A machine learning model of response to hypomethylating agents in myelodysplastic syndromes.

Hypomethylating agents (HMA) prolong survival and improve cytopenias in individuals with higher-risk myelodysplastic syndrome (MDS). Only 30-40% of patients, however, respond to HMAs, and responses may not occur for more than 6 months after HMA initiation. We developed a model to more rapidly assess HMA response by analyzing early changes in patients' blood counts. Three institutions' data were used to develop a model that assessed patients' response to therapy 90 days after the initiation using serial blood counts. The model was developed with a training cohort of 424 patients from 2 institutions and validated on an independent cohort of 90 patients. The final model achieved an area under the receiver operating characteristic curve (AUROC) of 0.79 in the train/test group and 0.84 in the validation group. The model provides cohort-wide and individual-level explanations for model predictions, and model certainty can be interrogated to gauge the reliability of a given prediction.

Why Machine Learning Should Be Taught in Medical Schools.

The rapid development of machine learning (ML) applications in healthcare promises to transform the landscape of healthcare. In order for ML advancements to be effectively utilized in clinical care, it is necessary for the medical workforce to be prepared to handle these changes. As physicians in training are exposed to a wide breadth of clinical tools during medical school, this offers an ideal opportunity to introduce ML concepts. A foundational understanding of ML will not only be practically useful for clinicians, but will also address ethical concerns for clinical decision making. While select medical schools have made effort to integrate ML didactics and practice into their curriculum, we argue that foundational ML principles should be taught broadly to medical students across the country.

A geno-clinical decision model for the diagnosis of myelodysplastic syndromes.

The differential diagnosis of myeloid malignancies is challenging and subject to interobserver variability. We used clinical and next-generation sequencing (NGS) data to develop a machine learning model for the diagnosis of myeloid malignancies independent of bone marrow biopsy data based on a 3-institution, international cohort of patients. The model achieves high performance, with model interpretations indicating that it relies on factors similar to those used by clinicians. In addition, we describe associations between NGS findings and clinically important phenotypes and introduce the use of machine learning algorithms to elucidate clinicogenomic relationships.

Personalized Prediction Model to Risk Stratify Patients With Myelodysplastic Syndromes.

PURPOSE: Patients with myelodysplastic syndromes (MDS) have a survival that can range from months to decades. Prognostic systems that incorporate advanced analytics of clinical, pathologic, and molecular data have the potential to more accurately and dynamically predict survival in patients receiving various therapies. METHODS: A total of 1,471 MDS patients with comprehensively annotated clinical and molecular data were included in a training cohort and analyzed using machine learning techniques. A random survival algorithm was used to build a prognostic model, which was then validated in external cohorts. The accuracy of the proposed model, compared with other established models, was assessed using a concordance (c)index. RESULTS: The median age for the training cohort was 71 years. Commonly mutated genes included SF3B1, TET2, and ASXL1. The algorithm identified chromosomal karyotype, platelet, hemoglobin levels, bone marrow blast percentage, age, other clinical variables, seven discrete gene mutations, and mutation number as having prognostic impact on overall and leukemia-free survivals. The model was validated in an independent external cohort of 465 patients, a cohort of patients with MDS treated in a prospective clinical trial, a cohort of patients with paired samples at different time points during the disease course, and a cohort of patients who underwent hematopoietic stem-cell transplantation. CONCLUSION: A personalized prediction model on the basis of clinical and genomic data outperformed established prognostic models in MDS. The new model was dynamic, predicting survival and leukemia transformation probabilities at different time points that are unique for a given patient, and can upstage and downstage patients into more appropriate risk categories.

Personalized Prediction of Hospital Mortality in COVID-19-Positive Patients.

OBJECTIVE: To develop predictive models for in-hospital mortality and length of stay (LOS) for coronavirus disease 2019 (COVID-19)-positive patients. PATIENTS AND METHODS: We performed a multicenter retrospective cohort study of hospitalized COVID-19-positive patients. A total of 764 patients admitted to 14 different hospitals within the Cleveland Clinic from March 9, 2020, to May 20, 2020, who had reverse transcriptase-polymerase chain reaction-proven coronavirus infection were included. We used LightGBM, a machine learning algorithm, to predict in-hospital mortality at different time points (after 7, 14, and 30 days of hospitalization) and in-hospital LOS. Our final cohort was composed of 764 patients admitted to 14 different hospitals within our system. RESULTS: The median LOS was 5 (range, 1-44) days for patients admitted to the regular nursing floor and 10 (range, 1-38) days for patients admitted to the intensive care unit. Patients who died during hospitalization were older, initially admitted to the intensive care unit, and more likely to be white and have worse organ dysfunction compared with patients who survived their hospitalization. Using the 10 most important variables only, the final model's area under the receiver operating characteristics curve was 0.86 for 7-day, 0.88 for 14-day, and 0.85 for 30-day mortality in the validation cohort. CONCLUSION: We developed a decision tool that can provide explainable and patient-specific prediction of in-hospital mortality and LOS for COVID-19-positive patients. The model can aid health care systems in bed allocation and distribution of vital resources.

Machine Learning Models to Predict Major Adverse Cardiovascular Events After Orthotopic Liver Transplantation: A Cohort Study.

OBJECTIVE: To develop machine learning models that can predict post-transplantation major adverse cardiovascular events (MACE), all-cause mortality, and cardiovascular mortality in patients undergoing liver transplantation (LT). DESIGN: Retrospective cohort study. SETTING: High-volume tertiary care center. PARTICIPANTS: The study comprised 1,459 consecutive patients undergoing LT between January 2008 and December 2019. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: MACE, all-cause mortality, and cardiovascular mortality were modeled using logistic regression, least absolute shrinkage and selection surgery regression, random forests, support vector machine, and gradient-boosted modeling (GBM). All models were built by splitting data into training and testing cohorts, and performance was assessed using five-fold cross-validation based on the area under the receiver operating characteristic curve and Harrell's C statistic. A total of 1,459 patients were included in the final cohort; 1,425 (97.7%) underwent index transplantation, 963 (66.0%) were female, the median age at transplantation was 57 (11-70) years, and the median Model for End-Stage Liver Disease score was 20 (6-40). Across all outcomes, the GBM model XGBoost achieved the highest performance, with an area under the receiver operating curve of 0.71 (95% confidence interval [CI] 0.63-0.79) for MACE, a Harrell's C statistic of 0.64 (95% CI 0.57-0.73) for overall survival, and 0.72 (95% CI 0.59-0.85) for cardiovascular mortality over a mean follow-up of 4.4 years. Examination of Shapley values for the GBM model revealed that on the cohort-wide level, the top influential factors for postoperative MACE were age at transplantation, diabetes, serum creatinine, cirrhosis caused by nonalcoholic steatohepatitis, right ventricular systolic pressure, and left ventricular ejection fraction. CONCLUSION: Machine learning models developed using data from a tertiary care transplantation center achieved good discriminant function in predicting post-LT MACE, all-cause mortality, and cardiovascular mortality. These models can support clinicians in recipient selection and help screen individuals who may be at elevated risk for post-transplantation MACE.

The Impact of Clinical Decision Support Alerts on Clostridioides difficile Testing: A Systematic Review.

BACKGROUND: Several studies have investigated the utility of electronic decision support alerts in diagnostic stewardship for Clostridioides difficile infection (CDI). However, it is unclear if alerts are effective in reducing inappropriate CDI testing and/or CDI rates. The aim of this systematic review was to determine if alerts related to CDI diagnostic stewardship are effective at reducing inappropriate CDI testing volume and CDI rates among hospitalized adult patients. METHODS: We searched Ovid Medline and 5 other databases for original studies evaluating the association between alerts for CDI diagnosis and CDI testing volume and/or CDI rate. Two investigators independently extracted data on study characteristics, study design, alert triggers, cointerventions, and study outcomes. RESULTS: Eleven studies met criteria for inclusion. Studies varied significantly in alert triggers and in study outcomes. Six of 11 studies demonstrated a statistically significant decrease in CDI testing volume, 6 of 6 studies evaluating appropriateness of CDI testing found a significant reduction in the proportion of inappropriate testing, and 4 of 7 studies measuring CDI rate demonstrated a significant decrease in the CDI rate in the postintervention vs preintervention period. The magnitude of the increase in appropriate CDI testing varied, with some studies reporting an increase with minimal clinical significance. CONCLUSIONS: The use of electronic alerts for diagnostic stewardship for C. difficile was associated with reductions in CDI testing, the proportion of inappropriate CDI testing, and rates of CDI in most studies. However, broader concerns related to alerts remain understudied, including unintended adverse consequences and alert fatigue.

Acute myeloid leukemia and artificial intelligence, algorithms and new scores.

Artificial intelligence, and more narrowly machine-learning, is beginning to expand humanity's capacity to analyze increasingly large and complex datasets. Advances in computer hardware and software have led to breakthroughs in multiple sectors of our society, including a burgeoning role in medical research and clinical practice. As the volume of medical data grows at an apparently exponential rate, particularly since the human genome project laid the foundation for modern genetic inquiry, informatics tools like machine learning are becoming crucial in analyzing these data to provide meaningful tools for diagnostic, prognostic, and therapeutic purposes. Within medicine, hematologic diseases can be particularly challenging to understand and treat given the increasingly complex and intercalated genetic, epigenetic, immunologic, and regulatory pathways that must be understood to optimize patient outcomes. In acute myeloid leukemia (AML), new developments in machine learning algorithms have enabled a deeper understanding of disease biology and the development of better prognostic and predictive tools. Ongoing work in the field brings these developments incrementally closer to clinical implementation.

Machine Learning in Oncology: What Should Clinicians Know?

The volume and complexity of scientific and clinical data in oncology have grown markedly over recent years, including but not limited to the realms of electronic health data, radiographic and histologic data, and genomics. This growth holds promise for a deeper understanding of malignancy and, accordingly, more personalized and effective oncologic care. Such goals require, however, the development of new methods to fully make use of the wealth of available data. Improvements in computer processing power and algorithm development have positioned machine learning, a branch of artificial intelligence, to play a prominent role in oncology research and practice. This review provides an overview of the basics of machine learning and highlights current progress and challenges in applying this technology to cancer diagnosis, prognosis, and treatment recommendations, including a discussion of current takeaways for clinicians.

Machine learning in haematological malignancies.

Machine learning is a branch of computer science and statistics that generates predictive or descriptive models by learning from training data rather than by being rigidly programmed. It has attracted substantial attention for its many applications in medicine, both as a catalyst for research and as a means of improving clinical care across the cycle of diagnosis, prognosis, and treatment of disease. These applications include the management of haematological malignancy, in which machine learning has created inroads in pathology, radiology, genomics, and the analysis of electronic health record data. As computational power becomes cheaper and the tools for implementing machine learning become increasingly democratised, it is likely to become increasingly integrated into the research and practice landscape of haematology. As such, machine learning merits understanding and attention from researchers and clinicians alike. This narrative Review describes important concepts in machine learning for unfamiliar readers, details machine learning's current applications in haematological malignancy, and summarises important concepts for clinicians to be aware of when appraising research that uses machine learning.

Lessons Learned in Using Laser Interstitial Thermal Therapy for Treatment of Brain Tumors: A Case Series of 238 Patients from a Single Institution.

BACKGROUND: Laser interstitial thermal therapy (LITT) is a novel, minimally invasive alternative to craniotomy, and as with any new technology, comes with a learning curve. OBJECTIVE: We present our experience detailing the evolution of this technology in our practice in one of the largest patient cohorts to date regarding LITT in neuro-oncology. METHODS: We reviewed 238 consecutive patients with brain tumor treated with LITT at our institution. Data on patient, surgery and tumor characteristics, and follow-up were collected. Patients were categorized into 2 cohorts: early (<2014, 100 patients) and recent (>2015, 138 patients). Median follow-up for the entire cohort was 8.4 months. RESULTS: The indications for LITT included gliomas (70.2%), radiation necrosis (21.0%), and metastasis (8.8%). Patient demographics stayed consistent between the 2 cohorts, with the exception of age (early, 54.3; recent, 58.4; P = 0.04). Operative time (6.6 vs. 3.5; P < 0.001) and number of trajectories (53.1% vs. 77.9% with 1 trajectory; P < 0.001) also decreased in the recent cohort. There was a significant decrease in permanent motor deficits over time (15.5 vs. 4.4%; P = 0.005) and 30-day mortality (4.1% vs. 1.5%) also decreased (not statistically significant) in the recent cohort. In terms of clinical outcomes, poor preoperative Karnofsky Performance Status (≤70) were significantly correlated with increased permanent deficits (P = 0.001) and decreased overall survival (P < 0.001 for all time points). CONCLUSIONS: We observed improvement in operative efficiency and permanent deficits over time and also patients with poor preoperative Karnofsky Performance Status achieved suboptimal outcomes with LITT. As many other treatment modalities, patient selection is important in this procedure.

Artificial Intelligence in Hematology: Current Challenges and Opportunities.

PURPOSE OF REVIEW: Artificial intelligence (AI), and in particular its subcategory machine learning, is finding an increasing number of applications in medicine, driven in large part by an abundance of data and powerful, accessible tools that have made AI accessible to a larger circle of investigators. RECENT FINDINGS: AI has been employed in the analysis of hematopathological, radiographic, laboratory, genomic, pharmacological, and chemical data to better inform diagnosis, prognosis, treatment planning, and foundational knowledge related to benign and malignant hematology. As more widespread implementation of clinical AI nears, attention has also turned to the effects this will have on other areas in medicine. AI offers many promising tools to clinicians broadly, and specifically in the practice of hematology. Ongoing research into its various applications will likely result in an increasing utilization of AI by a broader swath of clinicians.

Virus-like Particles Identify an HIV V1V2 Apex-Binding Neutralizing Antibody that Lacks a Protruding Loop.

Most HIV-1-specific neutralizing antibodies isolated to date exhibit unusual characteristics that complicate their elicitation. Neutralizing antibodies that target the V1V2 apex of the HIV-1 envelope (Env) trimer feature unusually long protruding loops, which enable them to penetrate the HIV-1 glycan shield. As antibodies with loops of requisite length are created through uncommon recombination events, an alternative mode of apex binding has been sought. Here, we isolated a lineage of Env apex-directed neutralizing antibodies, N90-VRC38.01-11, by using virus-like particles and conformationally stabilized Env trimers as B cell probes. A crystal structure of N90-VRC38.01 with a scaffolded V1V2 revealed a binding mode involving side-chain-to-side-chain interactions that reduced the distance the antibody loop must traverse the glycan shield, thereby facilitating V1V2 binding via a non-protruding loop. The N90-VRC38 lineage thus identifies a solution for V1V2-apex binding that provides a more conventional B cell pathway for vaccine design.

Mimicry of an HIV broadly neutralizing antibody epitope with a synthetic glycopeptide.

A goal for an HIV-1 vaccine is to overcome virus variability by inducing broadly neutralizing antibodies (bnAbs). One key target of bnAbs is the glycan-polypeptide at the base of the envelope (Env) third variable loop (V3). We have designed and synthesized a homogeneous minimal immunogen with high-mannose glycans reflective of a native Env V3-glycan bnAb epitope (Man9-V3). V3-glycan bnAbs bound to Man9-V3 glycopeptide and native-like gp140 trimers with similar affinities. Fluorophore-labeled Man9-V3 glycopeptides bound to bnAb memory B cells and were able to be used to isolate a V3-glycan bnAb from an HIV-1-infected individual. In rhesus macaques, immunization with Man9-V3 induced V3-glycan-targeted antibodies. Thus, the Man9-V3 glycopeptide closely mimics an HIV-1 V3-glycan bnAb epitope and can be used to isolate V3-glycan bnAbs.