A comparative study of zero-shot inference with large language models and supervised modeling in breast cancer pathology classification.

Although supervised machine learning is popular for information extraction from clinical notes, creating large, annotated datasets requires extensive domain expertise and is time-consuming. Meanwhile, large language models (LLMs) have demonstrated promising transfer learning capability. In this study, we explored whether recent LLMs can reduce the need for large-scale data annotations. We curated a manually labeled dataset of 769 breast cancer pathology reports, labeled with 13 categories, to compare zero-shot classification capability of the GPT-4 model and the GPT-3.5 model with supervised classification performance of three model architectures: random forests classifier, long short-term memory networks with attention (LSTM-Att), and the UCSF-BERT model. Across all 13 tasks, the GPT-4 model performed either significantly better than or as well as the best supervised model, the LSTM-Att model (average macro F1 score of 0.83 vs. 0.75). On tasks with a high imbalance between labels, the differences were more prominent. Frequent sources of GPT-4 errors included inferences from multiple samples and complex task design. On complex tasks where large annotated datasets cannot be easily collected, LLMs can reduce the burden of large-scale data labeling. However, if the use of LLMs is prohibitive, the use of simpler supervised models with large annotated datasets can provide comparable results. LLMs demonstrated the potential to speed up the execution of clinical NLP studies by reducing the need for curating large annotated datasets. This may increase the utilization of NLP-based variables and outcomes in observational clinical studies.

Biological insights and novel biomarker discovery through deep learning approaches in breast cancer histopathology.

Breast cancer remains a highly prevalent disease with considerable inter- and intra-tumoral heterogeneity complicating prognostication and treatment decisions. The utilization and depth of genomic, transcriptomic and proteomic data for cancer has exploded over recent times and the addition of spatial context to this information, by understanding the correlating morphologic and spatial patterns of cells in tissue samples, has created an exciting frontier of research, histo-genomics. At the same time, deep learning (DL), a class of machine learning algorithms employing artificial neural networks, has rapidly progressed in the last decade with a confluence of technical developments - including the advent of modern graphic processing units (GPU), allowing efficient implementation of increasingly complex architectures at scale; advances in the theoretical and practical design of network architectures; and access to larger datasets for training - all leading to sweeping advances in image classification and object detection. In this review, we examine recent developments in the application of DL in breast cancer histology with particular emphasis of those producing biologic insights or novel biomarkers, spanning the extraction of genomic information to the use of stroma to predict cancer recurrence, with the aim of suggesting avenues for further advancing this exciting field.

A molecular guide to systemic therapy in salivary gland carcinoma.

Salivary gland carcinomas (SGC) are a rare and variable group of head and neck cancers with historically poor response to cytotoxic chemotherapy and immunotherapy in the recurrent, advanced, and metastatic settings. In the last decade, a number of targetable molecular alterations have been identified in SGCs including HER2 upregulation, androgen receptor overexpression, Notch receptor activation, NTRK gene fusions, and RET alterations which have dramatically improved treatment outcomes in this disease. Here, we review the landscape of precision therapy in SGC including current options for systemic management, ongoing clinical trials, and promising future directions.

The Impact of Time Horizon on Classification Accuracy: Application of Machine Learning to Prediction of Incident Coronary Heart Disease.

BACKGROUND: Many machine learning approaches are limited to classification of outcomes rather than longitudinal prediction. One strategy to use machine learning in clinical risk prediction is to classify outcomes over a given time horizon. However, it is not well-known how to identify the optimal time horizon for risk prediction. OBJECTIVE: In this study, we aim to identify an optimal time horizon for classification of incident myocardial infarction (MI) using machine learning approaches looped over outcomes with increasing time horizons. Additionally, we sought to compare the performance of these models with the traditional Framingham Heart Study (FHS) coronary heart disease gender-specific Cox proportional hazards regression model. METHODS: We analyzed data from a single clinic visit of 5201 participants of a cardiovascular health study. We examined 61 variables collected from this baseline exam, including demographic and biologic data, medical history, medications, serum biomarkers, electrocardiographic, and echocardiographic data. We compared several machine learning methods (eg, random forest, L1 regression, gradient boosted decision tree, support vector machine, and k-nearest neighbor) trained to predict incident MI that occurred within time horizons ranging from 500-10,000 days of follow-up. Models were compared on a 20% held-out testing set using area under the receiver operating characteristic curve (AUROC). Variable importance was performed for random forest and L1 regression models across time points. We compared results with the FHS coronary heart disease gender-specific Cox proportional hazards regression functions. RESULTS: There were 4190 participants included in the analysis, with 2522 (60.2%) female participants and an average age of 72.6 years. Over 10,000 days of follow-up, there were 813 incident MI events. The machine learning models were most predictive over moderate follow-up time horizons (ie, 1500-2500 days). Overall, the L1 (Lasso) logistic regression demonstrated the strongest classification accuracy across all time horizons. This model was most predictive at 1500 days follow-up, with an AUROC of 0.71. The most influential variables differed by follow-up time and model, with gender being the most important feature for the L1 regression and weight for the random forest model across all time frames. Compared with the Framingham Cox function, the L1 and random forest models performed better across all time frames beyond 1500 days. CONCLUSIONS: In a population free of coronary heart disease, machine learning techniques can be used to predict incident MI at varying time horizons with reasonable accuracy, with the strongest prediction accuracy in moderate follow-up periods. Validation across additional populations is needed to confirm the validity of this approach in risk prediction.

Orbital atherectomy of calcified coronary ostial lesions.

OBJECTIVES: To evaluate the feasibility and safety of coronary orbital atherectomy (OA) for the treatment of calcified ostial lesions. BACKGROUND: Percutaneous coronary intervention (PCI) is increasingly being completed in complex patients and lesions. OA is effective for severely calcified coronary lesions; however, there is a dearth of evidence on the use of OA in ostial lesions, especially with long-term outcome data. METHODS: Data were obtained from a retrospective analysis of patients who underwent OA of heavily calcified ostial lesions followed by stent implantation from December 2010 to June 2019 at two high-volume PCI centers. Kaplan-Meier analysis was utilized to assess the primary endpoints of 30-day, 1-year, and 2-year freedom-from (FF) major adverse cardiac events (MACE: death, myocardial infarction, or target vessel revascularization), stroke, and stent thrombosis (ST). RESULTS: A total of 56 patients underwent OA to treat heavily calcified ostial coronary lesions. The mean age was 72 years with a high prevalence of diabetes (55%) and heart failure (36%), requiring hemodynamic support (14%). There was high FF angiographic complications (93%), and at 30-day, 1-year, and 2-year, a high FF-MACE (96%, 91%, and 88%), stroke (98%, 96%, and 96%), and ST (100%), respectively. CONCLUSIONS: This study represents the largest real-world experience of coronary OA use in heavily calcified ostial lesions with long-term outcomes over 2 years. The main finding in this retrospective analysis is that, despite the complex patients and lesions included in this analysis, OA appears to be a feasible and safe treatment option for calcified coronary ostial lesions.

Quantifying The Medical Impact of A Missed Diagnosis of Non-Small Cell Lung Cancer on Chest Imaging.

If a chest lesion is noted to have been visible on imaging conducted prior to a definitive diagnosis of non-small cell lung cancer, medico-legal action directed against those considered to have missed the initial diagnosis may ensue. Evidence-based approaches to determine the medical impact of the resulting delay are limited. This article reviews strategies for quantifying the medical impact of missed diagnoses and identifies areas for future research. If no nodal or metastatic disease is present at the time of the definitive diagnosis, the potential impact of the delay is sometimes deduced from the differing 5-year overall survival rates of the T status-associated cancer stage at each time point. However, relapse-free survival, specific lung cancer subtype, time from diagnosis and the medical condition of the patient when the evaluation is being made may also have to be considered. In the absence of T-status change, medical impact from any delay is unlikely to be significant, although the effect of changes in patient fitness on outcomes, emotional distress and lost time for the patient's preparation may be argued. When nodal or metastatic involvement is noted at the time of definitive diagnosis, arguments may be made that these did not exist at the time of the missed diagnosis. However, more nuanced calculations considering differences in the risk of spread based on T-stage at each time point would be preferable. Large datasets to inform T to N-status correlations for such calculations already exist, but data to inform T to M-status correlations are limited.

EFFICACY OF INTRAVITREAL AFLIBERCEPT INJECTIONS IN THE TREATMENT OF IDIOPATHIC RETINAL VASCULITIS, ANEURYSMS, AND NEURORETINITIS SYNDROME.

PURPOSE: To present a case of idiopathic retinal vasculitis, aneurysms, and neuroretinitis syndrome that was successfully managed with serial intravitreal aflibercept injections. METHODS: Ophthalmic imaging and visual acuity were used to monitor disease state and track treatment methods to determine the most valuable combination of treatment medication and treatment interval. RESULTS: A 28-year-old woman with idiopathic retinal vasculitis, aneurysms, and neuroretinitis syndrome status after panretinal photocoagulation of both eyes presented with bilateral cystoid macular edema. We demonstrate successful management of retinal cystoid macular edema associated with idiopathic retinal vasculitis, aneurysms, and neuroretinitis syndrome using serial intravitreal aflibercept injections. CONCLUSION: Intravitreal aflibercept has a useful role in managing the potential retinal complications associated with idiopathic retinal vasculitis, aneurysms, and neuroretinitis syndrome and provides further insights into treatment of the later stages of this rare disease.

Prediction of Drug-Induced Long QT Syndrome Using Machine Learning Applied to Harmonized Electronic Health Record Data.

BACKGROUND: Drug-induced QT prolongation is a potentially preventable cause of morbidity and mortality, however there are no widespread clinical tools utilized to predict which individuals are at greatest risk. Machine learning (ML) algorithms may provide a method for identifying these individuals, and could be automated to directly alert providers in real time. OBJECTIVE: This study applies ML techniques to electronic health record (EHR) data to identify an integrated risk-prediction model that can be deployed to predict risk of drug-induced QT prolongation. METHODS: We examined harmonized data from the UCHealth EHR and identified inpatients who had received a medication known to prolong the QT interval. Using a binary outcome of the development of a QTc interval >500 ms within 24 hours of medication initiation or no ECG with a QTc interval >500 ms, we compared multiple machine learning methods by classification accuracy and performed calibration and rescaling of the final model. RESULTS: We identified 35,639 inpatients who received a known QT-prolonging medication and an ECG performed within 24 hours of administration. Of those, 4,558 patients developed a QTc > 500 ms and 31,081 patients did not. A deep neural network with random oversampling of controls was found to provide superior classification accuracy (F1 score 0.404; AUC 0.71) for the development of a long QT interval compared with other methods. The optimal cutpoint for prediction was determined and was reasonably accurate (sensitivity 71%; specificity 73%). CONCLUSIONS: We found that deep neural networks applied to EHR data provide reasonable prediction of which individuals are most susceptible to drug-induced QT prolongation. Future studies are needed to validate this model in novel EHRs and within the physician order entry system to assess the ability to improve patient safety.

Prediction of incident myocardial infarction using machine learning applied to harmonized electronic health record data.

BACKGROUND: With cardiovascular disease increasing, substantial research has focused on the development of prediction tools. We compare deep learning and machine learning models to a baseline logistic regression using only 'known' risk factors in predicting incident myocardial infarction (MI) from harmonized EHR data. METHODS: Large-scale case-control study with outcome of 6-month incident MI, conducted using the top 800, from an initial 52 k procedures, diagnoses, and medications within the UCHealth system, harmonized to the Observational Medical Outcomes Partnership common data model, performed on 2.27 million patients. We compared several over- and under- sampling techniques to address the imbalance in the dataset. We compared regularized logistics regression, random forest, boosted gradient machines, and shallow and deep neural networks. A baseline model for comparison was a logistic regression using a limited set of 'known' risk factors for MI. Hyper-parameters were identified using 10-fold cross-validation. RESULTS: Twenty thousand Five hundred and ninety-one patients were diagnosed with MI compared with 2.25 million who did not. A deep neural network with random undersampling provided superior classification compared with other methods. However, the benefit of the deep neural network was only moderate, showing an F1 Score of 0.092 and AUC of 0.835, compared to a logistic regression model using only 'known' risk factors. Calibration for all models was poor despite adequate discrimination, due to overfitting from low frequency of the event of interest. CONCLUSIONS: Our study suggests that DNN may not offer substantial benefit when trained on harmonized data, compared to traditional methods using established risk factors for MI.

Darolutamide: An Evidenced-Based Review of Its Efficacy and Safety in the Treatment of Prostate Cancer.

Men treated with androgen deprivation therapy for rising PSA after failed local therapy will often develop castrate resistance, and the appearance of metastases predicts a poor prognosis. Thus, researchers have long sought to prolong the onset of metastasis in patients with nonmetastatic castration-resistant prostate cancer (CRPC). Until 2018, patients in this group had no FDA-approved treatment options. They were typically managed with androgen-deprivation therapy (ADT) to maintain castrate systemic testosterone levels and given approved therapies for metastatic CRPC once metastases appeared. However, third-generation androgen receptor inhibitors (ARIs) have dramatically changed the treatment paradigm, having shown the ability to extend metastasis-free survival (MFS) significantly over ADT alone in Phase 3 trials. The newest of these, darolutamide, prolonged MFS 22 months over placebo while also improving a host of secondary and exploratory endpoints such as overall survival (OS), prostate-specific antigen (PSA) progression and time to pain progression, chemotherapy initiation, and symptomatic skeletal events. Among third-generation ARIs, darolutamide is unique in that it incorporates two pharmacologically active diastereomers and has demonstrated resistance to all known androgen receptor (AR) mutations. Additionally, patients taking darolutamide appear to experience comparatively few central nervous system-related adverse events (AEs) such as fatigue and falls, and no increases in seizures have been reported in the drug's clinical or preclinical development. Various authors attribute the low incidence of CNS-related AEs to darolutamide's minimal penetration of the blood-brain barrier (BBB). Other side effects ranging from hot flashes to hypothyroidism also occurred at rates similar to those of the placebo arm in Phase 3. As ADT in itself raises cardiovascular risk, the cardiovascular safety of third-generation antiandrogens as a category warrants continued scrutiny. In total, however, published data suggest that darolutamide provides a reasonable option for patients with nonmetastatic CRPC. Ongoing research will determine darolutamide's potential role in additional disease states such as localized and castration-sensitive PCa.

3D printed mitral valve models: affordable simulation for robotic mitral valve repair.

OBJECTIVES: 3D printed mitral valve (MV) models that capture the suture response of real tissue may be utilized as surgical training tools. Leveraging clinical imaging modalities, 3D computerized modelling and 3D printing technology to produce affordable models complements currently available virtual simulators and paves the way for patient- and pathology-specific preoperative rehearsal. METHODS: We used polyvinyl alcohol, a dissolvable thermoplastic, to 3D print moulds that were casted with liquid platinum-cure silicone yielding flexible, low-cost MV models capable of simulating valvular tissue. Silicone-moulded MV models were fabricated for 2 morphologies: the normal MV and the P2 flail. The moulded valves were plication and suture tested in a laparoscopic trainer box with a da Vinci Si robotic surgical system. One cardiothoracic surgery fellow and 1 attending surgeon qualitatively evaluated the ability of the valves to recapitulate tissue feel through surveys utilizing the 5-point Likert-type scale to grade impressions of the valves. RESULTS: Valves produced with the moulding and casting method maintained anatomical dimensions within 3% of directly 3D printed acrylonitrile butadiene styrene controls for both morphologies. Likert-type scale mean scores corresponded with a realistic material response to sutures (5.0/5), tensile strength that is similar to real MV tissue (5.0/5) and anatomical appearance resembling real MVs (5.0/5), indicating that evaluators 'agreed' that these aspects of the model were appropriate for training. Evaluators 'somewhat agreed' that the overall model durability was appropriate for training (4.0/5) due to the mounting design. Qualitative differences in repair quality were notable between fellow and attending surgeon. CONCLUSIONS: 3D computer-aided design, 3D printing and fabrication techniques can be applied to fabricate affordable, high-quality educational models for technical training that are capable of differentiating proficiency levels among users.

A Case of Persistent Low Back Pain in a Young Female Caused by a Trauma-Induced Schmorl's Node in the Lumbar Spine Five Vertebra.

Physicians are often faced with managing difficult conditions such as chronic lower back pain. Intervertebral disk herniation typically occurs horizontally, leading to impingement of the spinal cord which can potentially cause radicular symptoms or other spinal cord pathologies; however, disk herniations can also occur vertically and extend through the endplate of an adjacent cranial or caudal vertebra: a phenomenon known as a Schmorl's node. Although Schmorl's nodes can be seen in many asymptomatic individuals, they can be a cause of degenerative disk disease and low back pain. An 18-year-old female with a history of trauma presented to urgent care with increasing lower back pain for the past six weeks. Four months prior, she was struck by a motor vehicle while riding her bicycle, and she had residual back pain since then. Plain radiography at the time of the accident showed no acute abnormalities. She had no other associated symptoms. On presentation, her vital signs were within normal limits, and her physical examination was largely unremarkable except for point tenderness along the lumbar (L4-L5) region of the spine. A complete blood count showed no leukocytosis and plain radiography of the lumbosacral spine showed a Schmorl's node in the inferior endplate of L5. The patient was diagnosed with a trauma-induced Schmorl's node and was treated with physical therapy, ice packs, and non-steroidal anti-inflammatory drugs. Her symptoms improved over the next several months. For patients with a history of axial load trauma and persistent back pain, clinicians should consider the possibility of a trauma-induced Schmorl's node. Plain radiography or magnetic resonance imaging can help with the diagnosis and guide further management.