Natural Language Processing of Computed Tomography Reports to Label Metastatic Phenotypes With Prognostic Significance in Patients With Colorectal Cancer.

PURPOSE: Natural language processing (NLP) applied to radiology reports can help identify clinically relevant M1 subcategories of patients with colorectal cancer (CRC). The primary purpose was to compare the overall survival (OS) of CRC according to American Joint Committee on Cancer TNM staging and explore an alternative classification. The secondary objective was to estimate the frequency of metastasis for each organ. METHODS: Retrospective study of CRC who underwent computed tomography (CT) chest, abdomen, and pelvis between July 1, 2009, and March 26, 2019, at a tertiary cancer center, previously labeled for the presence or absence of metastasis by an NLP prediction model. Patients were classified in M0, M1a, M1b, and M1c (American Joint Committee on Cancer), or an alternative classification on the basis of the metastasis organ number: M1, single; M2, two; M3, three or more organs. Cox regression models were used to estimate hazard ratios; Kaplan-Meier curves were used to visualize survival curves using the two M1 subclassifications. RESULTS: Nine thousand nine hundred twenty-eight patients with a total of 48,408 CT chest, abdomen, and pelvis reports were included. On the basis of NLP prediction, the median OS of M1a, M1b, and M1c was 4.47, 1.72, and 1.52 years, respectively. The median OS of M1, M2, and M3 was 4.24, 2.05, and 1.04 years, respectively. Metastases occurred most often in liver (35.8%), abdominopelvic lymph nodes (32.9%), lungs (29.3%), peritoneum (22.0%), thoracic nodes (19.9%), bones (9.2%), and pelvic organs (7.5%). Spleen and adrenal metastases occurred in < 5%. CONCLUSION: NLP applied to a large radiology report database can identify clinically relevant metastatic phenotypes and be used to investigate new M1 substaging for CRC. Patients with three or more metastatic disease organs have the worst prognosis, with an OS of 1 year.

Developing a Cancer Digital Twin: Supervised Metastases Detection From Consecutive Structured Radiology Reports.

The development of digital cancer twins relies on the capture of high-resolution representations of individual cancer patients throughout the course of their treatment. Our research aims to improve the detection of metastatic disease over time from structured radiology reports by exposing prediction models to historical information. We demonstrate that Natural language processing (NLP) can generate better weak labels for semi-supervised classification of computed tomography (CT) reports when it is exposed to consecutive reports through a patient's treatment history. Around 714,454 structured radiology reports from Memorial Sloan Kettering Cancer Center adhering to a standardized departmental structured template were used for model development with a subset of the reports included for validation. To develop the models, a subset of the reports was curated for ground-truth: 7,732 total reports in the lung metastases dataset from 867 individual patients; 2,777 reports in the liver metastases dataset from 315 patients; and 4,107 reports in the adrenal metastases dataset from 404 patients. We use NLP to extract and encode important features from the structured text reports, which are then used to develop, train, and validate models. Three models-a simple convolutional neural network (CNN), a CNN augmented with an attention layer, and a recurrent neural network (RNN)-were developed to classify the type of metastatic disease and validated against the ground truth labels. The models use features from consecutive structured text radiology reports of a patient to predict the presence of metastatic disease in the reports. A single-report model, previously developed to analyze one report instead of multiple past reports, is included and the results from all four models are compared based on accuracy, precision, recall, and F1-score. The best model is used to label all 714,454 reports to generate metastases maps. Our results suggest that NLP models can extract cancer progression patterns from multiple consecutive reports and predict the presence of metastatic disease in multiple organs with higher performance when compared with a single-report-based prediction. It demonstrates a promising automated approach to label large numbers of radiology reports without involving human experts in a time- and cost-effective manner and enables tracking of cancer progression over time.

Analysis of a Remote Monitoring Program for Symptoms Among Adults With Cancer Receiving Antineoplastic Therapy.

IMPORTANCE: Electronic patient-reported outcomes (ePROs) may have the potential to improve cancer care delivery by enhancing patient quality of life, reducing acute care visits, and extending overall survival. However, the optimal cadence of ePRO assessments is unknown. OBJECTIVE: To determine patient response preferences and the clinical value associated with a daily cadence for ePROs for patients receiving antineoplastic treatment. DESIGN, SETTING, AND PARTICIPANTS: This quality improvement study of adult patients undergoing antineoplastic treatment assessed a remote monitoring program using ePROs that was developed to manage cancer therapy-related symptoms. ePRO data submitted between October 16, 2018 to February 29, 2020, from a single regional site within the Memorial Sloan Kettering Cancer Center network were included. Data were analyzed from April 2020 to January 2022. EXPOSURE: While undergoing active treatment, patients received a daily ePRO assessment that, based on patient responses, generated yellow (moderate) or red (severe) symptom alerts that were sent to clinicians. MAIN OUTCOMES AND MEASURES: The main outcomes assessed included patient response rate, symptom alert frequency, and an analysis of the clinical value of daily ePROs. RESULTS: A total of 217 patients (median [range] age, 66 [31-92] years; 103 [47.5%] women and 114 [52.5%] men) initiating antineoplastic therapy at high risk for symptoms were monitored for a median (range) of 91 (2-369) days. Most patients had thoracic (59 patients [27.2%]), head and neck (48 patients [22.1%]), or gastrointestinal (43 patients [19.8%]) malignant neoplasms. Of 14 603 unique symptom assessments completed, 7349 (50.3%) generated red or yellow symptom alerts. Symptoms commonly generating alerts included pain (665 assessments [23.0%]) and functional status (465 assessments [16.1%]). Most assessments (8438 assessments [57.8%]) were completed at home during regular clinic hours (ie, 9 am-5 pm), with higher response rates on weekdays (58.4%; 95% CI, 57.5%-59.5%) than on weekend days (51.3%; 95% CI, 49.5%-53.1%). Importantly, 284 of 630 unique red alerts (45.1%) surfaced without a prior yellow alert for the same symptom within the prior 7 days; symptom severity fluctuated over the course of a week, and symptom assessments generating a red alert were followed by an acute care visit within 7 days 8.7% of the time compared with 2.9% for assessments without a red alert. CONCLUSIONS AND RELEVANCE: These findings suggest that daily ePRO assessments were associated with increased insight into symptom management in patients undergoing antineoplastic treatment and symptom alerts were associated with risk of acute care.

Natural Language Processing of Large-Scale Structured Radiology Reports to Identify Oncologic Patients With or Without Splenomegaly Over a 10-Year Period.

PURPOSE: To assess the accuracy of a natural language processing (NLP) model in extracting splenomegaly described in patients with cancer in structured computed tomography radiology reports. METHODS: In this retrospective study between July 2009 and April 2019, 3,87,359 consecutive structured radiology reports for computed tomography scans of the chest, abdomen, and pelvis from 91,665 patients spanning 30 types of cancer were included. A randomized sample of 2,022 reports from patients with colorectal cancer, hepatobiliary cancer (HB), leukemia, Hodgkin lymphoma (HL), and non-HL patients was manually annotated as positive or negative for splenomegaly. NLP model training/testing was performed on 1,617/405 reports, and a new validation set of 400 reports from all cancer subtypes was used to test NLP model accuracy, precision, and recall. Overall survival was compared between the patient groups (with and without splenomegaly) using Kaplan-Meier curves. RESULTS: The final cohort included 3,87,359 reports from 91,665 patients (mean age 60.8 years; 51.2% women). In the testing set, the model achieved accuracy of 92.1%, precision of 92.2%, and recall of 92.1% for splenomegaly. In the validation set, accuracy, precision, and recall were 93.8%, 92.9%, and 86.7%, respectively. In the entire cohort, splenomegaly was most frequent in patients with leukemia (32.5%), HB (17.4%), non-HL (9.1%), colorectal cancer (8.5%), and HL (5.6%). A splenomegaly label was associated with an increased risk of mortality in the entire cohort (hazard ratio 2.10; 95% CI, 1.98 to 2.22; P < .001). CONCLUSION: Automated splenomegaly labeling by NLP of radiology report demonstrates good accuracy, precision, and recall. Splenomegaly is most frequently reported in patients with leukemia, followed by patients with HB.

Patterns of Metastatic Disease in Patients with Cancer Derived from Natural Language Processing of Structured CT Radiology Reports over a 10-year Period.

Background Patterns of metastasis in cancer are increasingly relevant to prognostication and treatment planning but have historically been documented by means of autopsy series. Purpose To show the feasibility of using natural language processing (NLP) to gather accurate data from radiology reports for assessing spatial and temporal patterns of metastatic spread in a large patient cohort. Materials and Methods In this retrospective longitudinal study, consecutive patients who underwent CT from July 2009 to April 2019 and whose CT reports followed a departmental structured template were included. Three radiologists manually curated a sample of 2219 reports for the presence or absence of metastases across 13 organs; these manually curated reports were used to develop three NLP models with an 80%-20% split for training and test sets. A separate random sample of 448 manually curated reports was used for validation. Model performance was measured by accuracy, precision, and recall for each organ. The best-performing NLP model was used to generate a final database of metastatic disease across all patients. For each cancer type, statistical descriptive reports were provided by analyzing the frequencies of metastatic disease at the report and patient levels. Results In 91 665 patients (mean age ± standard deviation, 61 years ± 15; 46 939 women), 387 359 reports were labeled. The best-performing NLP model achieved accuracies from 90% to 99% across all organs. Metastases were most frequently reported in abdominopelvic (23.6% of all reports) and thoracic (17.6%) nodes, followed by lungs (14.7%), liver (13.7%), and bones (9.9%). Metastatic disease tropism is distinct among common cancers, with the most common first site being bones in prostate and breast cancers and liver among pancreatic and colorectal cancers. Conclusion Natural language processing may be applied to cancer patients' CT reports to generate a large database of metastatic phenotypes. Such a database could be combined with genomic studies and used to explore prognostic imaging phenotypes with relevance to treatment planning. © RSNA, 2021 Online supplemental material is available for this article.

Building a Clinically Relevant Risk Model: Predicting Risk of a Potentially Preventable Acute Care Visit for Patients Starting Antineoplastic Treatment.

PURPOSE: To create a risk prediction model that identifies patients at high risk for a potentially preventable acute care visit (PPACV). PATIENTS AND METHODS: We developed a risk model that used electronic medical record data from initial visit to first antineoplastic administration for new patients at Memorial Sloan Kettering Cancer Center from January 2014 to September 2018. The final time-weighted least absolute shrinkage and selection operator model was chosen on the basis of clinical and statistical significance. The model was refined to predict risk on the basis of 270 clinically relevant data features spanning sociodemographics, malignancy and treatment characteristics, laboratory results, medical and social history, medications, and prior acute care encounters. The binary dependent variable was occurrence of a PPACV within the first 6 months of treatment. There were 8,067 observations for new-start antineoplastic therapy in our training set, 1,211 in the validation set, and 1,294 in the testing set. RESULTS: A total of 3,727 patients experienced a PPACV within 6 months of treatment start. Specific features that determined risk were surfaced in a web application, riskExplorer, to enable clinician review of patient-specific risk. The positive predictive value of a PPACV among patients in the top quartile of model risk was 42%. This quartile accounted for 35% of patients with PPACVs and 51% of potentially preventable inpatient bed days. The model C-statistic was 0.65. CONCLUSION: Our clinically relevant model identified the patients responsible for 35% of PPACVs and more than half of the inpatient beds used by the cohort. Additional research is needed to determine whether targeting these high-risk patients with symptom management interventions could improve care delivery by reducing PPACVs.

Preoperative risk prediction for intraductal papillary mucinous neoplasms by quantitative CT image analysis.

BACKGROUND: Intraductal papillary mucinous neoplasms (IPMNs) are radiographically identifiable potential precursor lesions of pancreatic adenocarcinoma. While resection is recommended when main duct dilation is present, management of branch duct IPMN (BD-IPMN) remains controversial. This study sought to evaluate whether preoperative quantitative imaging features of BD-IPMNs could distinguish low-risk disease (low- and intermediate-grade dysplasia) from high-risk disease (high-grade dysplasia and invasive carcinoma). METHODS: Patients who underwent resection between 2005 and 2015 with pathologically proven BD-IPMN and a preoperative CT scan were included in the study. Quantitative image features were extracted using texture analysis and a novel quantitative mural nodularity feature developed for the study. Significant features on univariate analysis were combined with clinical variables to build a multivariate prediction model. RESULTS: Within the study group of 103 patients, 76 (74%) had low-risk disease and 27 (26%) had high-risk disease. Quantitative imaging features were prognostic of low-vs. high-risk disease. The model based only on clinical variables achieved an AUC of 0.67 and 0.79 with the addition of quantitative imaging features. CONCLUSION: Quantitative image analysis of BD-IPMNs is a novel method that may enable risk stratification. External validation may provide a reliable non-invasive prognostic tool for clinicians.

CT radiomics to predict high-risk intraductal papillary mucinous neoplasms of the pancreas.

PURPOSE: Intraductal papillary mucinous neoplasms (IPMNs) are radiographically visible precursor lesions of pancreatic cancer. Despite standard criteria for assessing risk, only 18% of cysts are malignant at resection. Thus, a large number of patients undergo unnecessary invasive surgery for benign disease. The ability to identify IPMNs with low or high risk of transforming into invasive cancer would optimize patient selection and improve surgical decision-making. The purpose of this study was to investigate quantitative CT imaging features as markers for objective assessment of IPMN risk. METHODS: This retrospective study analyzed pancreatic cyst and parenchyma regions extracted from CT scans in 103 patients to predict IPMN risk. Patients who underwent resection between 2005 and 2015 with pathologically proven branch duct (BD)-IPMN and a preoperative CT scan were included in the study. Expert pathologists categorized IPMNs as low or high risk following resection as part of routine clinical care. We extracted new radiographically inspired features as well as standard texture features and designed prediction models for the categorization of high- and low-risk IPMNs. Five clinical variables were also combined with imaging features to design prediction models. RESULTS: Using images from 103 patients and tenfold cross-validation technique, the novel radiographically inspired imaging features achieved an area under the receiver operating characteristic curve (AUC) of 0.77, demonstrating their predictive power. The combination of these features with clinical variables obtained the best performance (AUC = 0.81). CONCLUSION: The present study demonstrates that features extracted from pretreatment CT images can predict the risk of IPMN. Development of a preoperative model to discriminate between low-risk and high-risk IPMN will improve surgical decision-making.

Misery Loves Company: Presenting Symptom Clusters to Urgent Care by Patients Receiving Antineoplastic Therapy.

PURPOSE: The Centers for Medicare & Medicaid Services (CMS) identifies suboptimal management of treatment toxicities as a care gap and proposes the measurement of hospital performance on the basis of emergency department visits for 10 common symptoms. Current management strategies do not address symptom co-occurrence. METHODS: We evaluated symptom co-occurrence in three patient cohorts that presented to a cancer hospital urgent care center in 2016. We examined both the CMS-identified symptoms and an expanded clinician-identified set defined as symptoms that could be safely managed in the outpatient setting if identified early and managed proactively. The cohorts included patients who presented with a CMS-defined symptom within 30 days of treatment, patients who presented within 30 days of treatment with a symptom from the expanded set, and patients who presented with a symptom from the expanded set within 30 days of treatment start. Symptom co-occurrence was measured by Jaccard index. A community detection algorithm was used to identify symptom clusters on the basis of a random walk process, and network visualizations were used to illustrate symptom dynamics. RESULTS: There were 6,429 presentations in the CMS symptom-defined cohort. The network analysis identified two distinct symptom clusters centered around pain and fever. In the expanded symptom cohort, there were 5,731 visits and six symptom clusters centered around fever, emesis/nausea, fatigue, deep vein thrombosis, pain, and ascites. For patients who newly initiated treatment, there were 1,154 visits and four symptom clusters centered around fever, nausea/emesis, fatigue, and deep vein thrombosis. CONCLUSION: Uncontrolled symptoms are associated with unplanned acute care. Recognition of the complexity of symptom co-occurrence can drive improved management strategies.