MRI imaging features for predicting macrotrabecular-massive subtype hepatocellular carcinoma: a systematic review and meta-analysis.

PURPOSE: To identify significant MRI features associated with macrotrabecular-massive hepatocellular carcinoma (MTM-HCC), and to assess the distribution of Liver Imaging Radiology and Data System (LI-RADS, LR) category assignments. METHODS: PubMed and EMBASE were searched up to March 28, 2023. Random-effects model was constructed to calculate pooled diagnostic odds ratios (DORs) and 95% confidence intervals (CIs) for each MRI feature for differentiating MTM-HCC from NMTM-HCC. The pooled proportions of LI-RADS category assignments in MTM-HCC and NMTM-HCC were compared using z-test. RESULTS: Ten studies included 1978 patients with 2031 HCCs (426 (20.9%) MTM-HCC and 1605 (79.1%) NMTM-HCC). Six MRI features showed significant association with MTM-HCC: tumor in vein (TIV) (DOR = 2.4 [95% CI, 1.6-3.5]), rim arterial phase hyperenhancement (DOR =2.6 [95% CI, 1.4-5.0]), corona enhancement (DOR = 2.6 [95% CI, 1.4-4.5]), intratumoral arteries (DOR = 2.6 [95% CI, 1.1-6.3]), peritumoral hypointensity on hepatobiliary phase (DOR = 2.2 [95% CI, 1.5-3.3]), and necrosis (DOR = 4.2 [95% CI, 2.0-8.5]). The pooled proportions of LI-RADS categories in MTM-HCC were LR-3, 0% [95% CI, 0-2%]; LR-4, 11% [95% CI, 6-16%]; LR-5, 63% [95% CI, 55-71%]; LR-M, 12% [95% CI, 6-19%]; and LR-TIV, 13% [95% CI, 6-22%]. In NMTM-HCC, the pooled proportions of LI-RADS categories were LR-3, 1% [95% CI, 0-2%]; LR-4, 8% [95% CI, 3-15%]; LR-5, 77% [95% CI, 71-82%]; LR-M, 5% [95% CI, 3-7%]; and LR-TIV, 6% [95% CI, 2-11%]. MTM-HCC had significantly lower proportion of LR-5 and higher proportion of LR-M and LR-TIV categories. CONCLUSIONS: Six MRI features showed significant association with MTM-HCC. Additionally, compared to NMTM-HCC, MTM-HCC are more likely to be categorized LR-M and LR-TIV and less likely to be categorized LR-5. CLINICAL RELEVANCE STATEMENT: Several MR imaging features can suggest macrotrabecular-massive hepatocellular carcinoma subtype, which can assist in guiding treatment plans and identifying potential candidates for clinical trials of new treatment strategies. KEY POINTS: • Macrotrabecular-massive hepatocellular carcinoma is a subtype of HCC characterized by its aggressive nature and unfavorable prognosis. • Tumor in vein, rim arterial phase hyperenhancement, corona enhancement, intratumoral arteries, peritumoral hypointensity on hepatobiliary phase, and necrosis on MRI are indicative of macrotrabecular-massive hepatocellular carcinoma. • Various MRI characteristics can be utilized for the diagnosis of the macrotrabecular-massive hepatocellular carcinoma subtype. This can prove beneficial in guiding treatment decisions and identifying potential candidates for clinical trials involving novel treatment approaches.

Preoperative CT and survival data for patients undergoing resection of colorectal liver metastases.

The liver is a common site for the development of metastases in colorectal cancer. Treatment selection for patients with colorectal liver metastases (CRLM) is difficult; although hepatic resection will cure a minority of CRLM patients, recurrence is common. Reliable preoperative prediction of recurrence could therefore be a valuable tool for physicians in selecting the best candidates for hepatic resection in the treatment of CRLM. It has been hypothesized that evidence for recurrence could be found via quantitative image analysis on preoperative CT imaging of the future liver remnant before resection. To investigate this hypothesis, we have collected preoperative hepatic CT scans, clinicopathologic data, and recurrence/survival data, from a large, single-institution series of patients (n = 197) who underwent hepatic resection of CRLM. For each patient, we also created segmentations of the liver, vessels, tumors, and future liver remnant. The largest of its kind, this dataset is a resource that may aid in the development of quantitative imaging biomarkers and machine learning models for the prediction of post-resection hepatic recurrence of CRLM.

Applying Natural Language Processing to Single-Report Prediction of Metastatic Disease Response Using the OR-RADS Lexicon.

Generating Real World Evidence (RWE) on disease responses from radiological reports is important for understanding cancer treatment effectiveness and developing personalized treatment. A lack of standardization in reporting among radiologists impacts the feasibility of large-scale interpretation of disease response. This study examines the utility of applying natural language processing (NLP) to the large-scale interpretation of disease responses using a standardized oncologic response lexicon (OR-RADS) to facilitate RWE collection. Radiologists annotated 3503 retrospectively collected clinical impressions from radiological reports across several cancer types with one of seven OR-RADS categories. A Bidirectional Encoder Representations from Transformers (BERT) model was trained on this dataset with an 80-20% train/test split to perform multiclass and single-class classification tasks using the OR-RADS. Radiologists also performed the classification to compare human and model performance. The model achieved accuracies from 95 to 99% across all classification tasks, performing better in single-class tasks compared to the multiclass task and producing minimal misclassifications, which pertained mostly to overpredicting the equivocal and mixed OR-RADS labels. Human accuracy ranged from 74 to 93% across all classification tasks, performing better on single-class tasks. This study demonstrates the feasibility of the BERT NLP model in predicting disease response in cancer patients, exceeding human performance, and encourages the use of the standardized OR-RADS lexicon to improve large-scale prediction accuracy.

Targeted Molecular Profiling of Circulating Cell-Free DNA in Patients With Advanced Hepatocellular Carcinoma.

PURPOSE: Next-generation sequencing (NGS) of tumor-derived, circulating cell-free DNA (cfDNA) may aid in diagnosis, prognostication, and treatment of patients with hepatocellular carcinoma (HCC). The operating characteristics of cfDNA mutational profiling must be determined before routine clinical implementation. METHODS: This was a single-center, retrospective study with the primary objective of defining genomic alterations in circulating cfDNA along with plasma-tissue genotype agreement between NGS of matched tumor samples in patients with advanced HCC. cfDNA was analyzed using a clinically validated 129-gene NGS assay; matched tissue-based NGS was analyzed with a US Food and Drug Administration-authorized NGS tumor assay. RESULTS: Fifty-three plasma samples from 51 patients with histologically confirmed HCC underwent NGS-based cfDNA analysis. Genomic alterations were detected in 92.2% of patients, with the most commonly mutated genes including TERT promoter (57%), TP53 (47%), CTNNB1 (37%), ARID1A (18%), and TSC2 (14%). In total, 37 (73%) patients underwent paired tumor NGS, and concordance was high for mutations observed in patient-matched plasma samples: TERT (83%), TP53 (94%), CTNNB1 (92%), ARID1A (100%), and TSC2 (71%). In 10 (27%) of 37 tumor-plasma samples, alterations were detected by cfDNA analysis that were not detected in the patient-matched tumors. Potentially actionable mutations were identified in 37% of all cases including oncogenic/likely oncogenic alterations in TSC1/2 (18%), BRCA1/2 (8%), and PIK3CA (8%). Higher average variant allele fraction was associated with elevated alpha-fetoprotein, increased tumor volume, and no previous systemic therapy, but did not correlate with overall survival in treatment-naïve patients. CONCLUSION: Tumor mutation profiling of cfDNA in HCC represents an alternative to tissue-based genomic profiling, given the high degree of tumor-plasma NGS concordance; however, genotyping of both blood and tumor may be required to detect all clinically actionable genomic alterations.

Registration-based biomarkers for neoadjuvant treatment response of pancreatic cancer via longitudinal image registration.

Purpose: Pancreatic ductal adenocarcinoma (PDAC) frequently presents as hypo- or iso-dense masses with poor contrast delineation from surrounding parenchyma, which decreases reproducibility of manual dimensional measurements obtained during conventional radiographic assessment of treatment response. Longitudinal registration between pre- and post-treatment images may produce imaging biomarkers that more reliably quantify treatment response across serial imaging. Approach: Thirty patients who prospectively underwent a neoadjuvant chemotherapy regimen as part of a clinical trial were retrospectively analyzed in this study. Two image registration methods were applied to quantitatively assess longitudinal changes in tumor volume and tumor burden across the neoadjuvant treatment interval. Longitudinal registration errors of the pancreas were characterized, and registration-based treatment response measures were correlated to overall survival (OS) and recurrence-free survival (RFS) outcomes over 5-year follow-up. Corresponding biomarker assessments via manual tumor segmentation, the standardized response evaluation criteria in solid tumors (RECIST), and pathological examination of post-resection tissue samples were analyzed as clinical comparators. Results: Average target registration errors were 2.56±2.45 mm for a biomechanical image registration algorithm and 4.15±3.63 mm for a diffeomorphic intensity-based algorithm, corresponding to 1-2 times voxel resolution. Cox proportional hazards analysis showed that registration-derived changes in tumor burden were significant predictors of OS and RFS, while none of the alternative comparators, including manual tumor segmentation, RECIST, or pathological variables were associated with consequential hazard ratios. Additional ROC analysis at 1-, 2-, 3-, and 5-year follow-up revealed that registration-derived changes in tumor burden between pre- and post-treatment imaging were better long-term predictors for OS and RFS than the clinical comparators. Conclusions: Volumetric changes measured by longitudinal deformable image registration may yield imaging biomarkers to discriminate neoadjuvant treatment response in ill-defined tumors characteristic of PDAC. Registration-based biomarkers may help to overcome visual limits of radiographic evaluation to improve clinical outcome prediction and inform treatment selection.

Artificial Intelligence in CT and MR Imaging for Oncological Applications.

Cancer care increasingly relies on imaging for patient management. The two most common cross-sectional imaging modalities in oncology are computed tomography (CT) and magnetic resonance imaging (MRI), which provide high-resolution anatomic and physiological imaging. Herewith is a summary of recent applications of rapidly advancing artificial intelligence (AI) in CT and MRI oncological imaging that addresses the benefits and challenges of the resultant opportunities with examples. Major challenges remain, such as how best to integrate AI developments into clinical radiology practice, the vigorous assessment of quantitative CT and MR imaging data accuracy, and reliability for clinical utility and research integrity in oncology. Such challenges necessitate an evaluation of the robustness of imaging biomarkers to be included in AI developments, a culture of data sharing, and the cooperation of knowledgeable academics with vendor scientists and companies operating in radiology and oncology fields. Herein, we will illustrate a few challenges and solutions of these efforts using novel methods for synthesizing different contrast modality images, auto-segmentation, and image reconstruction with examples from lung CT as well as abdome, pelvis, and head and neck MRI. The imaging community must embrace the need for quantitative CT and MRI metrics beyond lesion size measurement. AI methods for the extraction and longitudinal tracking of imaging metrics from registered lesions and understanding the tumor environment will be invaluable for interpreting disease status and treatment efficacy. This is an exciting time to work together to move the imaging field forward with narrow AI-specific tasks. New AI developments using CT and MRI datasets will be used to improve the personalized management of cancer patients.

Computerized Diagnosis of Liver Tumors From CT Scans Using a Deep Neural Network Approach.

The liver is a frequent site of benign and malignant, primary and metastatic tumors. Hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (ICC) are the most common primary liver cancers, and colorectal liver metastasis (CRLM) is the most common secondary liver cancer. Although the imaging characteristic of these tumors is central to optimal clinical management, it relies on imaging features that are often non-specific, overlap, and are subject to inter-observer variability. Thus, in this study, we aimed to categorize liver tumors automatically from CT scans using a deep learning approach that objectively extracts discriminating features not visible to the naked eye. Specifically, we used a modified Inception v3 network-based classification model to classify HCC, ICC, CRLM, and benign tumors from pretreatment portal venous phase computed tomography (CT) scans. Using a multi-institutional dataset of 814 patients, this method achieved an overall accuracy rate of 96%, with sensitivity rates of 96%, 94%, 99%, and 86% for HCC, ICC, CRLM, and benign tumors, respectively, using an independent dataset. These results demonstrate the feasibility of the proposed computer-assisted system as a novel non-invasive diagnostic tool to classify the most common liver tumors objectively.

LI-RADS: Looking Back, Looking Forward.

Since its initial release in 2011, the Liver Imaging Reporting and Data System (LI-RADS) has evolved and expanded in scope. It started as a single algorithm for hepatocellular carcinoma (HCC) diagnosis with CT or MRI with extracellular contrast agents and has grown into a multialgorithm network covering all major liver imaging modalities and contexts of use. Furthermore, it has developed its own lexicon, report templates, and supplementary materials. This article highlights the major achievements of LI-RADS in the past 11 years, including adoption in clinical care and research across the globe, and complete unification of HCC diagnostic systems in the United States. Additionally, the authors discuss current gaps in knowledge, which include challenges in surveillance, diagnostic population definition, perceived complexity, limited sensitivity of LR-5 (definite HCC) category, management implications of indeterminate observations, challenges in reporting, and treatment response assessment following radiation-based therapies and systemic treatments. Finally, the authors discuss future directions, which will focus on mitigating the current challenges and incorporating advanced technologies. Tha authors envision that LI-RADS will ultimately transform into a probability-based system for diagnosis and prognostication of liver cancers that will integrate patient characteristics and quantitative imaging features, while accounting for imaging modality and contrast agent.

Postoperative surveillance of pancreatic ductal adenocarcinoma (PDAC) recurrence: practice pattern on standardized imaging and reporting from the society of abdominal radiology disease focus panel on PDAC.

PURPOSE: Surgical resection is the only potential curative treatment for patients with pancreatic ductal adenocarcinoma (PDAC), but unfortunately most patients recur within 5 years of surgery. This article aims to assess the practice patterns across major academic institutions and develop consensus recommendations for postoperative imaging and interpretation in patients with PDAC. METHODS: The consensus recommendations for postoperative imaging surveillance following PDAC resection were developed using the Delphi method. Members of the Society of Abdominal Radiology (SAR) PDAC Disease Focused Panel (DFP) underwent three rounds of surveys followed by live webinar group discussions to develop consensus recommendations. RESULTS: Significant variations currently exist in the postoperative surveillance of PDAC, even among academic institutions. Differentiating common postoperative inflammatory and fibrotic changes from tumor recurrence remains a diagnostic challenge, and there is no reliable size threshold or growth rate of imaging findings that can provide differentiation. A new liver lesion or peritoneal nodule should be considered suspicious for tumor recurrence, and the imaging features should be interpreted in the appropriate clinical context (e.g., CA 19-9, clinical presentation, pathologic staging). CONCLUSION: Postoperative imaging following PDAC resection is challenging to interpret due to the presence of confounding postoperative inflammatory changes. A standardized reporting template for locoregional findings and report impression may improve communication of relaying risk of recurrence with referring providers, which merits validation in future studies.

Interim Analysis of a Prospective Validation of 2 Blood-Based Genomic Assessments (PPQ and NETest) to Determine the Clinical Efficacy of 177Lu-DOTATATE in Neuroendocrine Tumors.

Reliable biomarkers for neuroendocrine tumor (NET) management during peptide receptor radionuclide therapy (PRRT) are lacking. We validated the role of 2 circulating biomarkers: the PRRT prediction quotient (PPQ) as a predictive marker for response and the NETest as a monitoring biomarker. Furthermore, we evaluated whether tissue-based genetic alterations are effective in predicting progression-free survival (PFS). Methods: Data were prospectively collected on patients at the Memorial Sloan Kettering Cancer Center with 177Lu-DOTATATE-treated somatostatin receptor (SSTR)-positive gastroenteropancreatic and lung NETs (n = 67; median age, 66 y; 52% female; 42% pancreatic, 39% small-bowel; 78% grade 1 or 2). All cases were metastatic (89% liver) and had received 1-8 prior treatments (median, 3), including somatostatin analogs (91%), surgery (55%), or chemotherapy (49%). Treatment response included PFS. According to RECIST, version 1.1, responders had stable disease or a partial response (disease-control rate) and nonresponders had progression. Blood was collected before each cycle and at follow-up. Samples were deidentified and assayed and underwent masked analyses. The gene expression assays included RNA isolation, real-time quantitative polymerase chain reaction, and multialgorithm analyses. The PPQ (positive predicts a responder; negative predicts a nonresponder) at baseline was determined. The NETest (0-100 score) was performed. Statistics were analyzed using Mann-Whitney U testing (2-tailed) or Kaplan-Meier survival testing (PFS). In patients with archival tumor tissue, next-generation sequencing was performed through an institutional platform (Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets). Results: Forty-one patients (61%) were responders. PPQ accurately predicted 96% (64/67). The hazard ratio for prediction was 24.4 (95% CI, 8.2-72.5). Twelve-month disease control was 97% for PPQ-positive patients versus 26% for PPQ-negative patients (P < 0.0001). Median progression-free survival was not reached in those predicted to respond (PPQ-positive, n = 40) but was 8 mo in those predicted not to respond (PPQ-negative, n = 27). The NETest result in responders was 67 ± 25 at baseline and significantly (P < 0.05) decreased (-37 ± 44%) at follow-up. The NETest result in nonresponders was 44 ± 23 at baseline and significantly (P < 0.05) increased (+76% ± 56%) at progression. Overall, the NETest changes (increases or decreases) were 90% accurate. Thirty patients underwent next-generation sequencing. Tumors were microsatellite-stable, and the median mutational burden was 1.8. Alterations involved mainly the mTOR/PTEN/TSC pathway (30%). No relationship was associated with PRRT response. Conclusion: Our interim analysis confirmed that PPQ is an accurate predictor of 177Lu-DOTATATE responsiveness (radiosensitivity) and that NETest changes accurately correlated with treatment response. Tissue-based molecular genetic information had little value in PRRT prediction. Blood-based gene signatures may improve the management of patients undergoing 177Lu-DOTATATE by providing information on tumor radiosensitivity and disease course, thus allowing individualized strategies.

Hepatic Adenoma Subtypes on Hepatobiliary Phase of Gadoxetic Acid-Enhanced MRI: Systematic Review and Meta-Analysis.

BACKGROUND. Accumulating evidence indicates that hepatocellular adenoma (HCA) may have a higher frequency of hepatobiliary phase (HBP) iso- or hyperintensity than previously reported. OBJECTIVE. The purpose of this study was to evaluate the proportion of HCA that shows iso- or hyperintensity in the HBP of gadoxetic acid-enhanced MRI, stratified by HCA subtype (HNF1a-inactivated [H-HCA], inflammatory [I-HCA], ß-catenin-activated [B-HCA], and unclassified [U-HCA] HCA), and to assess the diagnostic performance of HBP iso- or hyperintensity for differentiating focal nodular hyperplasia (FNH) from HCA. EVIDENCE ACQUISITION. PubMed, Embase, and Cochrane Central Register of Controlled Trials were searched through February 14, 2022, for articles reporting HBP signal intensity on gadoxetic acid-enhanced MRI among pathologically proven HCAs, stratified by subtype. The pooled proportion of HBP iso- or hyperintensity was determined for each subtype and compared using metaregression. Diagnostic performance of HBP iso- or hyperintensity for differentiating FNH from all HCA subtypes combined and from B-HCA and U-HCA combined was assessed using bivariate modeling. EVIDENCE SYNTHESIS. Twenty-eight studies (12 original investigations, 16 case reports or case series) were included, yielding 364 patients with 410 HCAs (112 H-HCAs, 203 I-HCAs, 33 B-HCAs, 62 U-HCAs). Pooled proportion of HBP iso- or hyperintensity was 14% (95% CI, 4-26%) among all HCAs, 0% (95% CI, 0-2%) among H-HCAs, 11% (95% CI, 0-29%) among U-HCAs, 14% (95% CI, 2-31%) among I-HCAs, and 59% (95% CI, 26-88%) among B-HCAs; metaregression showed significant difference among subtypes (p < .001). In four studies reporting diagnostic performance information, HBP iso- or hyperintensity had sensitivity of 99% (95% CI, 57-100%) and specificity of 89% (95% CI, 82-94%) for differentiating FNH from all HCA subtypes and sensitivity of 99% (95% CI, 53-100%) and specificity of 65% (95% CI, 44-80%) for differentiating FNH from B-HCA or U-HCA. CONCLUSION. HCA subtypes other than H-HCA show proportions of HBP iso- or hyperintensity ranging from 11% (U-HCA) to 59% (B-HCA). Low prevalence of B-HCA has contributed to prior reports of high diagnostic performance of HBP iso- or hyperintensity for differentiating FNH from HCA. CLINICAL IMPACT. Radiologists should recognize the low specificity of HBP iso- or hyperintensity on gadoxetic acid-enhanced MRI for differentiating FNH from certain HCA subtypes.

The Medical Segmentation Decathlon.

International challenges have become the de facto standard for comparative assessment of image analysis algorithms. Although segmentation is the most widely investigated medical image processing task, the various challenges have been organized to focus only on specific clinical tasks. We organized the Medical Segmentation Decathlon (MSD)-a biomedical image analysis challenge, in which algorithms compete in a multitude of both tasks and modalities to investigate the hypothesis that a method capable of performing well on multiple tasks will generalize well to a previously unseen task and potentially outperform a custom-designed solution. MSD results confirmed this hypothesis, moreover, MSD winner continued generalizing well to a wide range of other clinical problems for the next two years. Three main conclusions can be drawn from this study: (1) state-of-the-art image segmentation algorithms generalize well when retrained on unseen tasks; (2) consistent algorithmic performance across multiple tasks is a strong surrogate of algorithmic generalizability; (3) the training of accurate AI segmentation models is now commoditized to scientists that are not versed in AI model training.

Liver imaging: it is time to adopt standardized terminology.

Liver imaging plays a vital role in the management of patients at risk for hepatocellular carcinoma (HCC); however, progress in the field is challenged by nonuniform and inconsistent terminology in the published literature. The Steering Committee of the American College of Radiology (ACR)'s Liver Imaging Reporting And Data System (LI-RADS), in conjunction with the LI-RADS Lexicon Writing Group and the LI-RADS International Working Group, present this consensus document to establish a single universal liver imaging lexicon. The lexicon is intended for use in research, education, and clinical care of patients at risk for HCC (i.e., the LI-RADS population) and in the general population (i.e., even when LI-RADS algorithms are not applicable). We anticipate that the universal adoption of this lexicon will provide research, educational, and clinical benefits. KEY POINTS: •To standardize terminology, we encourage authors of research and educational materials on liver imaging to use the standardized LI-RADS Lexicon. •We encourage reviewers to promote the use of the standardized LI-RADS Lexicon for publications on liver imaging. •We encourage radiologists to use the standardized LI-RADS Lexicon for liver imaging in clinical care.

Recurrence After Resection of Pancreatic Cancer: Can Radiomics Predict Patients at Greatest Risk of Liver Metastasis?

BACKGROUND: Liver metastasis (LM) after pancreatic ductal adenocarcinoma (PDAC) resection is common but difficult to predict and has grave prognosis. We combined preoperative clinicopathological variables and quantitative analysis of computed tomography (CT) imaging to predict early LM. METHODS: We retrospectively evaluated patients with PDAC submitted to resection between 2005 and 2014 and identified clinicopathological variables associated with early LM. We performed liver radiomic analysis on preoperative contrast-enhanced CT scans and developed a logistic regression classifier to predict early LM (< 6 months). RESULTS: In 688 resected PDAC patients, there were 516 recurrences (75%). The cumulative incidence of LM at 5 years was 41%, and patients who developed LM first (n = 194) had the lowest 1-year overall survival (OS) (34%), compared with 322 patients who developed extrahepatic recurrence first (61%). Independent predictors of time to LM included poor tumor differentiation (hazard ratio (HR) = 2.30; P < 0.001), large tumor size (HR = 1.17 per 2-cm increase; P = 0.048), lymphovascular invasion (HR = 1.50; P = 0.015), and liver Fibrosis-4 score (HR = 0.89 per 1-unit increase; P = 0.029) on multivariate analysis. A model using radiomic variables that reflect hepatic parenchymal heterogeneity identified patients at risk for early LM with an area under the receiver operating characteristic curve (AUC) of 0.71; the performance of the model was improved by incorporating preoperative clinicopathological variables (tumor size and differentiation status; AUC = 0.74, negative predictive value (NPV) = 0.86). CONCLUSIONS: We confirm the adverse survival impact of early LM after resection of PDAC. We further show that a model using radiomic data from preoperative imaging combined with tumor-related variables has great potential for identifying patients at high risk for LM and may help guide treatment selection.

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.

Machine learning radiomics can predict early liver recurrence after resection of intrahepatic cholangiocarcinoma.

BACKGROUND: Most patients recur after resection of intrahepatic cholangiocarcinoma (IHC). We studied whether machine-learning incorporating radiomics and tumor size could predict intrahepatic recurrence within 1-year. METHODS: This was a retrospective analysis of patients with IHC resected between 2000 and 2017 who had evaluable computed tomography imaging. Texture features (TFs) were extracted from the liver, tumor, and future liver remnant (FLR). Random forest classification using training (70.3%) and validation cohorts (29.7%) was used to design a predictive model. RESULTS: 138 patients were included for analysis. Patients with early recurrence had a larger tumor size (7.25 cm [IQR 5.2-8.9] vs. 5.3 cm [IQR 4.0-7.2], P = 0.011) and a higher rate of lymph node metastasis (28.6% vs. 11.6%, P = 0.041), but were not more likely to have multifocal disease (21.4% vs. 17.4%, P = 0.643). Three TFs from the tumor, FD1, FD30, and IH4 and one from the FLR, ACM15, were identified by feature selection. Incorporation of TFs and tumor size achieved the highest AUC of 0.84 (95% CI 0.73-0.95) in predicting recurrence in the validation cohort. CONCLUSION: This study demonstrates that radiomics and machine-learning can reliably predict patients at risk for early intrahepatic recurrence with good discrimination accuracy.

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.