Imaging Reporting Research Articles

Evaluation and Use of Natural Language Processing (NLP) Reasoning and Classification Models to Support Clinical Trial Patient Identification and Enrollment in the Community Oncology Setting

Clinical trial research in oncology relies heavily on clinical documentation within the electronic medical record (EMR) to ascertain patient eligibility in clinical trials based on inclusion and exclusion criteria. The structured data elements within the EMR serve as the primary information source for defining patient cohorts, with clinical cancer stage and performance status being two pivotal criteria determining trial eligibility. The challenge arises from the inconsistent availability of clinical stage and performance status data within the structured fields of the EMR despite their consistent presence in clinical notes. Additionally, there is a deficiency of standardization of this data that exists in the unstructured field. Hence, due to lack of structured data and standardization of said data, there are limitations in developing artificial intelligence (AI) models. To increase the comprehensiveness of clinical records, a clinical research team at a community oncology practice was consulted to identify requirements and extract essential clinical features from de-identified data. The methods outlined in this paper focused on eliminating false positives to allow future development of Large Language Models (LLM) using the outputted structured fields which resulted in an increase in patient record completeness with high accuracy. The accuracy ranged from 97.5-97.75% for the models that were developed. Out of the 60,000+ patients, the numerical staging, TNM (tumor, node, metastasis) staging, and Karnofsky performance score models added a structured field for 29.62%, 21.01%, and 40.64% patients respectively. Additionally, a semi-supervised NLP algorithm was applied on the performance status algorithm which achieved a mean absolute error (MAE) of 1.57. This work demonstrates the use case of natural language processing (NLP) in optimizing the clinical research enrollment process by providing an efficient and accurate method to detect key clinical values in unstructured patient data. Similar methodology with more advanced algorithms such as LLM can be employed to detect additional patient elements such as molecular biomarkers, imaging reports, postoperative surgical outcomes (i.e., clear margins etc.) and patient treatment outcomes using the extracted structured fields.

A New Perspective in Diagnosis of Neuroendocrine Tumors and Pseudo Pheochromocytomas, Understanding the Biology Behind a Mislabeled Medical Terminology

Recently the incidence of undiagnosed hypertension that does not fit into either category of essential hypertension or secondary hypertension is becoming largely increased. Most of these cases are attributed to anxiety or forms of panic attacks associated with high blood pressure. Some in fact are lethal yet clinicians have been challenged to know the real cause of the disease. The term pseudopheochromocytoma widely used for patients who have typical presentation of pheochromocytoma but yet no concrete evidence of disease manifested in their imaging or lab reports to suggest increased production of adrenaline or neurepinephrine, as seen in typical pheochromocytomas. These patients suffer from life threatening attacks of hypertension, yet the only medical revenue offered are anxiolytics, or tricyclic antidepressants. Here we review a case of suspected pheochromocytoma and we present a novel theory about the real player behind all cases of pseudopheochromocytomas. Since in contrast to all other tumors, performing a tissue biopsy is detrimental in the case of pheochromocytomas, and therefore no one logically considers a tissue biopsy for adrenal gland, liquid biopsy( NETest by WREN Laboratory) is a great substitute with high accuracy to replace the tissue biopsy. We conclude that performing liquid biopsy on all cases of pseudopheochromocytoma is essentially life saving and can lead the clinician to appropriate course of action, when the other labs and functional imaging are inconclusive or negative. This is the very perfect example of why the traditional thinking of required presence of metanephrines in urine or blood, and nuclear studies(Dotetate scan/ Octrascan/ MiBG scan, etc…) for diagnosis of pheochromocytoma and surgical interventions depending on such tests is outdated.

The Horizon of Thyroid Imaging Reporting and Data System in the Diagnostic Performance of Thyroid Nodules: Clinical Application and Future Perspectives

The Horizon of Thyroid Imaging Reporting and Data System in the Diagnostic Performance of Thyroid Nodules: Clinical Application and Future Perspectives

Rare giant ovarian thecoma presented as atypical/incomplete Meigs' syndrome: A case image report

Rare giant ovarian thecoma presented as atypical/incomplete Meigs' syndrome: A case image report

Performance of breast MRI for high-risk screening during lactation.

To assess the diagnostic performance of breast MRI during lactation in the setting of high-risk breast cancer screening. Screening breast MRIs performed between April 2008 and March 2024 were retrospectively reviewed. Background parenchymal enhancement (BPE) grade was compared between lactating patients and patients who recently stopped lactating using the Mann-Whitney test. Breast Imaging Reporting and Data System (BI-RADS) scores prevalence rates were compared between lactating patients and controls encompassing young non-lactating patients using the Chi-square test. Diagnostic performance was calculated for patients with a biopsy reference or a 1-year radiologic follow-up. One-hundred forty-two screening breast MRIs were performed in lactating patients (n = 104, median age, 36.0 ± 6.0 years). Marked BPE appeared in 82% of cases (116/142), with a higher BPE grade in exams performed during lactation as compared with those performed in patients who had recently ceased lactating (p < 0.001). Screening MRIs performed during lactation had a higher rate of BI-RADS 3 scores (40/142, 28% vs. 683/8922, 7%, p < 0.001) and a lower rate of BI-RADS 1/2 scores (88/142, 62% vs. 7549/8922, 84.6%, p = 0.002) compared with those performed in controls (n = 8922). One pregnancy-associated breast cancer was detected, and one interval-cancer occurred. All MRI-guided biopsies were negative (n = 13). Screening breast MRI during lactation had 50% sensitivity (1/2), 60% specificity (72/120), 2.0% positive predictive value (1/49), and 98.6% negative predictive value (71/82). The efficacy of breast MRI for high-risk screening during lactation is limited by prominent BPE, leading to an increased rate of BI-RADS 3 categorization and diminished overall specificity. Question Studies on breast MRI during lactation were solely focused on studies conducted in patients with known cancer but not in the screening setting. Findings Screening breast MRI during lactation usually results in marked background parenchymal enhancement, negatively impacting its diagnostic performance. Clinical relevance Despite the lower performance, and amidst the significant risk of pregnancy-associated breast cancer, this screening approach remains relevant for lactating patients with high-risk profiles, such as BReast Cancer (BRCA) carriers. Radiologists should be familiar with the normal appearance of breast MRI duringlactation.

Diagnostic significance of peritumoral enhancement in distinguishing between muscle-invasive and non-muscle-invasive bladder cancer.

The aim of this study was to assess the prevalence of peritumoral enhancement (PTE) in patients with muscle-invasive bladder cancer (MIBC) and non-MIBC (NMIBC) and to propose a modified diagnostic criterion for Vesical Imaging Reporting and Data System (VI-RADS) that incorporates PTE. This retrospective study included 95 patients with bladder cancer (age, 72 ± 11; 77 men; 36 MIBCs and 59 Non-MIBCs) who underwent multiparametric MRI in our referral center between 2011 and 2023. The images were interpreted by four radiologists. The readers classified the possibility of muscle layer invasion into categories 1-5, based on the VI-RADS categorical diagnostic criterion. PTE was defined as a linear contrast enhancement observed at the edge of tumor invasion which is convex outward from the normal bladder wall and contrasts more than the normal muscle layer and tumor. A modified VI-RADS that upgrades the final VI-RADS category to 4 if PTE is present when the original VI-RADS category is 3 or less was proposed. The frequency of PTE in the MIBC and NMIBC groups was compared using the Fisher's exact test. Sensitivity and specificity for the diagnosis of MIBC were compared with the original VI-RADS using McNemar test. Pathologic diagnosis was used as the reference standard. PTE was present in 70-81% (25/36-29/36) of MIBC and absent in 92-98% (54/59-58/59) of non-MIBC. For all readers, the PTE was significantly more frequent (p < 0.001) in the MIBC group than the NMIBC group. The sensitivities of modified VI-RADS (75.0-86.1%) were significantly higher than those of original VI-RADS (41.7-55.6%) (p = 0.002-0.008). The specificity of modified VI-RADS (91.5-98.3%) were not statistically different from original VI-RADS (98.3-100%). In conclusion, PTE is a highly specific finding for MIBC. modified VI-RADS incorporating PTE increases sensitivity for MIBC.

Probing clarity: AI-generated simplified breast imaging reports for enhanced patient comprehension powered by ChatGPT-4o

BackgroundTo assess the reliability and comprehensibility of breast radiology reports simplified by artificial intelligence using the large language model (LLM) ChatGPT-4o.MethodsA radiologist with 20 years’ experience selected 21 anonymized breast radiology reports, 7 mammography, 7 breast ultrasound, and 7 breast magnetic resonance imaging (MRI), categorized according to breast imaging reporting and data system (BI-RADS). These reports underwent simplification by prompting ChatGPT-4o with “Explain this medical report to a patient using simple language”. Five breast radiologists assessed the quality of these simplified reports for factual accuracy, completeness, and potential harm with a 5-point Likert scale from 1 (strongly agree) to 5 (strongly disagree). Another breast radiologist evaluated the text comprehension of five non-healthcare personnel readers using a 5-point Likert scale from 1 (excellent) to 5 (poor). Descriptive statistics, Cronbach’s α, and the Kruskal–Wallis test were used.ResultsMammography, ultrasound, and MRI showed high factual accuracy (median 2) and completeness (median 2) across radiologists, with low potential harm scores (median 5); no significant group differences (p ≥ 0.780), and high internal consistency (α > 0.80) were observed. Non-healthcare readers showed high comprehension (median 2 for mammography and MRI and 1 for ultrasound); no significant group differences across modalities (p = 0.368), and high internal consistency (α > 0.85) were observed. BI-RADS 0, 1, and 2 reports were accurately explained, while BI-RADS 3–6 reports were challenging.ConclusionThe model demonstrated reliability and clarity, offering promise for patients with diverse backgrounds. LLMs like ChatGPT-4o could simplify breast radiology reports, aid in communication, and enhance patient care.Relevance statementSimplified breast radiology reports generated by ChatGPT-4o show potential in enhancing communication with patients, improving comprehension across varying educational backgrounds, and contributing to patient-centered care in radiology practice.Key PointsAI simplifies complex breast imaging reports, enhancing patient understanding.Simplified reports from AI maintain accuracy, improving patient comprehension significantly.Implementing AI reports enhances patient engagement and communication in breast imaging.Graphical

Inter-Reader Agreement for Contrast-Enhanced Ultrasound Liver Imaging Reporting and Data System Major Features and Final Categorization: A Subanalysis From a Prospective Multicenter Study.

Contrast-enhanced ultrasound (CEUS) Liver Imaging Reporting and Data System (LI-RADS) is used to definitively diagnose hepatocellular carcinoma (HCC) in patients at risk. However, the user variability associated with CEUS LI-RADS has not been validated in North American and European patients. This study aims to evaluate the inter-reader agreements of CEUS LI-RADS features for, and final categorization of, HCC in patients at risk. This retrospective multicenter clinical study used the database of a previous prospective multinational study, evaluating the accuracy of CEUS LI-RADS for HCC diagnosis in patients at risk. All cases were first evaluated by a site physician performing/supervising the CEUS examination. Randomly selected cases were re-evaluated by a blinded central reader. Final diagnosis was confirmed with the reference standard, which was a composite of imaging tests and histology. Cohen's kappa test was used to evaluate inter-reader agreement. This study included 150 liver nodules and 58.0% (87/150) were confirmed as HCC, 4.7% (7/150) non-HCC malignancies, 22.7% (34/150) had no confirmed final diagnosis, and 14.7% (22/150) were nonmalignant. Inter-reader agreements were substantial for CEUS LI-RADS categorization (κ = 0.61; 95% confidence interval [CI]: 0.51-0.71) and major features assessment (ranged κ = 0.64-0.78), LR-5 (κ = 0.65; 95% CI: 0.52-0.77), and LR-M (κ = 0.67; 95% CI: 0.44-0.90), while for LR-1 and LR-2 categorization was almost perfect (κ = 0.85; 95% CI: 0.65-1.00). Our study reported a substantial inter-reader agreement for overall CEUS LI-RADS categorization, especially for LR-5 and LR-M, and major imaging features of HCC, further confirming CEUS LI-RADS as a valuable and reliable tool for diagnosis of HCC.

Role of Image-Guided Biopsy in Nonpalpable Breast Lesions: A Study in the Sub-Himalayan Region of North India

Background Nonpalpable breast lesions pose a challenge in their early diagnosis. Image-guided biopsy is preferred in these cases so that a pathological diagnosis of breast carcinoma is reached timely for a better prognosis as the disease has an increased chance of successful outcome with early identification and treatment. Objective The study aims at evaluating the role of stereotactic core needle biopsy (CNB) and percutaneous ultrasound-guided core needle biopsy (US-CNB) in diagnosing suspicious nonpalpable breast lesions. Methods Our study included 35 patients with nonpalpable breast lesions and having a Breast Imaging Reporting and Data System (BI-RADS) risk assessment category IV or V on mammography or sonography. These 35 lesions were subjected to percutaneous image-guided (stereotactic or US) biopsy for histopathological analysis. Results Out of a total of 35 cases, 17 were pathologically malignant (48.6%), with the most common subtype being invasive ductal carcinoma (82.3%). Twenty-nine cases underwent US-CNB, 16 (55.1%) of which were malignant and 13 (44.8%) were benign on histopathological evaluation (HPE). The remaining six cases, which on mammography showed no mass but suspicious malignant calcification only, were subjected to stereotactic CNB, out of which one (16.6%) was malignant and five (83.3%) were benign on HPE. Hence, the lesions visible on sonography were more likely to be malignant. Conclusion Sonography and mammography play a complimentary role in detecting breast carcinoma. Percutaneous biopsy under image guidance can be used as an accurate diagnostic alternative to open surgical excisional biopsy to avoid diagnostic delay.

Automated breast density assessment for full-field digital mammography and digital breast tomosynthesis.

Mammographic density is a strong risk factor for breast cancer (BC) and is reported clinically as part of Breast Imaging Reporting and Data System (BI-RADS) results issued by radiologists. Automated assessment of density is needed that can be used for both full-field digital mammography (FFDM) and digital breast tomosynthesis (DBT) as both types of exams are acquired in standard clinical practice. We trained a deep learning model to automate the estimation of BI-RADS density from a prospective Washington University (WashU) clinic-based cohort of 9,714 women, entering into the cohort in 2013 with follow-up through, October 31, 2020. The cohort included 27% non-Hispanic Black women. The trained algorithm was assessed in an external validation cohort that included 18,360 women screened at Emory from January 1, 2013 and followed through December 31, 2020 that included 42% non-Hispanic Black women. Our model-estimated BI-RADS density demonstrated substantial agreement with the density as assessed by radiologist. In the external validation, the agreement with radiologists for category B 81% and C 77% for FFDM and B 83% and C 74% for DBT show important distinction for separation of women with dense breast. We obtained a Cohen's κ of 0.72 (95% CI, 0.71, 0.73) in FFDM and 0.71 (95% CI 0.69, 0.73) in DBT. We provided a consistent and fully automated BI-RADS estimation for both FFDM and DBT using a deep learning model. The software can be easily implemented anywhere for clinical use and risk prediction.

PERFIL DOS RESULTADOS DE ESCORE DE CÁLCIO DE CORONÁRIAS EM UM SERVIÇO PRIVADO DE ARACAJU

INTRODUCTION: Cardiovascular diseases (CVD) are responsible for one third of all deaths worldwide, with immutable risk factors such as age and gender, but also mutable factors such as sedentary lifestyles, obesity, hypertension and alcohol consumption. Given this scenario, there is a need to stratify the risk of individuals. The Coronary Calcium Score (CCS) is a tool used in non-contrast CT scans to stratify cardiovascular risk by quantifying the amount of calcium in the coronary arteries. OBJECTIVE: To analyze the coronary calcium score profile of patients at a private service in Aracaju/SE. METHODS: This is a quantitative, cross-sectional, documentary, descriptive and retrospective study, produced by evaluating tomography reports for calcium score, performed on outpatients, in a private service in Aracaju-SE, between January 1 and December 31, 2023. A total of 227 examination reports were obtained. The inclusion criteria were imaging reports from January to December 2023. Duplicate exams, exams of other areas of the body and exams outside the period listed for the research were excluded. RESULTS: Our sample consisted of 125 women and 102 men, with no difference in mean age between them, whose mean calcification was 215.46. Men had higher overall mean values for calcium score and calcification volume than women, as well as for the arterial territories (DA, CX and CD), with the exception of the TCE, for which there was no difference between the genders. CONCLUSION: These data reveal a higher proportion of women undergoing examinations compared to men. Although most of the individuals in the study did not have significant coronary calcification, it was observed that, among the cases of significant calcification, men were more affected than women.

TECRR: a benchmark dataset of radiological reports for BI-RADS classification with machine learning, deep learning, and large language model baselines

BackgroundRecently, machine learning (ML), deep learning (DL), and natural language processing (NLP) have provided promising results in the free-form radiological reports’ classification in the respective medical domain. In order to classify radiological reports properly, a high-quality annotated and curated dataset is required. Currently, no publicly available breast imaging-based radiological dataset exists for the classification of Breast Imaging Reporting and Data System (BI-RADS) categories and breast density scores, as characterized by the American College of Radiology (ACR). To tackle this problem, we construct and annotate a breast imaging-based radiological reports dataset and its benchmark results. The dataset was originally in Spanish. Board-certified radiologists collected and annotated it according to the BI-RADS lexicon and categories at the Breast Radiology department, TecSalud Hospitals Monterrey, Mexico. Initially, it was translated into English language using Google Translate. Afterwards, it was preprocessed by removing duplicates and missing values. After preprocessing, the final dataset consists of 5046 unique reports from 5046 patients with an average age of 53 years and 100% women. Furthermore, we used word-level NLP-based embedding techniques, term frequency-inverse document frequency (TF-IDF) and word2vec to extract semantic and syntactic information. We also compared the performance of ML, DL and large language models (LLMs) classifiers for BI-RADS category classification.ResultsThe final breast imaging-based radiological reports dataset contains 5046 unique reports. We compared K-Nearest Neighbour (KNN), Support Vector Machine (SVM), Naive Bayes (NB), Random Forest (RF), Adaptive Boosting (AdaBoost), Gradient-Boosting (GB), Extreme Gradient Boosting (XGB), Long Short-Term Memory (LSTM), Bidirectional Encoder Representations from Transformers (BERT) and Biomedical Generative Pre-trained Transformer (BioGPT) classifiers. It is observed that the BioGPT classifier with preprocessed data performed 6% better with a mean sensitivity of 0.60 (95% confidence interval (CI), 0.391-0.812) compared to the second best performing classifier BERT, which achieved mean sensitivity of 0.54 (95% CI, 0.477-0.607).ConclusionIn this work, we propose a curated and annotated benchmark dataset that can be used for BI-RADS and breast density category classification. We also provide baseline results of most ML, DL and LLMs models for BI-RADS classification that can be used as a starting point for future investigation. The main objective of this investigation is to provide a repository for the investigators who wish to enter the field to push the boundaries further.

Lessons learned: strategies for implementing and the ongoing use of LI-RADS in your practice.

The establishment of the Liver Imaging Reporting and Data System (LI-RADS) in 2011 provided a comprehensive approach to standardized imaging, interpretation, and reporting of liver observations in patients diagnosed with or at risk for hepatocellular carcinoma (HCC). Each set of algorithms provides criteria pertinent to the various components of HCC management including surveillance, diagnosis, staging, and treatment response supported by a detailed lexicon of terms applicable to a wide range of liver imaging scenarios. Before its widespread adoption, the variability in the terminology of diagnostic criteria and definitions of imaging features led to significant challenges in patient management and made it difficult to replicate findings or apply them consistently. The integration of LI-RADS into the clinical setting has enhanced the efficiency and clarity of communication between radiologists, referring providers, and patients by employing a uniform language that averts miscommunications. LI-RADS has been strengthened with its integration into the American Association for Study of Liver Diseases practice guidelines. We will provide the background on the initial development of LI-RADS and reasons for development to serve as a starting point for conveying the system's benefits and evolution over the years. We will also suggest strategies for the implementation and maintenance of a LI-RADS program will be discussed.

Comparison in prostate cancer diagnosis with PSA 4–10 ng/mL: radiomics-based model VS. PI-RADS v2.1

BackgroundTo evaluate accuracy of MRI-based radiomics in diagnosing prostate cancer (PCa) in patients with PSA levels between 4 and 10 ng/mL and compare it with the latest Prostate Imaging Reporting and Data System (PI-RADS v2.1) score.Methods221 patients with prostate lesions and PSA levels in 4–10 ng/mL, including 154 and 67 cases in the training and validation groups. Pathological confirmation of all patients was accomplished by the use of MRI-TRUS fusion targeted biopsy or systematic transrectal ultrasound (TRUS) guided biopsy. 851 radiomic features were extracted from each lesion of ADC and T2WI images. The least absolute shrinkage and selection operator (LASSO) regression algorithm and logistic regression were employed to select features and build the ADC and T2WI model. The combined model was obtained based on the ADC and T2WI features. The clinical benefit and diagnostic accuracy of the three radiomics models and PI-RADS v2.1 score were evaluated.Results10 radiomic features were ultimately selected from the ADC images, 13 from the T2WI images and 7 from the combined models. The ADC, T2WI and combined models achieved satisfactory diagnostic accuracy in the training [AUC:0.945 (ADC), 0.939 (T2WI), 0.979 (combined)] and validation groups [AUC: 0.942 (ADC), 0.943 (T2WI), 0.959 (combined)], which was significantly higher than those in PI-RADS v2.1 model (0.825 for training cohort and 0.853 for validation cohort). Compared with the PI-RADS v2.1 score, the three radiomics models generated superior PCa diagnostic performance in both the training (p = 0.002, p = 0.005, p < 0.001) and validation groups (p = 0.045, p = 0.035, p = 0.015).ConclusionRadiomics based on ADC and T2WI images can better identify PCa in patients with PSA 4–10 ng/mL, and MRI-based radiomics significantly outperforms the PI-RADS v2.1 score.Clinical trial numberNot applicable.

Efficient labeling of french mammogram reports with MammoBERT

Recent advances in deep learning and natural language processing (NLP) have broadened opportunities for automatic text processing in the medical field. However, the development of models for low-resource languages like French is challenged by limited datasets, often due to legal restrictions. Large-scale training of medical imaging models often requires extracting labels from radiology text reports. Current methods for report labeling primarily rely on sophisticated feature engineering based on medical domain knowledge or manual annotations by radiologists. These methods can be labor-intensive. In this work, we introduce a BERT-based approach for the efficient labeling of French mammogram image reports. Our method leverages both the expansive scale of existing rule-based systems and the precision of radiologist annotations. Our experimental results showcase the superiority of the proposed approach. It was initially fine-tuned on a limited dataset of radiologist annotations. Then, it underwent training on annotations generated by a rule-based labeler. Our findings reveal that our final model, MammoBERT, significantly outperforms the rule-based labeler while simultaneously reducing the necessity for radiologist annotations during training. This research not only advances the state of the art in medical image report labeling but also offers an efficient and effective solution for large-scale medical imaging model development.

CT/MRI technical pitfalls for diagnosis and treatment response assessment using LI-RADS and how to optimize.

Hepatocellular carcinoma (HCC), the most common primary liver cancer, is a significant global health burden. Accurate imaging is crucial for diagnosis and treatment response assessment, often eliminating the need for biopsy. The Liver Imaging Reporting and Data System (LI-RADS) standardizes the interpretation and reporting of liver imaging for diagnosis and treatment response assessment, categorizing observations using defined categories that are based on the probability of malignancy or post-treatment tumor viability. Optimized imaging protocols are essential for accurate visualization and characterization of liver findings by LI-RADS. Common technical pitfalls, such as suboptimal postcontrast phase timing, and MRI-specific challenges like subtraction misregistration artifacts, can significantly reduce image quality and diagnostic accuracy. The use of hepatobiliary contrast agents introduces additional challenges including arterial phase degradation and suboptimal uptake in advanced cirrhosis. This review provides radiologists with comprehensive insights into the technical aspects of liver imaging for LI-RADS. We discuss common pitfalls encountered in routine clinical practice and offer practical solutions to optimize imaging techniques. We also highlight technical advances in liver imaging, including multi-arterial MR acquisition and compressed sensing. By understanding and addressing these technical aspects, radiologists can improve accuracy and confidence in the diagnosis and treatment response assessment for hepatocellular carcinoma.

Ring-Enhancement on CEUS: Is it Useful in the Differential Diagnosis of Solid Thyroid Nodules?

To investigate the efficiency of contrast-enhanced ultrasound (CEUS) features, particularly ring-enhancement patterns, in the differential diagnosis of thyroid nodules. 302 nodules with CEUS ring-enhancement were retrospectively enrolled, including 135 benign and 167 malignant ones. The ring-enhancement patterns were classified into regular and irregular hyper- or hypo-ring enhancement. Comparative analyses of ultrasound (US) and CEUS features between benign and malignant nodules were performed. The diagnostic performances of the ring-enhancement patterns and Chinese Thyroid Imaging Reporting and Data System (C-TIRADS) were compared in nodules with different sizes. Irregular hypo-ring enhancement was much more common in malignancies than that in benign ones, and it was an independent predictor for thyroid malignant nodules. With irregular hypo-ring enhancement as the diagnostic criteria for malignant nodules, the specificity was higher than that of C-TIRADS (85.2% vs. 75.6%, p = .037) while the AUC was comparable (0.845 vs. 0.803, p = .136) in all nodules. When the nodule size was taken into account, the specificity and AUC were both significantly higher than those of C-TIRADS (92.8% vs. 81.1%, p = .021; 0.907 vs. 0.823, p = .026) in nodules ≥10 mm, which can decrease the unnecessary FNA rate. Irregular hypo-ring enhancement was a valuable CEUS feature for the differential diagnosis of thyroid nodules, especially in nodules ≥10 mm.

SWOT analysis of the application of artificial intelligence in radiologic technology and radiology

Artificial intelligence (AI) is bringing changes to radiology and radiologic technology, enabling the development of programs and algorithms that facilitate diagnosis and decisions. It is essential to understand how AI can improve patient outcomes, increase the efficiency of investigations and reduce costs. Machine and deep learning have proven to be extremely useful in the detection and characterization of lesions, improving routine imaging techniques and facilitating the work of radiologists by reducing workload and improving the quality of reporting. The practical application of artificial intelligence in radiology and radiologic technology has been slowed down by the lack of integrated solutions and well-structured data archives, as well as challenges such as non-transparency of decision-making systems and large amounts of quality data needed to train artificial intelligence models. There are concerns about the potential impact of AI on the work of radiologists and radiologic technologists, which may hinder the development and implementation of this technology. Textual data derived from image reports can provide valuable healthcare insights. Natural language processing (NLP), a subset of artificial intelligence, offers promising solutions for handling unstructured text in these reports, opening a new era in extracting information from medical images and related reports. Due to the challenges in training experts for the application of AI in healthcare, a multidisciplinary approach will have to be used and investments in collaboration and education will have to be made. There are outstanding issues of liability and regulation regarding data storage and privacy, particularly in the case of cloud storage. The concern of the workforce and their lack of education about artificial intelligence represents an obstacle to its adoption, but also offers an opportunity for the technological advancement of the profession.

Risk of Bias in Liver Imaging Reporting and Data System Studies Using QUADAS-2.

Purpose: Use a tailored version of the Quality Assessment of Diagnostic Accuracy Studies tool to evaluate risk of bias and applicability across LIRADS related publications. Method: A tailored QUADAS-2 tool was created through consensus approach to assess risk of bias and applicability across 37 LI-RADS related publications. Studies were selected from 2017 to 2022 using the assistance of experienced hospital librarians to search for studies evaluating the diagnostic accuracy of CT, MRI, or contrast-enhanced ultrasound for HCC using LI-RADS through multiple different databases. QUADAS-2 assessments were performed in duplicate and independently by 2 authors with experience using the QUADAS-2 tool. Disagreements were resolved with a third expert reviewer. Consensus QUADAS-2 assessments were tabulated for each domain. Results: Using the tailored QUADAS-2 tool, 31 of the 37 included LI-RADS studies were assessed as high risk of bias, and 9 out of 37 studies demonstrated concerns for applicability. Patient selection (21 out of 37 studies) and flow/timing (24 out of 37 studies) domains demonstrated the highest risk of bias. 6 out of 37 studies in the index domain demonstrated high risk of bias. 2 out of 37 studies showed high risk of bias in the reference standard domain. Conclusion: A significant proportion of LI-RADS research is at risk of bias with concerns for applicability. Identifying risk of bias in such research is essential to recognize limitations of a study that may affect the validity of the results. Areas for improvement in LI-RADS research include reducing selection bias, avoiding inappropriate exclusions, and decreasing verification bias.

Assessment of the Diagnostic Performance of a Commercially Available Artificial Intelligence Algorithm for Risk Stratification of Thyroid Nodules on Ultrasound.

Background: Thyroid nodules are challenging to accurately characterize on ultrasound (US), though the emergence of risk stratification systems and more recently artificial intelligence (AI) algorithms has improved nodule classification. The purpose of this study was to evaluate the performance of a recent Food and Drug Administration (FDA)-cleared AI tool for detection of malignancy in thyroid nodules on US. Methods: One year of consecutive thyroid US with ≥1 nodule from Duke University Hospital and its affiliate community hospital (649 nodules from 347 patients) were retrospectively evaluated. Included nodules had ground truth diagnoses by surgical pathology, fine needle aspiration (FNA), or three-year follow-up US showing stability. An FDA-cleared AI tool (Koios DS Thyroid) analyzed each nodule to generate (i) American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) descriptors, scores, and follow-up recommendations and (ii) an AI-adapter score to further adjust risk assessments and recommendations. Four groups were then compared: (i) Koios with AI-adapter, (ii) Koios without AI-adapter, (iii) clinical radiology report, and (iv) radiology report combined with AI-adapter. Performance of the final recommendations (FNA or no FNA) was determined based on ground truth, and comparison between the four groups was made using sensitivity, specificity, and receiver-operating-curve analysis. Results: Of 649 nodules, 32 were malignant and 617 were benign. Performance of Koios with AI-adapter enabled was similar to radiologists (area under the curve [AUC] 0.70 for both, [CI 0.60-0.81] and [0.60-0.79], respectively). Koios with AI-adapter had improved specificity compared to radiologists (0.63 [CI: 0.59-0.67] versus 0.43 [CI: 0.38-0.48]) but decreased sensitivity (0.69 [CI: 0.50-0.83) versus 0.81 [CI: 0.61, 0.92]). Highest performance was seen when the radiology interpretation was combined with the AI-adapter (AUC 0.76, [CI: 0.67-0.85]). Combined with the AI-adapter, radiologist specificity improved from 0.43 ([CI: 0.38-0.48]) to 0.53 ([CI: 0.49-0.58]) (McNemar's test p < 0.001), resulting in 17% fewer FNA recommendations, with unchanged sensitivity (0.81, p = 1). Conclusion: Koios DS demonstrated standalone performance similar to radiologists, though with lower sensitivity and higher specificity. Performance was best when radiologist interpretations were combined with the AI-adapter component, with improved specificity and reduced unnecessary FNA recommendations.