classification and molecular characterization of neuroblastic tumors from digital histopathology.

A deep learning model using attention-based multiple instance learning (aMIL) and self-supervised learning (SSL) was developed to perform pathologic classification of neuroblastic tumors and assess MYCN-amplification status using H&E-stained whole slide digital images. The model demonstrated strong performance in identifying diagnostic category, grade, mitosis-karyorrhexis index (MKI), and MYCN-amplification on an external test dataset. This AI-based approach establishes a valuable tool for automating diagnosis and precise classification of neuroblastoma tumors.

Utility of the Endoscopic Healing Index in Identifying Active Inflammation in Patients with Crohn's Disease: Real World Data from a Tertiary Center.

BACKGROUND: The Endoscopic Healing Index (EHI) analyzes biomarkers in a patient's peripheral blood to assess mucosal healing. We aimed to characterize the effectiveness of the EHI as a predictor of disease activity in a real world clinical setting. METHODS: This retrospective study looked at patients treated and followed up at the University of Chicago Medicine IBD center who had EHI tests done as part of routine clinical care. The results of the EHI were compared with radiological imaging or endoscopy performed within 3 months of the EHI in order to determine accuracy at diagnosing active inflammation. RESULTS: Fifty-five patients with CD and with an available EHI were included in this study. Four (50%) patients with an EHI of < 20 (n = 8) had evidence of objective inflammation. A cutoff of ≤ 20 had a sensitivity of 89% and specificity of 23.5% for predicting no evidence of any objective inflammation with an AUROC of 0.69. This score had a negative predictive value (NPV) of 50% and positive predictive value (PPV) of 72.3%. A cutoff EHI of 30 tended to classify patients as either having objective evidence of inflammation or not more often than FCAL (Correctly classifying inflammation: 89% vs 64%, respectively; p = 0.32). CONCLUSION: In this real world analysis, the EHI showed poor predictive value for the absence of active inflammation as assessed by imaging or endoscopy, has limited utility in confirming deep remission and should be used with another objective modality.

Acquired resistance to immunotherapy and chemoradiation in MYC amplified head and neck cancer.

The proto-oncogene MYC encodes a nuclear transcription factor that has an important role in a variety of cellular processes, such as cell cycle progression, proliferation, metabolism, adhesion, apoptosis, and therapeutic resistance. MYC amplification is consistently observed in aggressive forms of several solid malignancies and correlates with poor prognosis and distant metastases. While the tumorigenic effects of MYC in patients with head and neck squamous cell carcinoma (HNSCC) are well known, the molecular mechanisms by which the amplification of this gene may confer treatment resistance, especially to immune checkpoint inhibitors, remains under-investigated. Here we present a unique case of a patient with recurrent/metastatic (R/M) HNSCC who, despite initial response to nivolumab-based treatment, developed rapidly progressive metastatic disease after the acquisition of MYC amplification. We conducted comparative transcriptomic analysis of this patient's tumor at baseline and upon progression to interrogate potential molecular processes through which MYC may confer resistance to immunotherapy and/or chemoradiation and used TCGA-HNSC dataset and an institutional cohort to further explore clinicopathologic features and key molecular networks associated with MYC amplification in HNSCC. This study highlights MYC amplification as a potential mechanism of immune checkpoint inhibitor resistance and suggest its use as a predictive biomarker and potential therapeutic target in R/M HNSCC.

Demographic bias in misdiagnosis by computational pathology models.

Despite increasing numbers of regulatory approvals, deep learning-based computational pathology systems often overlook the impact of demographic factors on performance, potentially leading to biases. This concern is all the more important as computational pathology has leveraged large public datasets that underrepresent certain demographic groups. Using publicly available data from The Cancer Genome Atlas and the EBRAINS brain tumor atlas, as well as internal patient data, we show that whole-slide image classification models display marked performance disparities across different demographic groups when used to subtype breast and lung carcinomas and to predict IDH1 mutations in gliomas. For example, when using common modeling approaches, we observed performance gaps (in area under the receiver operating characteristic curve) between white and Black patients of 3.0% for breast cancer subtyping, 10.9% for lung cancer subtyping and 16.0% for IDH1 mutation prediction in gliomas. We found that richer feature representations obtained from self-supervised vision foundation models reduce performance variations between groups. These representations provide improvements upon weaker models even when those weaker models are combined with state-of-the-art bias mitigation strategies and modeling choices. Nevertheless, self-supervised vision foundation models do not fully eliminate these discrepancies, highlighting the continuing need for bias mitigation efforts in computational pathology. Finally, we demonstrate that our results extend to other demographic factors beyond patient race. Given these findings, we encourage regulatory and policy agencies to integrate demographic-stratified evaluation into their assessment guidelines.

Slideflow: deep learning for digital histopathology with real-time whole-slide visualization.

Deep learning methods have emerged as powerful tools for analyzing histopathological images, but current methods are often specialized for specific domains and software environments, and few open-source options exist for deploying models in an interactive interface. Experimenting with different deep learning approaches typically requires switching software libraries and reprocessing data, reducing the feasibility and practicality of experimenting with new architectures. We developed a flexible deep learning library for histopathology called Slideflow, a package which supports a broad array of deep learning methods for digital pathology and includes a fast whole-slide interface for deploying trained models. Slideflow includes unique tools for whole-slide image data processing, efficient stain normalization and augmentation, weakly-supervised whole-slide classification, uncertainty quantification, feature generation, feature space analysis, and explainability. Whole-slide image processing is highly optimized, enabling whole-slide tile extraction at 40x magnification in 2.5 s per slide. The framework-agnostic data processing pipeline enables rapid experimentation with new methods built with either Tensorflow or PyTorch, and the graphical user interface supports real-time visualization of slides, predictions, heatmaps, and feature space characteristics on a variety of hardware devices, including ARM-based devices such as the Raspberry Pi.

Comparing scientific abstracts generated by ChatGPT to real abstracts with detectors and blinded human reviewers.

Large language models such as ChatGPT can produce increasingly realistic text, with unknown information on the accuracy and integrity of using these models in scientific writing. We gathered fifth research abstracts from five high-impact factor medical journals and asked ChatGPT to generate research abstracts based on their titles and journals. Most generated abstracts were detected using an AI output detector, 'GPT-2 Output Detector', with % 'fake' scores (higher meaning more likely to be generated) of median [interquartile range] of 99.98% 'fake' [12.73%, 99.98%] compared with median 0.02% [IQR 0.02%, 0.09%] for the original abstracts. The AUROC of the AI output detector was 0.94. Generated abstracts scored lower than original abstracts when run through a plagiarism detector website and iThenticate (higher scores meaning more matching text found). When given a mixture of original and general abstracts, blinded human reviewers correctly identified 68% of generated abstracts as being generated by ChatGPT, but incorrectly identified 14% of original abstracts as being generated. Reviewers indicated that it was surprisingly difficult to differentiate between the two, though abstracts they suspected were generated were vaguer and more formulaic. ChatGPT writes believable scientific abstracts, though with completely generated data. Depending on publisher-specific guidelines, AI output detectors may serve as an editorial tool to help maintain scientific standards. The boundaries of ethical and acceptable use of large language models to help scientific writing are still being discussed, and different journals and conferences are adopting varying policies.

Fifteen-minute consultation: Recognition and management of eating disorders presenting to the emergency department.

Eating disorder presentations in children and young people during the COVID-19 pandemic have increased, and this has become a common presentation to paediatric emergency departments (EDs). We cover a structured approach on identifying and managing these presentations within the ED including history taking, what to look for on examination, what investigations are needed and how to decide who requires admission to hospital.

Diagnostic and Management Considerations for the IPAA With Crohn's Disease-Like Features.

BACKGROUND: Patients with ulcerative colitis often develop medically refractory colonic inflammation or colorectal neoplasia, and approximately 10% to 15% of patients require surgery. The most common surgical procedure is a restorative proctocolectomy with IPAA. Even if the preoperative diagnosis is ulcerative colitis, approximately 10% of patients can develop inflammatory pouch conditions resembling a Crohn's disease phenotype. OBJECTIVE: This study aimed to review the diagnostic approach, prognosis, and management of IPAA with Crohn's disease-like features. DATA SOURCES: The data sources include search in electronic databases. STUDY SELECTION: This narrative review included studies focusing on pouches with Crohn's disease-like features. MAIN OUTCOME MEASURES: The main topics in this review included the pathogenesis, risk factors, diagnosis, phenotypes, prognosis, and medications of pouches with Crohn's disease-like features. RESULTS: A diagnostic approach for the pouch conditions resembling a Crohn's disease phenotype should be based on history-taking to evaluate its risk factors and endoscopic assessment of the pouch. Prior disease history and pathology, location of pouch complications, and timing of complications offer clues for the differential diagnosis of this phenotype. We advocate for the more descriptive term "pouch with Crohn's disease-like features" and reserve the term "Crohn's disease of the pouch" for patients who undergo IPAA and have a precolectomy diagnosis of Crohn's disease or whose colectomy pathology revealed Crohn's disease. Medications, which are often used for traditional Crohn's disease, show efficacy in pouches with Crohn's disease-like features as well. The poor prognosis associated with pouches with Crohn's disease-like features, particularly the fistulizing phenotype, underscores the importance of proactive monitoring and therapeutic intervention. LIMITATIONS: The limitations include no explicit criteria for article selection. CONCLUSIONS: This review suggests future research should seek to understand the natural history and meaningful shorter and longer term therapeutic targets for these types of pouch phenotypes. Long-term follow-up and prospective preoperative and postoperative interventional trials of treatments and prevention strategies are needed.

IL-23 Monoclonal Antibodies for IBD: So Many, So Different?

Interleukin [IL]-23 is a member of the IL-12 family of cytokines and has been implicated in multiple inflammatory disorders including psoriasis, psoriatic arthritis, and the inflammatory bowel diseases [IBDs]. Blockade of both IL-12 and IL-23 using an antibody that targets a shared subunit is highly effective in treating psoriasis, and recent data suggest similar efficacy in IBD with minimal adverse events. In this review, we summarise published data on the efficacy of anti-IL-12/23 therapies in IBD as well as emerging data on more selective anti-IL-23 specific therapies. Last, we discuss novel therapeutics under development which target the IL-23 pathway in unique ways and suggest that a biomarker-driven approach will soon guide clinicians to prescribe anti-IL-23 therapies to the patients most likely to respond to them.

Factors Associated With Fecal Calprotectin Sample Collection Compliance: An IBD Center Quality Improvement Project.

Background: Fecal calprotectin (Fcal) is a noninvasive, inexpensive biomarker of disease activity. However, patient compliance with this test is variable and incompletely described. We assessed compliance rates with Fcal tests and identified factors associated with noncompliance. Methods: A retrospective chart review of patients with inflammatory bowel disease (IBD) who had a Fcal test ordered through our center between August 2021 and December 2021 was conducted. Demographic, clinical, disease, and test-related information were recorded. Patients with incomplete Fcal orders were sent a survey to better understand their reasons for noncompliance. Simple statistical analysis and and multivariable logistic regression modeling were performed. Results: Of 303 patients, 165 (54.4%) had an order for Fcal. Of the Fcal tests ordered, 55 (33.3%) were not completed. Remission of IBD, no prior Fcal completion, and tests ordered at a distant site were all associated with test noncompletion. A multivariable logistic regression revealed that history of a prior completed Fcal test is associated with subsequent test completion (odds ratio = 2.1, 95% confidence interval 1.9-35.5, P = .004). Patients who did not complete the test described the pandemic and third-party testing center issues as the most common reasons for noncompliance. Conclusions: In this single center experience with Fcal testing in patients with IBD, we identified that a history of incomplete Fcal testing and distant location of lab testing were significantly associated with noncompletion of the test. We provide practical guidance for future utilization and compliance, including the impact of home-based testing.

FmhA and FmhC of Staphylococcus aureus incorporate serine residues into peptidoglycan cross-bridges.

Staphylococcal peptidoglycan is characterized by pentaglycine cross-bridges that are cross-linked between adjacent wall peptides by penicillin-binding proteins to confer robustness and flexibility. In Staphylococcus aureus, pentaglycine cross-bridges are synthesized by three proteins: FemX adds the first glycine, and the homodimers FemA and FemB sequentially add two Gly-Gly dipeptides. Occasionally, serine residues are also incorporated into the cross-bridges by enzymes that have heretofore not been identified. Here, we show that the FemA/FemB homologues FmhA and FmhC pair with FemA and FemB to incorporate Gly-Ser dipeptides into cross-bridges and to confer resistance to lysostaphin, a secreted bacteriocin that cleaves the pentaglycine cross-bridge. FmhA incorporates serine residues at positions 3 and 5 of the cross-bridge. In contrast, FmhC incorporates a single serine at position 5. Serine incorporation also lowers resistance toward oxacillin, an antibiotic that targets penicillin-binding proteins, in both methicillin-sensitive and methicillin-resistant strains of S. aureus FmhC is encoded by a gene immediately adjacent to lytN, which specifies a hydrolase that cleaves the bond between the fifth glycine of cross-bridges and the alanine of the adjacent stem peptide. In this manner, LytN facilitates the separation of daughter cells. Cell wall damage induced upon lytN overexpression can be alleviated by overexpression of fmhC. Together, these observations suggest that FmhA and FmhC generate peptidoglycan cross-bridges with unique serine patterns that provide protection from endogenous murein hydrolases governing cell division and from bacteriocins produced by microbial competitors.

Donor hearts in the Sydney Heart Bank: reliable control but is it 'normal' heart?

Human heart samples from the Sydney Heart Bank have become a de facto standard against which others can be measured. Crucially, the heart bank contains a lot of donor heart material: for most researchers this is the hardest to obtain and yet is necessary since we can only study the pathological human heart in comparison with a control, preferably a normal heart sample. It is not generally realised how important the control is for human heart studies. We review our studies on donor heart samples. We report the results obtained with 17 different donor samples collected from 1994 to 2011 and measured from 2005 to 2015 by our standard methodology for in vitro motility and troponin I phosphorylation measurements. The donor heart sample parameters are consistent between the hearts, over time and with different operators indicating that Sydney Heart Bank donor hearts are a valid baseline control for comparison with pathological heart samples. We also discuss to what extent donor heart samples are representative of the normal heart.

How to interpret urine toxicology tests.

A 13-year-old girl presents to the emergency department for the second time with an unresponsive episode. She has a GCS (Glasgow Coma Scale) score of 11 on arrival and all other observations are normal. The story is unclear, but there are ongoing safeguarding concerns and the family are known to social services. All investigations are normal. After a period of observation on the ward, her GCS returns to normal and she appears well. Both on the first presentation and this presentation ingestion of a toxin was suspected. However, this was denied by the patient and urine toxicology screen was negative. Does this rule out toxin ingestion? Will this change your management?

How to use C-reactive protein.

A 3-month-old baby is brought to the paediatric emergency department by their parents because of a fever. You decide to check their inflammatory markers. Their C-reactive protein (CRP) level comes back as 20 mg/L. Does this affect whether or not you start antibiotic therapy? Does it influence your decision to admit or discharge the patient? CRP is a commonly used biochemical test and yet its use is constantly debated and challenged. We look at the current evidence and suggest the best way to use this test in clinical practice.

Problem-solving in clinical practice: a baby who won't stop bleeding.

Spontaneous bleeding in the neonatal period is an unusual presentation, and yet one that can pose a challenge of both diagnosis and management to the general paediatrician. This case chronicles the diagnostic journey of an 8-day-old baby who presented with unrelenting bleeding from the umbilical cord and explores the clinical approach to bleeding in a neonate.

Fifteen-minute consultation: Monogenic forms of diabetes with onset after age 6 months.

Monogenic forms of diabetes (historically known as Maturity Onset Diabetes of the Young (MODY)) are caused by single gene mutations inherited in an autosomal dominant fashion that result in reduced pancreatic beta cell function. Children with these forms of diabetes may be misdiagnosed as having type 1 or 2 diabetes, which has important implications for treatment, genetic counselling, screening of family members and prognosis. Useful tools now exist to aid in their diagnosis and management. Here, we attempt to outline the clinical features that will help the physician make the differentiation from other diabetes subtypes.