Automated Radiology Report Summarization Using an Open-Source Natural Language Processing Pipeline.

Diagnostic radiologists are expected to review and assimilate findings from prior studies when constructing their overall assessment of the current study. Radiology information systems facilitate this process by presenting the radiologist with a subset of prior studies that are more likely to be relevant to the current study, usually by comparing anatomic coverage of both the current and prior studies. It is incumbent on the radiologist to review the full text report and/or images from those prior studies, a process that is time-consuming and confers substantial risk of overlooking a relevant prior study or finding. This risk is compounded when patients have dozens or even hundreds of prior imaging studies. Our goal is to assess the feasibility of natural language processing techniques to automatically extract asserted and negated disease entities from free-text radiology reports as a step towards automated report summarization. We compared automatically extracted disease mentions to a gold-standard set of manual annotations for 50 radiology reports from CT abdomen and pelvis examinations. The automated report summarization pipeline found perfect or overlapping partial matches for 86% of the manually annotated disease mentions (sensitivity 0.86, precision 0.66, accuracy 0.59, F1 score 0.74). The performance of the automated pipeline was good, and the overall accuracy was similar to the interobserver agreement between the two manual annotators.

ADAR1 promotes malignant progenitor reprogramming in chronic myeloid leukemia.

The molecular etiology of human progenitor reprogramming into self-renewing leukemia stem cells (LSC) has remained elusive. Although DNA sequencing has uncovered spliceosome gene mutations that promote alternative splicing and portend leukemic transformation, isoform diversity also may be generated by RNA editing mediated by adenosine deaminase acting on RNA (ADAR) enzymes that regulate stem cell maintenance. In this study, whole-transcriptome sequencing of normal, chronic phase, and serially transplantable blast crisis chronic myeloid leukemia (CML) progenitors revealed increased IFN-γ pathway gene expression in concert with BCR-ABL amplification, enhanced expression of the IFN-responsive ADAR1 p150 isoform, and a propensity for increased adenosine-to-inosine RNA editing during CML progression. Lentiviral overexpression experiments demonstrate that ADAR1 p150 promotes expression of the myeloid transcription factor PU.1 and induces malignant reprogramming of myeloid progenitors. Moreover, enforced ADAR1 p150 expression was associated with production of a misspliced form of GSK3ß implicated in LSC self-renewal. Finally, functional serial transplantation and shRNA studies demonstrate that ADAR1 knockdown impaired in vivo self-renewal capacity of blast crisis CML progenitors. Together these data provide a compelling rationale for developing ADAR1-based LSC detection and eradication strategies.

GLI2 inhibition abrogates human leukemia stem cell dormancy.

BACKGROUND: Dormant leukemia stem cells (LSC) promote therapeutic resistance and leukemic progression as a result of unbridled activation of stem cell gene expression programs. Thus, we hypothesized that 1) deregulation of the hedgehog (Hh) stem cell self-renewal and cell cycle regulatory pathway would promote dormant human LSC generation and 2) that PF-04449913, a clinical antagonist of the GLI2 transcriptional activator, smoothened (SMO), would enhance dormant human LSC eradication. METHODS: To test these postulates, whole transcriptome RNA sequencing (RNA-seq), microarray, qRT-PCR, stromal co-culture, confocal fluorescence microscopic, nanoproteomic, serial transplantation and cell cycle analyses were performed on FACS purified normal, chronic phase (CP) chronic myeloid leukemia (CML), blast crisis (BC) phase CML progenitors with or without PF-04449913 treatment. RESULTS: Notably, RNA-seq analyses revealed that Hh pathway and cell cycle regulatory gene overexpression correlated with leukemic progression. While lentivirally enforced GLI2 expression enhanced leukemic progenitor dormancy in stromal co-cultures, this was not observed with a mutant GLI2 lacking a transactivation domain, suggesting that GLI2 expression prevented cell cycle transit. Selective SMO inhibition with PF-04449913 in humanized stromal co-cultures and LSC xenografts reduced downstream GLI2 protein and cell cycle regulatory gene expression. Moreover, SMO inhibition enhanced cell cycle transit and sensitized BC LSC to tyrosine kinase inhibition in vivo at doses that spare normal HSC. CONCLUSION: In summary, while GLI2, forms part of a core HH pathway transcriptional regulatory network that promotes human myeloid leukemic progression and dormant LSC generation, selective inhibition with PF-04449913 reduces the dormant LSC burden thereby providing a strong rationale for clinical trials predicated on SMO inhibition in combination with TKIs or chemotherapeutic agents with the ultimate aim of obviating leukemic therapeutic resistance, persistence and progression.

Automated Identification and Measurement Extraction of Pancreatic Cystic Lesions from Free-Text Radiology Reports Using Natural Language Processing.

Purpose: To automatically identify a cohort of patients with pancreatic cystic lesions (PCLs) and extract PCL measurements from historical CT and MRI reports using natural language processing (NLP) and a question answering system. Materials and Methods: Institutional review board approval was obtained for this retrospective Health Insurance Portability and Accountability Act-compliant study, and the requirement to obtain informed consent was waived. A cohort of free-text CT and MRI reports generated between January 1991 and July 2019 that covered the pancreatic region were identified. A PCL identification model was developed by modifying a rule-based information extraction model; measurement extraction was performed using a state-of-the-art question answering system. The system's performance was evaluated against radiologists' annotations. Results: For this study, 430 426 free-text radiology reports from 199 783 unique patients were identified. The NLP model for identifying PCL was applied to 1000 test samples. The interobserver agreement between the model and two radiologists was almost perfect (Fleiss κ = 0.951), and the false-positive rate and true-positive rate were 3.0% and 98.2%, respectively, against consensus of radiologists' annotations as ground truths. The overall accuracy and Lin concordance correlation coefficient for measurement extraction were 0.958 and 0.874, respectively, against radiologists' annotations as ground truths. Conclusion: An NLP-based system was developed that identifies patients with PCLs and extracts measurements from a large single-institution archive of free-text radiology reports. This approach may prove valuable to study the natural history and potential risks of PCLs and can be applied to many other use cases.Keywords: Informatics, Abdomen/GI, Pancreas, Cysts, Computer Applications-General (Informatics), Named Entity Recognition Supplemental material is available for this article. © RSNA, 2022See also commentary by Horii in this issue.

Toward Reduction in False-Positive Thyroid Nodule Biopsies with a Deep Learning-based Risk Stratification System Using US Cine-Clip Images.

Purpose: To develop a deep learning-based risk stratification system for thyroid nodules using US cine images. Materials and Methods: In this retrospective study, 192 biopsy-confirmed thyroid nodules (175 benign, 17 malignant) in 167 unique patients (mean age, 56 years ± 16 [SD], 137 women) undergoing cine US between April 2017 and May 2018 with American College of Radiology (ACR) Thyroid Imaging Reporting and Data System (TI-RADS)-structured radiology reports were evaluated. A deep learning-based system that exploits the cine images obtained during three-dimensional volumetric thyroid scans and outputs malignancy risk was developed and compared, using fivefold cross-validation, against a two-dimensional (2D) deep learning-based model (Static-2DCNN), a radiomics-based model using cine images (Cine-Radiomics), and the ACR TI-RADS level, with histopathologic diagnosis as ground truth. The system was used to revise the ACR TI-RADS recommendation, and its diagnostic performance was compared against the original ACR TI-RADS. Results: The system achieved higher average area under the receiver operating characteristic curve (AUC, 0.88) than Static-2DCNN (0.72, P = .03) and tended toward higher average AUC than Cine-Radiomics (0.78, P = .16) and ACR TI-RADS level (0.80, P = .21). The system downgraded recommendations for 92 benign and two malignant nodules and upgraded none. The revised recommendation achieved higher specificity (139 of 175, 79.4%) than the original ACR TI-RADS (47 of 175, 26.9%; P < .001), with no difference in sensitivity (12 of 17, 71% and 14 of 17, 82%, respectively; P = .63). Conclusion: The risk stratification system using US cine images had higher diagnostic performance than prior models and improved specificity of ACR TI-RADS when used to revise ACR TI-RADS recommendation.Keywords: Neural Networks, US, Abdomen/GI, Head/Neck, Thyroid, Computer Applications-3D, Oncology, Diagnosis, Supervised Learning, Transfer Learning, Convolutional Neural Network (CNN) Supplemental material is available for this article. © RSNA, 2022.

A novel patient-derived intra-femoral xenograft model of bone metastatic prostate cancer that recapitulates mixed osteolytic and osteoblastic lesions.

UNLABELLED: Prostate cancer metastasizes to bone in the majority of patients with advanced disease leading to painfully debilitating fractures, spinal compression and rapid decline. In addition, prostate cancer bone metastases often become resistant to standard therapies including androgen deprivation, radiation and chemotherapy. There are currently few models to elucidate mechanisms of interaction between the bone microenvironment and prostate cancer. It is, thus, essential to develop new patient-derived, orthotopic models. Here we report the development and characterization of PCSD1 (Prostate Cancer San Diego 1), a novel patient-derived intra-femoral xenograft model of prostate bone metastatic cancer that recapitulates mixed osteolytic and osteoblastic lesions. METHODS: A femoral bone metastasis of prostate cancer was removed during hemiarthroplasty and transplanted into Rag2(-/-);γc(-/-) mice either intra-femorally or sub-cutaneously. Xenograft tumors that developed were analyzed for prostate cancer biomarker expression using RT-PCR and immunohistochemistry. Osteoblastic, osteolytic and mixed lesion formation was measured using micro-computed tomography (microCT). RESULTS: PCSD1 cells isolated directly from the patient formed tumors in all mice that were transplanted intra-femorally or sub-cutaneously into Rag2(-/-);γc(-/-) mice. Xenograft tumors expressed human prostate specific antigen (PSA) in RT-PCR and immunohistochemical analyses. PCSD1 tumors also expressed AR, NKX3.1, Keratins 8 and 18, and AMACR. Histologic and microCT analyses revealed that intra-femoral PCSD1 xenograft tumors formed mixed osteolytic and osteoblastic lesions. PCSD1 tumors have been serially passaged in mice as xenografts intra-femorally or sub-cutaneously as well as grown in culture. CONCLUSIONS: PCSD1 xenografts tumors were characterized as advanced, luminal epithelial prostate cancer from a bone metastasis using RT-PCR and immunohistochemical biomarker analyses. PCSD1 intra-femoral xenografts formed mixed osteoblastic/osteolytic lesions that closely resembled the bone lesions in the patient. PCSD1 is a new primary prostate cancer bone metastasis-derived xenograft model to study metastatic disease in the bone and to develop novel therapies for inhibiting prostate cancer growth in the bone-niche.

ADAR1 Activation Drives Leukemia Stem Cell Self-Renewal by Impairing Let-7 Biogenesis.

Post-transcriptional adenosine-to-inosine RNA editing mediated by adenosine deaminase acting on RNA1 (ADAR1) promotes cancer progression and therapeutic resistance. However, ADAR1 editase-dependent mechanisms governing leukemia stem cell (LSC) generation have not been elucidated. In blast crisis chronic myeloid leukemia (BC CML), we show that increased JAK2 signaling and BCR-ABL1 amplification activate ADAR1. In a humanized BC CML mouse model, combined JAK2 and BCR-ABL1 inhibition prevents LSC self-renewal commensurate with ADAR1 downregulation. Lentiviral ADAR1 wild-type, but not an editing-defective ADAR1(E912A) mutant, induces self-renewal gene expression and impairs biogenesis of stem cell regulatory let-7 microRNAs. Combined RNA sequencing, qRT-PCR, CLIP-ADAR1, and pri-let-7 mutagenesis data suggest that ADAR1 promotes LSC generation via let-7 pri-microRNA editing and LIN28B upregulation. A small-molecule tool compound antagonizes ADAR1's effect on LSC self-renewal in stromal co-cultures and restores let-7 biogenesis. Thus, ADAR1 activation represents a unique therapeutic vulnerability in LSCs with active JAK2 signaling.

Transcriptome sequencing of tumor subpopulations reveals a spectrum of therapeutic options for squamous cell lung cancer.

BACKGROUND: The only therapeutic options that exist for squamous cell lung carcinoma (SCC) are standard radiation and cytotoxic chemotherapy. Cancer stem cells (CSCs) are hypothesized to account for therapeutic resistance, suggesting that CSCs must be specifically targeted. Here, we analyze the transcriptome of CSC and non-CSC subpopulations by RNA-seq to identify new potential therapeutic strategies for SCC. METHODS: We sorted a SCC into CD133- and CD133+ subpopulations and then examined both by copy number analysis (CNA) and whole genome and transcriptome sequencing. We analyzed The Cancer Genome Atlas (TCGA) transcriptome data of 221 SCCs to determine the generality of our observations. RESULTS: Both subpopulations highly expressed numerous mRNA isoforms whose protein products are active drug targets for other cancers; 31 (25%) correspond to 18 genes under active investigation as mAb targets and an additional 4 (3%) are of therapeutic interest. Moreover, we found evidence that both subpopulations were proliferatively driven by very high levels of c-Myc and the TRAIL long isoform (TRAILL) and that normal apoptotic responses to high expression of these genes was prevented through high levels of Mcl-1L and Bcl-xL and c-FlipL-isoforms for which drugs are now in clinical development. SCC RNA-seq data (n = 221) from TCGA supported our findings. Our analysis is inconsistent with the CSC concept that most cells in a cancer have lost their proliferative potential. Furthermore, our study suggests how to target both the CSC and non-CSC subpopulations with one treatment strategy. CONCLUSIONS: Our study is relevant to SCC in particular for it presents numerous potential options to standard therapy that target the entire tumor. In so doing, it demonstrates how transcriptome sequencing provides insights into the molecular underpinnings of cancer propagating cells that, importantly, can be leveraged to identify new potential therapeutic options for cancers beyond what is possible with DNA sequencing.

A Pan-BCL2 inhibitor renders bone-marrow-resident human leukemia stem cells sensitive to tyrosine kinase inhibition.

Leukemia stem cells (LSCs) play a pivotal role in the resistance of chronic myeloid leukemia (CML) to tyrosine kinase inhibitors (TKIs) and its progression to blast crisis (BC), in part, through the alternative splicing of self-renewal and survival genes. To elucidate splice-isoform regulators of human BC LSC maintenance, we performed whole-transcriptome RNA sequencing, splice-isoform-specific quantitative RT-PCR (qRT-PCR), nanoproteomics, stromal coculture, and BC LSC xenotransplantation analyses. Cumulatively, these studies show that the alternative splicing of multiple prosurvival BCL2 family genes promotes malignant transformation of myeloid progenitors into BC LSCS that are quiescent in the marrow niche and that contribute to therapeutic resistance. Notably, sabutoclax, a pan-BCL2 inhibitor, renders marrow-niche-resident BC LSCs sensitive to TKIs at doses that spare normal progenitors. These findings underscore the importance of alternative BCL2 family splice-isoform expression in BC LSC maintenance and suggest that the combinatorial inhibition of prosurvival BCL2 family proteins and BCR-ABL may eliminate dormant LSCs and obviate resistance.

NOTCH1 signaling promotes human T-cell acute lymphoblastic leukemia initiating cell regeneration in supportive niches.

BACKGROUND: Leukemia initiating cells (LIC) contribute to therapeutic resistance through acquisition of mutations in signaling pathways, such as NOTCH1, that promote self-renewal and survival within supportive niches. Activating mutations in NOTCH1 occur commonly in T cell acute lymphoblastic leukemia (T-ALL) and have been implicated in therapeutic resistance. However, the cell type and context specific consequences of NOTCH1 activation, its role in human LIC regeneration, and sensitivity to NOTCH1 inhibition in hematopoietic microenvironments had not been elucidated. METHODOLOGY AND PRINCIPAL FINDINGS: We established humanized bioluminescent T-ALL LIC mouse models transplanted with pediatric T-ALL samples that were sequenced for NOTCH1 and other common T-ALL mutations. In this study, CD34(+) cells from NOTCH1(Mutated) T-ALL samples had higher leukemic engraftment and serial transplantation capacity than NOTCH1(Wild-type) CD34(+) cells in hematopoietic niches, suggesting that self-renewing LIC were enriched within the NOTCH1(Mutated) CD34(+) fraction. Humanized NOTCH1 monoclonal antibody treatment reduced LIC survival and self-renewal in NOTCH1(Mutated) T-ALL LIC-engrafted mice and resulted in depletion of CD34(+)CD2(+)CD7(+) cells that harbor serial transplantation capacity. CONCLUSIONS: These results reveal a functional hierarchy within the LIC population based on NOTCH1 activation, which renders LIC susceptible to targeted NOTCH1 inhibition and highlights the utility of NOTCH1 antibody targeting as a key component of malignant stem cell eradication strategies.

Cycling toward elimination of leukemic stem cells.

In two recent articles in Science Translational Medicine and Nature Biotechnology, Saito et al. (2010) identify a molecular signature of acute myeloid leukemia (AML) stem cells and demonstrate that quiescent AML stem cells become sensitized to chemotherapy after G-CSF stimulation.

Chronic myeloid leukemia stem cells.

Although rare, chronic myeloid leukemia (CML) represents an important paradigm for understanding the molecular events leading to malignant transformation of primitive hematopoietic progenitors. CML was the first cancer to be associated with a defined genetic abnormality, BCR-ABL, that is necessary and sufficient for initiating chronic phase disease as well as the first cancer to be treated with molecular targeted therapy. Malignant progenitors or leukemia stem cells (LSCs) evolve as a result of both epigenetic and genetic events that alter hematopoietic progenitor differentiation, proliferation, survival, and self-renewal. LSCs are rare and divide less frequently, and thus, represent a reservoir for relapse and resistance to a molecularly targeted single agent. On subverting developmental processes normally responsible for maintaining robust life-long hematopoiesis, the LSCs are able to evade the majority of current cancer treatments that target rapidly dividing cells. Enthusiasm for the enormous success of tyrosine kinase inhibitors at controlling the chronic phase disease is tempered somewhat by the persistence of the LSC pool in the majority of the patients. Combined therapies targeting aberrant properties of LSC may obviate therapeutic resistance and relapse in advanced phase and therapeutically recalcitrant CML.

The interferon-induced protein BST-2 restricts HIV-1 release and is downregulated from the cell surface by the viral Vpu protein.

The HIV-1 accessory protein Vpu counteracts a host factor that restricts virion release from infected cells. Here we show that the interferon-induced cellular protein BST-2/HM1.24/CD317 is such a factor. BST-2 is downregulated from the cell surface by Vpu, and BST-2 is specifically expressed in cells that support the vpu phenotype. Exogenous expression of BST-2 inhibits HIV-1 virion release, while suppression of BST-2 relieves the requirement for Vpu. Downregulation of BST-2 requires both the transmembrane/ion channel domain and conserved serines in the cytoplasmic domain of Vpu. Endogenous BST-2 colocalizes with the HIV-1 structural protein Gag in endosomes and at the plasma membrane, suggesting that BST-2 traps virions within and on infected cells. The unusual structure of BST-2, which includes a transmembrane domain and a lumenal GPI anchor, may allow it to retain nascent enveloped virions on cellular membranes, providing a mechanism of viral restriction counteracted by a specific viral accessory protein.