Systematic Literature Review of Machine Learning Algorithms Using Pretherapy Radiologic Imaging for Glioma Molecular Subtype Prediction.

BACKGROUND: The molecular profile of gliomas is a prognostic indicator for survival, driving clinical decision-making for treatment. Pathology-based molecular diagnosis is challenging because of the invasiveness of the procedure, exclusion from neoadjuvant therapy options, and the heterogeneous nature of the tumor. PURPOSE: We performed a systematic review of algorithms that predict molecular subtypes of gliomas from MR Imaging. DATA SOURCES: Data sources were Ovid Embase, Ovid MEDLINE, Cochrane Central Register of Controlled Trials, Web of Science. STUDY SELECTION: Per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, 12,318 abstracts were screened and 1323 underwent full-text review, with 85 articles meeting the inclusion criteria. DATA ANALYSIS: We compared prediction results from different machine learning approaches for predicting molecular subtypes of gliomas. Bias analysis was conducted for each study, following the Prediction model Risk Of Bias Assessment Tool (PROBAST) guidelines. DATA SYNTHESIS: Isocitrate dehydrogenase mutation status was reported with an area under the curve and accuracy of 0.88 and 85% in internal validation and 0.86 and 87% in limited external validation data sets, respectively. For the prediction of O6-methylguanine-DNA methyltransferase promoter methylation, the area under the curve and accuracy in internal validation data sets were 0.79 and 77%, and in limited external validation, 0.89 and 83%, respectively. PROBAST scoring demonstrated high bias in all articles. LIMITATIONS: The low number of external validation and studies with incomplete data resulted in unequal data analysis. Comparing the best prediction pipelines of each study may introduce bias. CONCLUSIONS: While the high area under the curve and accuracy for the prediction of molecular subtypes of gliomas are reported in internal and external validation data sets, limited use of external validation and the increased risk of bias in all articles may present obstacles for clinical translation of these techniques.

Clinical implementation of artificial intelligence in neuroradiology with development of a novel workflow-efficient picture archiving and communication system-based automated brain tumor segmentation and radiomic feature extraction.

Purpose: Personalized interpretation of medical images is critical for optimum patient care, but current tools available to physicians to perform quantitative analysis of patient's medical images in real time are significantly limited. In this work, we describe a novel platform within PACS for volumetric analysis of images and thus development of large expert annotated datasets in parallel with radiologist performing the reading that are critically needed for development of clinically meaningful AI algorithms. Specifically, we implemented a deep learning-based algorithm for automated brain tumor segmentation and radiomics extraction, and embedded it into PACS to accelerate a supervised, end-to- end workflow for image annotation and radiomic feature extraction. Materials and methods: An algorithm was trained to segment whole primary brain tumors on FLAIR images from multi-institutional glioma BraTS 2021 dataset. Algorithm was validated using internal dataset from Yale New Haven Health (YHHH) and compared (by Dice similarity coefficient [DSC]) to radiologist manual segmentation. A UNETR deep-learning was embedded into Visage 7 (Visage Imaging, Inc., San Diego, CA, United States) diagnostic workstation. The automatically segmented brain tumor was pliable for manual modification. PyRadiomics (Harvard Medical School, Boston, MA) was natively embedded into Visage 7 for feature extraction from the brain tumor segmentations. Results: UNETR brain tumor segmentation took on average 4 s and the median DSC was 86%, which is similar to published literature but lower than the RSNA ASNR MICCAI BRATS challenge 2021. Finally, extraction of 106 radiomic features within PACS took on average 5.8 ± 0.01 s. The extracted radiomic features did not vary over time of extraction or whether they were extracted within PACS or outside of PACS. The ability to perform segmentation and feature extraction before radiologist opens the study was made available in the workflow. Opening the study in PACS, allows the radiologists to verify the segmentation and thus annotate the study. Conclusion: Integration of image processing algorithms for tumor auto-segmentation and feature extraction into PACS allows curation of large datasets of annotated medical images and can accelerate translation of research into development of personalized medicine applications in the clinic. The ability to use familiar clinical tools to revise the AI segmentations and natively embedding the segmentation and radiomic feature extraction tools on the diagnostic workstation accelerates the process to generate ground-truth data.

Identifying clinically applicable machine learning algorithms for glioma segmentation: recent advances and discoveries.

Background: While there are innumerable machine learning (ML) research algorithms used for segmentation of gliomas, there is yet to be a US FDA cleared product. The aim of this study is to explore the systemic limitations of research algorithms that have prevented translation from concept to product by a review of the current research literature. Methods: We performed a systematic literature review on 4 databases. Of 11 727 articles, 58 articles met the inclusion criteria and were used for data extraction and screening using TRIPOD. Results: We found that while many articles were published on ML-based glioma segmentation and report high accuracy results, there were substantial limitations in the methods and results portions of the papers that result in difficulty reproducing the methods and translation into clinical practice. Conclusions: In addition, we identified that more than a third of the articles used the same publicly available BRaTS and TCIA datasets and are responsible for the majority of patient data on which ML algorithms were trained, which leads to limited generalizability and potential for overfitting and bias.

Machine Learning Models for Classifying High- and Low-Grade Gliomas: A Systematic Review and Quality of Reporting Analysis.

Objectives: To systematically review, assess the reporting quality of, and discuss improvement opportunities for studies describing machine learning (ML) models for glioma grade prediction. Methods: This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses of Diagnostic Test Accuracy (PRISMA-DTA) statement. A systematic search was performed in September 2020, and repeated in January 2021, on four databases: Embase, Medline, CENTRAL, and Web of Science Core Collection. Publications were screened in Covidence, and reporting quality was measured against the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Statement. Descriptive statistics were calculated using GraphPad Prism 9. Results: The search identified 11,727 candidate articles with 1,135 articles undergoing full text review and 85 included in analysis. 67 (79%) articles were published between 2018-2021. The mean prediction accuracy of the best performing model in each study was 0.89 ± 0.09. The most common algorithm for conventional machine learning studies was Support Vector Machine (mean accuracy: 0.90 ± 0.07) and for deep learning studies was Convolutional Neural Network (mean accuracy: 0.91 ± 0.10). Only one study used both a large training dataset (n>200) and external validation (accuracy: 0.72) for their model. The mean adherence rate to TRIPOD was 44.5% ± 11.1%, with poor reporting adherence for model performance (0%), abstracts (0%), and titles (0%). Conclusions: The application of ML to glioma grade prediction has grown substantially, with ML model studies reporting high predictive accuracies but lacking essential metrics and characteristics for assessing model performance. Several domains, including generalizability and reproducibility, warrant further attention to enable translation into clinical practice. Systematic Review Registration: PROSPERO, identifier CRD42020209938.

Machine Learning Applications for Differentiation of Glioma from Brain Metastasis-A Systematic Review.

Glioma and brain metastasis can be difficult to distinguish on conventional magnetic resonance imaging (MRI) due to the similarity of imaging features in specific clinical circumstances. Multiple studies have investigated the use of machine learning (ML) models for non-invasive differentiation of glioma from brain metastasis. Many of the studies report promising classification results, however, to date, none have been implemented into clinical practice. After a screening of 12,470 studies, we included 29 eligible studies in our systematic review. From each study, we aggregated data on model design, development, and best classifiers, as well as quality of reporting according to the TRIPOD statement. In a subset of eligible studies, we conducted a meta-analysis of the reported AUC. It was found that data predominantly originated from single-center institutions (n = 25/29) and only two studies performed external validation. The median TRIPOD adherence was 0.48, indicating insufficient quality of reporting among surveyed studies. Our findings illustrate that despite promising classification results, reliable model assessment is limited by poor reporting of study design and lack of algorithm validation and generalizability. Therefore, adherence to quality guidelines and validation on outside datasets is critical for the clinical translation of ML for the differentiation of glioma and brain metastasis.

The Utility of Imaging Parameters in Predicting Long-Term Clinical Improvement After Shunt Surgery in Patients with Idiopathic Normal Pressure Hydrocephalus.

OBJECTIVE: It is difficult to predict which patients with idiopathic normal pressure hydrocephalus (iNPH) will improve after shunt surgery. This study investigated the association between preoperative imaging parameters in patients with iNPH and long-term outcome after shunt placement. METHODS: Patients with iNPH who showed a response to large-volume cerebrospinal fluid drainage and subsequently underwent ventriculoperitoneal shunt surgery were reviewed. Long-term patient-reported outcomes were obtained by telephone interview. Preoperative computed tomography and/or magnetic resonance imaging were retrospectively reviewed to determine associations between imaging parameters and clinical outcome. RESULTS: The final analysis included 37 patients. The median duration between shunt surgery and telephone interview was 30 months (range, 12-56 months). Gait improvement after shunting was present more often in patients without focally dilated sulci (95% vs. 71%, P = 0.04), but a statistically significant relationship was not established after logistic regression. Patients with cognitive improvement after shunting had a higher preoperative Evans index (mean 0.41 vs. 0.36, P < 0.01), and Evans index was a predictor of cognitive improvement (odds ratio = 1.40, scale of 0.01, P = 0.01). CONCLUSIONS: Higher Evans index is a predictor of long-term cognitive improvement after shunt placement; however, no cutoff value demonstrates sufficient accuracy for the selection of shunt candidates. None of the evaluated imaging features was predictive of long-term gait or urinary improvement. The utility of imaging to predict a response to shunting is limited, and no imaging feature alone can be used to exclude patients from shunt surgery.

Trends in Development of Novel Machine Learning Methods for the Identification of Gliomas in Datasets That Include Non-Glioma Images: A Systematic Review.

PURPOSE: Machine learning has been applied to the diagnostic imaging of gliomas to augment classification, prognostication, segmentation, and treatment planning. A systematic literature review was performed to identify how machine learning has been applied to identify gliomas in datasets which include non-glioma images thereby simulating normal clinical practice. MATERIALS AND METHODS: Four databases were searched by a medical librarian and confirmed by a second librarian for all articles published prior to February 1, 2021: Ovid Embase, Ovid MEDLINE, Cochrane trials (CENTRAL), and Web of Science-Core Collection. The search strategy included both keywords and controlled vocabulary combining the terms for: artificial intelligence, machine learning, deep learning, radiomics, magnetic resonance imaging, glioma, as well as related terms. The review was conducted in stepwise fashion with abstract screening, full text screening, and data extraction. Quality of reporting was assessed using TRIPOD criteria. RESULTS: A total of 11,727 candidate articles were identified, of which 12 articles were included in the final analysis. Studies investigated the differentiation of normal from abnormal images in datasets which include gliomas (7 articles) and the differentiation of glioma images from non-glioma or normal images (5 articles). Single institution datasets were most common (5 articles) followed by BRATS (3 articles). The median sample size was 280 patients. Algorithm testing strategies consisted of five-fold cross validation (5 articles), and the use of exclusive sets of images within the same dataset for training and for testing (7 articles). Neural networks were the most common type of algorithm (10 articles). The accuracy of algorithms ranged from 0.75 to 1.00 (median 0.96, 10 articles). Quality of reporting assessment utilizing TRIPOD criteria yielded a mean individual TRIPOD ratio of 0.50 (standard deviation 0.14, range 0.37 to 0.85). CONCLUSION: Systematic review investigating the identification of gliomas in datasets which include non-glioma images demonstrated multiple limitations hindering the application of these algorithms to clinical practice. These included limited datasets, a lack of generalizable algorithm training and testing strategies, and poor quality of reporting. The development of more robust and heterogeneous datasets is needed for algorithm development. Future studies would benefit from using external datasets for algorithm testing as well as placing increased attention on quality of reporting standards. SYSTEMATIC REVIEW REGISTRATION: www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42020209938, International Prospective Register of Systematic Reviews (PROSPERO 2020 CRD42020209938).

The Subjective Experience of Patients Undergoing Shunt Surgery for Idiopathic Normal Pressure Hydrocephalus.

OBJECTIVE: The counseling of patients with idiopathic normal pressure hydrocephalus (iNPH) is difficult; there is variability in the diagnostic criteria, and a definitive diagnosis can be made only postoperatively. A patient's clinical response to shunting is also difficult to predict. This study examines the subjective experience of patients treated for iNPH, to identify the challenges patients face and to improve patient outcomes and satisfaction. METHODS: We reviewed a consecutive series of patients diagnosed with iNPH who underwent ventriculoperitoneal shunt surgery between January 2012 and March 2016 at our institution. Semistructured telephone interviews were conducted with 31 patients. Interviews were analyzed using the principles of grounded theory. RESULTS: Thirty-one patients who underwent shunt surgery for iNPH were interviewed to reach saturation of themes. Seven themes were identified: 1) long preoperative course causes morbidity; 2) the decision to have shunt surgery is easy to make; 3) patients primarily desire to gain independence; 4) patients show variable levels of anxiety; 5) comorbid conditions interfere with postoperative assessment; 6) patients stand by their decision to have shunt surgery; and 7) outside information is used before surgery. CONCLUSIONS: Patients often present to the neurosurgeon frustrated and desperate after a long preoperative course. It is important to acknowledge the uncertainty regarding diagnosis and response to shunting when counseling patients. Comorbid conditions interfere with the ability to assess progression of iNPH and the effectiveness of the shunt. Patient caregivers play a large role in decision making and clinical course and should be included when counseling patients.

Pretreatment predictors of adjuvant chemoradiation in patients receiving transoral robotic surgery for squamous cell carcinoma of the oropharynx: a case control study.

BACKGROUND: The purpose of this study was to identify preoperative patient characteristics associated with the incidence of positive surgical margins or lymph node extracapsular extension (ECE), which necessitate adjuvant chemoradiation after transoral robotic surgery (TORS). METHODS: We conducted a single institution retrospective study of 34 consecutive patients with primary oropharyngeal cancer who underwent TORS. All imaging was reviewed by a single neuroradiologist. Surgical margins and ECE status were determined by a single head and neck pathologist. Associations of preoperative patient characteristics with positive surgical margins and lymph node ECE were examined using univariate analysis. Independent predictors of these outcomes were determined using logistic regression. RESULTS: Preoperatively, the majority of patients had early-stage disease (7 cT1 and 21 cT2; 10 cN0). Positive margins occurred in 4 (12 %) patients. A clinically positive lymph node was seen in 23 (68 %) patients. Neck dissection was performed in 29 (85 %) patients, among whom 19 had a pathologically positive lymph node and 15 had nodal ECE. Logistic regression showed that larger preoperative lymph node size was an independent predictor of ECE (odds ratio, 13.32 [95 % CI, 1.46-121.43]). Among the 21 patients with a clinically positive lymph node who underwent neck dissection, ECE was present more often in patients with a preoperative node size ≥ 3.0 vs. < 3.0 cm (92 % vs. 44 %, P = 0.046). There was no patient characteristic associated with positive margins. CONCLUSIONS: Patients with a larger preoperative lymph node appear more likely to have ECE, and thus be treated with chemoradiation after TORS, with a potentially higher rate of toxicity. Lymph node size should be taken into account when deciding upon treatment approaches. Further research is needed to validate these results.

Microbiota-Dependent Priming of Antiviral Intestinal Immunity in Drosophila.

Enteric pathogens must overcome intestinal defenses to establish infection. In Drosophila, the ERK signaling pathway inhibits enteric virus infection. The intestinal microflora also impacts immunity but its role in enteric viral infection is unknown. Here we show that two signals are required to activate antiviral ERK signaling in the intestinal epithelium. One signal depends on recognition of peptidoglycan from the microbiota, particularly from the commensal Acetobacter pomorum, which primes the NF-kB-dependent induction of a secreted factor, Pvf2. However, the microbiota is not sufficient to induce this pathway; a second virus-initiated signaling event involving release of transcriptional paused genes mediated by the kinase Cdk9 is also required for Pvf2 production. Pvf2 stimulates antiviral immunity by binding to the receptor tyrosine kinase PVR, which is necessary and sufficient for intestinal ERK responses. These findings demonstrate that sensing of specific commensals primes inflammatory signaling required for epithelial responses that restrict enteric viral infections.

IGFBP3 promotes esophageal cancer growth by suppressing oxidative stress in hypoxic tumor microenvironment.

Insulin-like growth factor binding protein 3 (IGFBP3), a hypoxia-inducible gene, regulates a variety of cellular processes including cell proliferation, senescence, apoptosis and epithelial-mesenchymal transition (EMT). IGFBP3 has been linked to the pathogenesis of cancers. Most previous studies focus upon proapoptotic tumor suppressor activities of IGFBP3. Nevertheless, IGFBP3 is overexpressed in certain cancers including esophageal squamous cell carcinoma (ESCC), one of the most aggressive forms of squamous cell carcinomas (SCCs). The tumor-promoting activities of IGFBP3 remain poorly understood in part due to a lack of understanding as to how the tumor microenvironment may influence IGFBP3 expression and how IGFBP3 may in turn influence heterogeneous intratumoral cell populations. Here, we show that IGFBP3 overexpression is associated with poor postsurgical prognosis in ESCC patients. In xenograft transplantation models with genetically engineered ESCC cells, IGFBP3 contributes to tumor progression with a concurrent induction of a subset of tumor cells showing high expression of CD44 (CD44H), a major cell surface receptor for hyaluronic acid, implicated in invasion, metastasis and drug resistance. Our gain-of-function and loss-of-function experiments reveal that IGFBP3 mediates the induction of intratumoral CD44H cells. IGFBP3 cooperates with hypoxia to mediate the induction of CD44H cells by suppressing reactive oxygen species (ROS) in an insulin-like growth factor-independent fashion. Thus, our study sheds light on the growth stimulatory functions of IGFPB3 in cancer, gaining a novel mechanistic insight into the functional interplay between the tumor microenvironment and IGFBP3.

ERK signaling couples nutrient status to antiviral defense in the insect gut.

A unique facet of arthropod-borne virus (arbovirus) infection is that the pathogens are orally acquired by an insect vector during the taking of a blood meal, which directly links nutrient acquisition and pathogen challenge. We show that the nutrient responsive ERK pathway is both induced by and restricts disparate arboviruses in Drosophila intestines, providing insight into the molecular determinants of the antiviral "midgut barrier." Wild-type flies are refractory to oral infection by arboviruses, including Sindbis virus and vesicular stomatitis virus, but this innate restriction can be overcome chemically by oral administration of an ERK pathway inhibitor or genetically via the specific loss of ERK in Drosophila intestinal epithelial cells. In addition, we found that vertebrate insulin, which activates ERK in the mosquito gut during a blood meal, restricts viral infection in Drosophila cells and against viral invasion of the insect gut epithelium. We find that ERK's antiviral signaling activity is likely conserved in Aedes mosquitoes, because genetic or pharmacologic manipulation of the ERK pathway affects viral infection of mosquito cells. These studies demonstrate that ERK signaling has a broadly antiviral role in insects and suggest that insects take advantage of cross-species signals in the meal to trigger antiviral immunity.

Notch receptor inhibition reveals the importance of cyclin D1 and Wnt signaling in invasive esophageal squamous cell carcinoma.

Esophageal squamous cell carcinoma (ESCC) is one of the most aggressive forms of squamous cell carcinomas. Common genetic lesions in ESCC include p53 mutations and EGFR overexpression, both of which have been implicated in negative regulation of Notch signaling. In addition, cyclin D1 is overexpressed in ESCC and can be activated via EGFR, Notch and Wnt signaling. To elucidate how these genetic lesions may interact during the development and progression of ESCC, we tested a panel of genetically engineered human esophageal cells (keratinocytes) in organotypic 3D culture (OTC), a form of human tissue engineering. Notch signaling was suppressed in culture and mice by dominant negative Mastermind-like1 (DNMAML1), a genetic pan-Notch inhibitor. DNMAML1 mice were subjected to 4-Nitroquinoline 1-oxide-induced oral-esophageal carcinogenesis. Highly invasive characteristics of primary human ESCC were recapitulated in OTC as well as DNMAML1 mice. In OTC, cyclin D1 overexpression induced squamous hyperplasia. Concurrent EGFR overexpression and mutant p53 resulted in transformation and invasive growth. Interestingly, cell proliferation appeared to be regulated differentially between those committed to squamous-cell differentiation and those invading into the stroma. Invasive cells exhibited Notch-independent activation of cyclin D1 and Wnt signaling. Within the oral-esophageal squamous epithelia, Notch signaling regulated squamous-cell differentiation to maintain epithelial integrity, and thus may act as a tumor suppressor by preventing the development of a tumor-promoting inflammatory microenvironment.

Hypoxia induces IGFBP3 in esophageal squamous cancer cells through HIF-1α-mediated mRNA transcription and continuous protein synthesis.

Insulin-like growth factor binding protein (IGFBP)-3 regulates cell proliferation and apoptosis in esophageal squamous cell carcinoma (ESCC) cells. We have investigated how the hypoxic tumor microenvironment in ESCC fosters the induction of IGFBP3. RNA interference experiments revealed that hypoxia-inducible factor (HIF)-1α, but not HIF-2α, regulates IGFBP3 mRNA induction. By chromatin immunoprecipitation and transfection assays, HIF-1α was found to transactivate IGFBP3 through a novel hypoxia responsive element (HRE) located at 57 kb upstream from the transcription start site. Metabolic labeling experiments demonstrated hypoxia-mediated inhibition of global protein synthesis. 7-Methyl GTP-cap binding assays suggested that hypoxia suppresses cap-dependent translation. Experiments using pharmacological inhibitors for mammalian target of rapamycin (mTOR) suggested that a relatively weak mTOR activity may be sufficient for cap-dependent translation of IGFBP3 under hypoxic conditions. Bicistronic RNA reporter transfection assays did not validate the possibility of an internal ribosome entry site as a potential mechanism for cap-independent translation for IGFBP3 mRNA. Finally, IGFBP3 mRNA was found enriched to the polysomes. In aggregate, our study establishes IGFBP3 as a direct HIF-1α target gene and that polysome enrichment of IGFBP3 mRNA may permit continuous translation under hypoxic conditions.