Multimodal data fusion using sparse canonical correlation analysis and cooperative learning: a COVID-19 cohort study.

Through technological innovations, patient cohorts can be examined from multiple views with high-dimensional, multiscale biomedical data to classify clinical phenotypes and predict outcomes. Here, we aim to present our approach for analyzing multimodal data using unsupervised and supervised sparse linear methods in a COVID-19 patient cohort. This prospective cohort study of 149 adult patients was conducted in a tertiary care academic center. First, we used sparse canonical correlation analysis (CCA) to identify and quantify relationships across different data modalities, including viral genome sequencing, imaging, clinical data, and laboratory results. Then, we used cooperative learning to predict the clinical outcome of COVID-19 patients: Intensive care unit admission. We show that serum biomarkers representing severe disease and acute phase response correlate with original and wavelet radiomics features in the LLL frequency channel (cor(Xu1, Zv1) = 0.596, p value < 0.001). Among radiomics features, histogram-based first-order features reporting the skewness, kurtosis, and uniformity have the lowest negative, whereas entropy-related features have the highest positive coefficients. Moreover, unsupervised analysis of clinical data and laboratory results gives insights into distinct clinical phenotypes. Leveraging the availability of global viral genome databases, we demonstrate that the Word2Vec natural language processing model can be used for viral genome encoding. It not only separates major SARS-CoV-2 variants but also allows the preservation of phylogenetic relationships among them. Our quadruple model using Word2Vec encoding achieves better prediction results in the supervised task. The model yields area under the curve (AUC) and accuracy values of 0.87 and 0.77, respectively. Our study illustrates that sparse CCA analysis and cooperative learning are powerful techniques for handling high-dimensional, multimodal data to investigate multivariate associations in unsupervised and supervised tasks.

Multimodal Biomedical Data Fusion Using Sparse Canonical Correlation Analysis and Cooperative Learning: A Cohort Study on COVID-19.

Through technological innovations, patient cohorts can be examined from multiple views with high-dimensional, multiscale biomedical data to classify clinical phenotypes and predict outcomes. Here, we aim to present our approach for analyzing multimodal data using unsupervised and supervised sparse linear methods in a COVID-19 patient cohort. This prospective cohort study of 149 adult patients was conducted in a tertiary care academic center. First, we used sparse canonical correlation analysis (CCA) to identify and quantify relationships across different data modalities, including viral genome sequencing, imaging, clinical data, and laboratory results. Then, we used cooperative learning to predict the clinical outcome of COVID-19 patients. We show that serum biomarkers representing severe disease and acute phase response correlate with original and wavelet radiomics features in the LLL frequency channel (corr(Xu1, Zv1) = 0.596, p-value < 0.001). Among radiomics features, histogram-based first-order features reporting the skewness, kurtosis, and uniformity have the lowest negative, whereas entropy-related features have the highest positive coefficients. Moreover, unsupervised analysis of clinical data and laboratory results gives insights into distinct clinical phenotypes. Leveraging the availability of global viral genome databases, we demonstrate that the Word2Vec natural language processing model can be used for viral genome encoding. It not only separates major SARS-CoV-2 variants but also allows the preservation of phylogenetic relationships among them. Our quadruple model using Word2Vec encoding achieves better prediction results in the supervised task. The model yields area under the curve (AUC) and accuracy values of 0.87 and 0.77, respectively. Our study illustrates that sparse CCA analysis and cooperative learning are powerful techniques for handling high-dimensional, multimodal data to investigate multivariate associations in unsupervised and supervised tasks.

Long-term follow-up results of multiparametric prostate MRI and the prognostic value of PI-RADS: a single-center retrospective cohort study.

PURPOSE: We aim to examine the long-term outcomes of patients who underwent multiparametric prostate magnetic resonance imaging (mp-MRI) for suspected prostate cancer (PCa), specifically based on their initial Prostate Imaging Reporting and Data System (PI-RADS) categories and various clinical factors. Our secondary aim is to evaluate the prognostic value of the PI-RADS through the National Comprehensive Cancer Network (NCCN) risk group distribution. METHODS: This research was conducted as a single-center retrospective cohort study in a tertiary care hospital. A total of 1,359 cases having at least one histopathological examination after the initial mp-MRI and/or adequate clinical/radiological follow-up data were included in the clinically significant PCa (cs-PCa) diagnosis-free survival analysis. Initial mp-MRI dates were accepted as the start of follow-up for the time-to-event analysis. The event was defined as cs-PCa diagnosis (International Society of Urological Pathology ≥2). Patients who were not diagnosed with cs-PCa during follow-up were censored according to predefined literature-based criteria at the end of the maximum follow-up duration with no reasonable suspicion of PCa and no biopsy indication. The impact of various factors on survival was assessed using a log-rank test and multivariable Cox regression. Subsequently, 394 cases diagnosed with PCa during follow-up were evaluated, based on initial PI-RADS categories and NCCN risk groups. RESULTS: Three main risk factors for cs-PCa diagnosis during follow-up were an initial PI-RADS 5 category, initial PI-RADS 4 category, and high MRI-defined PSA density (mPSAD), with average hazard ratios of 29.52, 14.46, and 3.12, respectively. The PI-RADS 3 category, advanced age group, and biopsy-naïve status were identified as additional risk factors (hazard ratios: 2.03, 1.54-1.98, and 1.79, respectively). In the PI-RADS 1-2 cohort, 1, 3, and 5-year cs-PCa diagnosis-free survival rates were 99.1%, 96.5%, and 93.8%, respectively. For the PI-RADS 3 cohort, 1, 3, and 5-year cs-PCa diagnosis-free survival rates were 94.9%, 90.9%, and 89.1%, respectively. For the PI-RADS 4 cohort, 1, 3, and 5-year cs-PCa diagnosis-free survival rates were 56.6%, 55.1%, and 55.1%, respectively. These rates were found to all be 24.2% in the PI-RADS 5 cohort. Considering the 394 cases diagnosed with PCa during follow-up, PI-RADS ≥4 cases were more likely to harbor unfavorable PCa compared to PI-RADS ≤3 cases (P < 0.001). In the PI-RADS 3 subgroup analysis, a low mPSAD (<0.15 ng/mL2) was found to be a protective prognostic factor against unfavorable PCa (P = 0.005). CONCLUSION: The PI-RADS category has a significant impact on patient management and provides important diagnostic and prognostic information. Higher initial PI-RADS categories are associated with decreased follow-up losses, a shorter time to PCa diagnosis, increased biopsy rates, a higher likelihood of developing cs-PCa during follow-up, and a worse PCa prognosis. Combining mPSAD with PI-RADS categories could enhance diagnostic stratification in the identification of cs-PCa.

Associations Between Hepatic and Pancreatic Steatosis with Lumbar Spinal Bone Marrow Fat: A Single-Center Magnetic Resonance Imaging Study.

BACKGROUND: To evaluate the associations between hepatic, pancreatic steatosis, and lumbar spinal bone marrow fat determined by magnetic resonance imaging-proton density fat fraction in patients with no known or suspected liver disease. METHODS: A total of 200 patients who were referred to our radiology department for upper abdominal magnetic resonance imaging between November 2015 and November 2017 were included in this study. All patients underwent a magnetic resonance imaging-proton density fat fraction on a 1.5-T magnetic resonance imaging system. RESULTS: The mean liver, pancreas, and lumbar magnetic resonance imaging-proton density fat fraction were 7.52 ± 4.82%, 5.25 ± 5.44%, and 46.85 ± 10.38% in the study population. There were significant correlations between liver and pancreas (rs = 0.180, P = .036), liver and lumbar (rs = 0.317, P < .001), and pancreas and lumbar magnetic resonance imaging-proton density fat fraction (rs = 0.215, P = .012) in female patients. A weak correlation was observed between liver and lumbar magnetic resonance imaging-proton density fat fraction (rs = 0.174, P = .014) in the total population. The prevalence of hepatic and pancreatic steatosis was 42.5% and 29%, respectively. The prevalence of pancreatic steatosis (42.9% vs. 22.8%, P = .004) was higher in male patients compared to female patients. In subgroup analysis, in patients with hepatic steatosis, there were higher pancreas magnetic resonance imaging-proton density fat fraction (6.07 ± 6.42% vs. 4.66 ± 4.53%, P = .036) and lumbar magnetic resonance imaging-proton density fat fraction (48.81 ± 10.01% vs. 45.40 ± 10.46%, P =.029) compared to patients without hepatic steatosis. In patients with pancreatic steatosis, there were higher liver (9.07 ± 6.08 vs. 6.87 ± 4.06, P = .009) and lumbar magnetic resonance imaging-proton density fat fraction (49.31 ± 9.13% vs.45.83 ± 10.76%, P = .032) in comparison with patients without pancreatic steatosis. CONCLUSION: Based on the results of the present study, fat accumulation in liver, pancreas, and lumbar vertebra have associations with more evident in females.

Infarcts and ischemia in the abdomen: an imaging perspective with an emphasis on cross-sectional imaging findings.

Infarcts and ischemia of abdominal organs may present with acute abdominal pain, and early diagnosis is crucial to prevent morbidity and mortality. Unfortunately, some of these patients present in poor clinical conditions to the emergency department, and imaging specialists are crucial for optimal outcomes. Although the radiological diagnosis of abdominal infarcts is often straightforward, it is vital to use the appropriate imaging modalities and correct imaging techniques for their detection. Additionally, some non-infarct-related abdominal pathologies may mimic infarcts, cause diagnostic confusion, and result in delayed diagnosis or misdiagnosis. In this article, we aimed to outline the general imaging approach, present cross-sectional imaging findings of infarcts and ischemia in several abdominal organs, including but not limited to, liver, spleen, kidneys, adrenals, omentum, and intestinal segments with relevant vascular anatomy, discuss possible differential diagnoses and emphasize important clinical/radiological clues that may assist radiologists in the diagnostic process.

A snapshot of teleradiology practice in Turkey: the results of a survey among radiologists.

PURPOSE: This study featured a survey that offers a snapshot of various teleradiology practices in Turkey, a Group of Twenty country that has undertaken a major transformation of its health care system during the last two decades and is currently the world leader in terms of the combined number of per capita magnetic resonance imaging and computed tomography examinations performed (which represent the bulk of teleradiology services worldwide). METHODS: The study data was collected from 4736 Turkish Society of Radiology (TSR) members via an electronic platform in the web environment through a questionnaire consisting of 24 questions. The survey was conducted in a 3-month time window (March-May 2021). Statistical tools were used for the analysis of the quantitative data. RESULTS: Responses from 156 members of the TSR comprised the study data, revealing that teleradiology is used for various applications in Turkey. Almost half of the participants (49%) performed teleradiology only in the private sector. Half of the respondents (51%) stated that they reported images at home for multiple centers. Moreover, 38% of the participants had been reporting more than 50 examinations per day, and 74% of the respondents earned less than 0.50 Euro per examination they reported. The overall satisfaction with teleradiology among the teleradiologists was, on average, 4.7 out of 10 points. CONCLUSION: The results are both promising for the future (i.e., concerning the propensity for adopting new technology) and alarming for the current state of affairs (i.e., insufficient radiologist reimbursement and lack of licensing and accreditation of teleradiology service providers). Periodic surveys performed in countries with different health care systems concerning financial, technical, and medicolegal aspects might reveal an up-to-date landscape of teleradiology practices worldwide and help guide local and regional decision-makers.

Imaging findings of spontaneous intraabdominal hemorrhage: neoplastic and non-neoplastic causes.

Contrary to traumatic and iatrogenic intraabdominal hemorrhages, spontaneous intraabdominal hemorrhage is a challenging clinical situation. A variety of neoplastic and non-neoplastic conditions may cause spontaneous intraabdominal bleeding. Imaging findings vary depending on the source of bleeding and the underlying cause. In this article, we aim to increase the awareness of imagers to the most common causes of spontaneous intraabdominal hemorrhage by using representative cases.

Identifying the deceiver: the non-neoplastic mimickers of genital system neoplasms.

Tumors of the genital system are common and imaging is of crucial importance for their detection and diagnosis. Several non-neoplastic diseases may mimic these tumors and differential diagnosis may be difficult in certain cases. Misdiagnosing non-neoplastic diseases as tumor may prompt unnecessary medical treatment or surgical interventions. In this article, we aimed to present the imaging characteristics of non-neoplastic diseases of the male and female genital systems that may mimic neoplastic processes. Increasing awareness of the imaging specialists to these entities may have a severe positive impact on the management of these patients.

TSR guidelines for the practice of teleradiology: 2021 update.

This update of Turkish Society of Radiology's (TSR) guidelines for the practice of teleradiology is intended to provide a reference framework for all parties involved in delivering imaging services away from the immediate vicinity of the patient. It includes relevant definitions and general principles, features organizational modes and qualifications of the practicing parties, lists technical issues, and addresses such management and legal aspects as archiving and documentation, security and privacy, reliability, responsibilities, quality inspection and improvement, reimbursement and accountability.

Cross-sectional imaging findings of splenic infections: is differential diagnosis possible?

The spleen plays an important role in the immunological homeostasis of the body. Several neoplastic and non-neoplastic diseases may affect this organ, and imaging is of fundamental importance for diagnosis. Infectious diseases of the spleen can be encountered in daily radiology practice, and differential diagnosis may sometimes be challenging. Infectious involvement of the spleen can be primary or secondary to a different source outside the spleen. Despite the fact that different infectious diseases may cause similar imaging findings, we believe that differential diagnosis between different causes may also be possible in certain patients with imaging. Early diagnosis may potentially enhance patients' treatment and outcome. In this review, we aimed to increase imaging specialists' awareness of splenic infections by describing the multimodality imaging features of common and atypical infections of the spleen with their differential diagnoses.

Imaging findings of diseases affecting the gastrohepatic ligament: not as acquiscent as it seems.

The gastrohepatic ligament, a major part of the lesser omentum with hepatoduodenal ligament, is an important anatomic structure connected to several other intraabdominal organs and ligaments. It is at the crossroads of several different anatomic structures and may be affected by different diseases. In this article, we aim to increase the awareness of imagers to this small anatomic structure and provide clues for correct diagnosis and assessment of diseases that may affect this area. We will examine various diseases involving the gastrohepatic ligament in detail and try to address its importance using representative cases.

Cross-sectional imaging features of unusual adrenal lesions: a radiopathological correlation.

The detection rates of adrenal masses are increasing with the common and widespread use of cross-sectional imaging. Adrenal adenomas, myelolipomas, metastases, pheochromocytomas, and adrenocortical tumors are well-known and relatively common adrenal tumors. However, there are many less-known neoplastic and nonneoplastic adrenal diseases that might affect the adrenal glands in addition to these common lesions. These rare entities include, but are not limited to, hydatid cysts, congenital adrenal hyperplasia, Wolman disease, adrenal tuberculosis, primary adrenal lymphoma. This article aims to present imaging findings of these unusual lesions in accordance with their pathologic characteristics. We think that the simultaneous presentation of the pathological findings with the imaging features may facilitate the learning process and may potentially enhance the recognition of these entities.

Immunoglobulin G4-related systemic disease: mesenteric and peritoneal involvement with radiopathological correlation and differential diagnoses.

Since its first introduction in 2003 by Kamisawa et al., IgG4-related disease has gained wide interest in the imaging community, and several manuscripts have been published regarding its imaging features. In addition to initial observations in the pancreaticobiliary system, it is now well known that the disease may involve every organ system in the body. There is not much information in the imaging literature about the involvement of mesentery, omentum, and peritoneum in this disease. This article aims to provide more information about the imaging findings of IgG4-related disease regarding these areas by making radiopathological correlations and discussing the possible differential diagnoses.

A sling of a normal right inferior vena cava around the abdominal aorta.

Variations in the anatomy of inferior vena cava (IVC) may have important clinical implications. In-depth knowledge of its embryology and variations are of fundamental importance to prevent any potential medical complications related to anatomic variations of the IVC. In this article, we described a previously unreported, to the best of our knowledge, a variation of IVC. In the case we presented, the IVC was seen almost completely encircling the abdominal aorta. We decided to call this anatomic variation as "a sling of a normal right IVC around the abdominal aorta". Cross-sectional imaging is a prompt and highly reliable method to evaluate IVC anatomy and may have significant clinical importance to prevent any potential complications related to IVC during surgery or interventional radiology procedures.

Can we differentiate neoplastic and non-neoplastic spontaneous adrenal bleeding? Imaging findings with radiopathologic correlation.

Spontaneous adrenal bleeding is a rare clinical event with non-specific clinical features. Life-threatening bleeding in the adrenal glands may be promptly diagnosed with imaging. Computed tomography (CT) is generally the first imaging modality to be used in these patients. However, in the acute phase of bleeding, it may be difficult to detect the underlying mass from the large hematoma. In these patients, additional imaging studies such as magnetic resonance imaging or positron emission tomography/CT may be utilized to rule out a neoplastic mass as the source of bleeding. In patients where an underlying neoplastic mass could not be identified at the time of initial diagnosis, follow-up imaging may be helpful after the acute phase subsides.

Role of imaging in visceral vascular emergencies.

Differential diagnosis in non-traumatic acute abdomen is broad and unrelated diseases may simulate each other from a clinical perspective. Despite the fact that they are not as common, acute abdominal pain due to diseases related to visceral vessels may be life-threating if not detected and treated promptly. Thrombosis, dissection, and aneurysm in the abdominal visceral arteries or thrombosis in visceral veins may cause acute abdominal pain. Imaging with appropriate protocoling plays a fundamental role in both early diagnosis and determination of the treatment approach in these cases where early treatment can be life-saving. Computed tomography (CT) appears to be the most effective modality for the diagnosis as it provides high detail images in a very short time. Patient cooperation is also a less concern as compared to magnetic resonance imaging (MRI). As the imaging findings may sometimes be really subtle, diagnosis may be difficult especially to inexperienced imagers. Correct protocoling is also very critical to detect arterial abnormalities as visceral arterial abnormalities may not be detectable in portal phase only abdominal CT scans. In this article, we aimed to increase awareness among imaging specialists to these not very common causes of acute abdomen.

Tumor in the veins: an abdominal perspective with an emphasis on CT and MR imaging.

Endovenous tumor thrombus in abdomen should be accurately diagnosed as it is a significant finding that may change medical and surgical treatment approaches. As some underlying reasons for endovenous tumor thrombi are relatively rare and imaging findings may be quite subtle, they can be easily overlooked which may have important clinical consequences. In this paper, we described the various imaging aspects of endovenous tumor thrombi originating from various tumor types in different venous structures of the abdomen.

Non-neoplastic hepatopancreatobiliary lesions simulating malignancy: can we differentiate?

Despite the success of cross-sectional imaging in evaluating hepatopancreatobiliary system malignancies, several non-malignant disease processes may closely mimic malignancy. Differentiating these benign diseases from malignancy may be difficult, or even impossible, even in the hands of experienced imagers. In this manuscript, we present benign mimics involving the hepatopancreatobiliary system and try to increase awareness of these potential pitfalls.

Imaging findings of infectious and inflammatory diseases of the urinary system mimicking neoplastic diseases.

Neoplastic diseases affecting the urinary organs are common, and diagnosis by imaging is usually straightforward. However, infectious/inflammatory processes also commonly affect these organs and can be mistaken for a neoplasm. Familiarity with these potential mimickers and awareness of their imaging presentations are key for correct diagnosis. We present the imaging findings of non-neoplastic infectious/inflammatory diseases that can mimic a neoplastic process.