Merlin: A Vision Language Foundation Model for 3D Computed Tomography.

Over 85 million computed tomography (CT) scans are performed annually in the US, of which approximately one quarter focus on the abdomen. Given the current shortage of both general and specialized radiologists, there is a large impetus to use artificial intelligence to alleviate the burden of interpreting these complex imaging studies while simultaneously using the images to extract novel physiological insights. Prior state-of-the-art approaches for automated medical image interpretation leverage vision language models (VLMs) that utilize both the image and the corresponding textual radiology reports. However, current medical VLMs are generally limited to 2D images and short reports. To overcome these shortcomings for abdominal CT interpretation, we introduce Merlin - a 3D VLM that leverages both structured electronic health records (EHR) and unstructured radiology reports for pretraining without requiring additional manual annotations. We train Merlin using a high-quality clinical dataset of paired CT scans (6+ million images from 15,331 CTs), EHR diagnosis codes (1.8+ million codes), and radiology reports (6+ million tokens) for training. We comprehensively evaluate Merlin on 6 task types and 752 individual tasks. The non-adapted (off-the-shelf) tasks include zero-shot findings classification (31 findings), phenotype classification (692 phenotypes), and zero-shot cross-modal retrieval (image to findings and image to impressions), while model adapted tasks include 5-year chronic disease prediction (6 diseases), radiology report generation, and 3D semantic segmentation (20 organs). We perform internal validation on a test set of 5,137 CTs, and external validation on 7,000 clinical CTs and on two public CT datasets (VerSe, TotalSegmentator). Beyond these clinically-relevant evaluations, we assess the efficacy of various network architectures and training strategies to depict that Merlin has favorable performance to existing task-specific baselines. We derive data scaling laws to empirically assess training data needs for requisite downstream task performance. Furthermore, unlike conventional VLMs that require hundreds of GPUs for training, we perform all training on a single GPU. This computationally efficient design can help democratize foundation model training, especially for health systems with compute constraints. We plan to release our trained models, code, and dataset, pending manual removal of all protected health information.

Automated abdominal CT contrast phase detection using an interpretable and open-source artificial intelligence algorithm.

OBJECTIVES: To develop and validate an open-source artificial intelligence (AI) algorithm to accurately detect contrast phases in abdominal CT scans. MATERIALS AND METHODS: Retrospective study aimed to develop an AI algorithm trained on 739 abdominal CT exams from 2016 to 2021, from 200 unique patients, covering 1545 axial series. We performed segmentation of five key anatomic structures-aorta, portal vein, inferior vena cava, renal parenchyma, and renal pelvis-using TotalSegmentator, a deep learning-based tool for multi-organ segmentation, and a rule-based approach to extract the renal pelvis. Radiomics features were extracted from the anatomical structures for use in a gradient-boosting classifier to identify four contrast phases: non-contrast, arterial, venous, and delayed. Internal and external validation was performed using the F1 score and other classification metrics, on the external dataset "VinDr-Multiphase CT". RESULTS: The training dataset consisted of 172 patients (mean age, 70 years ± 8, 22% women), and the internal test set included 28 patients (mean age, 68 years ± 8, 14% women). In internal validation, the classifier achieved an accuracy of 92.3%, with an average F1 score of 90.7%. During external validation, the algorithm maintained an accuracy of 90.1%, with an average F1 score of 82.6%. Shapley feature attribution analysis indicated that renal and vascular radiodensity values were the most important for phase classification. CONCLUSION: An open-source and interpretable AI algorithm accurately detects contrast phases in abdominal CT scans, with high accuracy and F1 scores in internal and external validation, confirming its generalization capability. CLINICAL RELEVANCE STATEMENT: Contrast phase detection in abdominal CT scans is a critical step for downstream AI applications, deploying algorithms in the clinical setting, and for quantifying imaging biomarkers, ultimately allowing for better diagnostics and increased access to diagnostic imaging. KEY POINTS: Digital Imaging and Communications in Medicine labels are inaccurate for determining the abdominal CT scan phase. AI provides great help in accurately discriminating the contrast phase. Accurate contrast phase determination aids downstream AI applications and biomarker quantification.

Cardiovascular mortality in peritoneal dialysis: the impact of mineral disorders / Mortalidade cardiovascular em diálise peritoneal: o impacto dos distúrbios minerais

Abstract Introduction: Mineral and bone disorders (MBD) are associated with higher mortality in dialysis patients. The main guidelines related to the subject, Kidney Disease Outcomes Quality Initiative (KDOQI) and Kidney Disease: Improving Global Outcomes (KDIGO), were elaborated based on published information from hemodialysis participants. The aim of our study was to evaluate the impact of calcium (Ca), phosphorus (P), and parathyroid hormone (PTH) (according to guideline ranges from KDOQI and KDIGO) on the cardiovascular mortality of peritoneal dialysis (PD) patients. Methods: We used the BRAZPDII database, an observational multi-centric prospective study, which assessed participants on PD between December 2004 and January 2011. Amongst 9,905 participants included in this database, we analyzed 4424 participants who were on PD for at least 6 months. The appropriate confounding variables were entered into the model. Serum levels of Ca, P, and PTH were the variables of interest for the purposes of the current study. Results: We found a significant association between high P serum levels, categorized by KDOQI and KDIGO (P above 5.5 mg/dL), and cardiovascular survival (p < 0.01). Likewise, a compelling association was found between lower levels of PTH, categorized by guidelines (KDOQI and KDIGO - PTH less than 150 pg/mL, p < 0.01), and cardiovascular survival. Conclusion: In conclusion, levels of P above and PTH below the values proposed by KDOQI and KDIGO were associated with cardiovascular mortality in PD patients.

Resumo Introdução: Os distúrbios minerais e ósseos (DMO) estão associados a maior mortalidade em pacientes de diálise. As principais diretrizes relacionadas ao assunto, Kidney Disease Outcomes Quality Initiative (KDOQI) e Kidney Disease: Improving Global Outcomes (KDIGO) foram elaboradas com base em informações publicadas de pacientes em hemodiálise. O objetivo do nosso estudo foi avaliar o impacto do cálcio (Ca), fósforo (P) e paratormônio (PTH) (de acordo com as faixas propostas pelas diretrizes do KDOQI e KDIGO) na mortalidade cardiovascular de pacientes em diálise peritoneal (DP). Métodos: Utilizamos o banco de dados BRAZPDII, um estudo prospectivo observacional multicêntrico, que avaliou participantes de DP entre dezembro de 2004 e janeiro de 2011. Entre os 9.905 participantes incluídos neste banco de dados, analisamos 4.424 que estavam em DP há pelo menos 6 meses. As variáveis de confusão apropriadas foram inseridas no modelo. Os níveis séricos de Ca, P e PTH foram as variáveis de interesse para os fins do presente estudo. Resultados: Encontramos uma associação significativa entre níveis séricos de P elevados, categorizados por KDOQI e KDIGO (P acima de 5,5 mg/dL), e sobrevivência cardiovascular (p < 0,01). Da mesma forma, foi encontrada uma associação convincente entre níveis mais baixos de PTH, categorizados por diretrizes (KDOQI e KDIGO - PTH inferior a 150 pg/mL, p < 0,01), e sobrevivência cardiovascular. Conclusão: Em conclusão, níveis de P acima e PTH abaixo dos valores propostos por KDOQI e KDIGO foram associados à mortalidade cardiovascular em pacientes de DP.

High prevalence of biochemical disturbances of chronic kidney disease - mineral and bone disorders (CKD-MBD) in a nation-wide peritoneal dialysis cohort: are guideline goals too hard to achieve? / Alta Prevalência de Distúrbios Bioquímicos Minerais e Ósseos na Doença Renal Crônica em uma coorte de diálise peritoneal nacional: as metas das diretrizes são muito difíceis de alcançar?

Abstract Introduction: Chronic kidney disease - mineral and bone disorders (CKD-MBD) are common in dialysis patients. Definition of targets for calcium (Ca), phosphorus (P), parathormone (iPTH), and alkaline phosphatase (ALP) and their treatment recommendations, are provided by international guidelines. There are few studies analyzing CKD-MBD in peritoneal dialysis (PD) patients and the impact of guidelines on mineral metabolism control. The aim of our study was to describe the prevalence of biomarkers for CKD-MBD in a large cohort of PD patients in Brazil. Methods: Data from the nation-wide prospective observational cohort BRAZPD II was used. Incident patients were followed between December 2004 and January 2011. According to KDOQI recommendations, reference ranges for total Ca were 8.4 to 9.5 mg/dL, for P, 3.5 to 5.5 mg/dL, for iPTH, 150-300 pg/mL, and for ALP, 120 U/L. Results: Mean age was 59.8 ± 16 years, 48% were male, and 43% had diabetes. In the beginning, Ca was 8.9 ± 0.9 mg/dL, and 48.3% were on the KODQI target. After 1 year, Ca increased to 9.1 ± 0.9 mg/dL and 50.4% were in the KDOQI preferred range. P at baseline was 5.2 ± 1.6 mg/dL, with 52.8% on target, declining to 4.9 ± 1.5 mg/dL after one year, when 54.7% were on target. Median iPTH at baseline was 238 (P25% 110 - P75% 426 pg/mL) and it remained stable throughout the first year; patients within target ranged from 26 to 28.5%. At the end of the study, 80% was in 3.5 meq/L Ca dialysate concentration, 66.9% of patients was taking any phosphate binder, and 25% was taking activated vitamin D. Conclusions: We observed a significant prevalence of biochemical disorders related to CKD-MBD in this dialysis population.

Resumo Introdução: Os distúrbios minerais e ósseos da doença renal crônica (DMO-DRC) são comuns em pacientes em diálise. A definição de metas para cálcio (Ca), fósforo (P), paratormônio (PTHi) e fosfatase alcalina (FA) e suas recomendações de tratamento são fornecidas por diretrizes internacionais. Há poucos estudos analisando o DMO-DRC em pacientes em diálise peritoneal (DP) e o impacto das diretrizes no controle do metabolismo mineral. O objetivo do nosso estudo foi descrever a prevalência de alterações nos marcadores para DMO-DRC em uma grande coorte de pacientes em DP no Brasil. Métodos: Foram utilizados dados da coorte observacional prospectiva nacional BRAZPD II. Pacientes incidentes foram acompanhados entre Dezembro de 2004 e Janeiro de 2011. De acordo com as recomendações do KDOQI, os intervalos de referência para Ca total foram de 8,4 a 9,5 mg/dL, para P, 3,5 a 5,5 mg/dL, para PTHi, 150-300 pg/mL, e para FA, 120 U/L. Resultados: A idade média foi de 59,8 ± 16 anos, 48% eram homens e 43% tinham diabetes. No início, o Ca era de 8,9 ± 0,9 mg/dL, e 48,3% estavam na meta do KODQI. Após 1 ano, o Ca aumentou para 9,1 ± 0,9 mg/dL e 50,4% estavam na faixa preferida do KDOQI. P basal era 5,2 ± 1,6 mg/dL, com 52,8% na meta, diminuindo para 4,9 ± 1,5 mg/dL após um ano, quando 54,7% estavam na meta. O PTHi basal mediano foi de 238 (P25% 110 - P75% 426 pg/mL) e permaneceu estável durante o primeiro ano; os pacientes dentro da meta variaram de 26 a 28,5%. No final do estudo, 80% estavam na concentração de 3,5 meq/L de Ca dialisato, 66,9% dos pacientes estavam tomando qualquer quelante de fosfato, e 25% estavam tomando vitamina D ativada. Conclusões: Observamos uma prevalência significativa de distúrbios bioquímicos relacionados ao DMO-DRC nesta população em diálise.

High prevalence of biochemical disturbances of chronic kidney disease - mineral and bone disorders (CKD-MBD) in a nation-wide peritoneal dialysis cohort: are guideline goals too hard to achieve?

INTRODUCTION: Chronic kidney disease - mineral and bone disorders (CKD-MBD) are common in dialysis patients. Definition of targets for calcium (Ca), phosphorus (P), parathormone (iPTH), and alkaline phosphatase (ALP) and their treatment recommendations, are provided by international guidelines. There are few studies analyzing CKD-MBD in peritoneal dialysis (PD) patients and the impact of guidelines on mineral metabolism control. The aim of our study was to describe the prevalence of biomarkers for CKD-MBD in a large cohort of PD patients in Brazil. METHODS: Data from the nation-wide prospective observational cohort BRAZPD II was used. Incident patients were followed between December 2004 and January 2011. According to KDOQI recommendations, reference ranges for total Ca were 8.4 to 9.5 mg/dL, for P, 3.5 to 5.5 mg/dL, for iPTH, 150-300 pg/mL, and for ALP, 120 U/L. RESULTS: Mean age was 59.8 ± 16 years, 48% were male, and 43% had diabetes. In the beginning, Ca was 8.9 ± 0.9 mg/dL, and 48.3% were on the KODQI target. After 1 year, Ca increased to 9.1 ± 0.9 mg/dL and 50.4% were in the KDOQI preferred range. P at baseline was 5.2 ± 1.6 mg/dL, with 52.8% on target, declining to 4.9 ± 1.5 mg/dL after one year, when 54.7% were on target. Median iPTH at baseline was 238 (P25% 110 - P75% 426 pg/mL) and it remained stable throughout the first year; patients within target ranged from 26 to 28.5%. At the end of the study, 80% was in 3.5 meq/L Ca dialysate concentration, 66.9% of patients was taking any phosphate binder, and 25% was taking activated vitamin D. CONCLUSIONS: We observed a significant prevalence of biochemical disorders related to CKD-MBD in this dialysis population.

Cardiovascular mortality in peritoneal dialysis: the impact of mineral disorders.

INTRODUCTION: Mineral and bone disorders (MBD) are associated with higher mortality in dialysis patients. The main guidelines related to the subject, Kidney Disease Outcomes Quality Initiative (KDOQI) and Kidney Disease: Improving Global Outcomes (KDIGO), were elaborated based on published information from hemodialysis participants. The aim of our study was to evaluate the impact of calcium (Ca), phosphorus (P), and parathyroid hormone (PTH) (according to guideline ranges from KDOQI and KDIGO) on the cardiovascular mortality of peritoneal dialysis (PD) patients. METHODS: We used the BRAZPDII database, an observational multi-centric prospective study, which assessed participants on PD between December 2004 and January 2011. Amongst 9,905 participants included in this database, we analyzed 4424 participants who were on PD for at least 6 months. The appropriate confounding variables were entered into the model. Serum levels of Ca, P, and PTH were the variables of interest for the purposes of the current study. RESULTS: We found a significant association between high P serum levels, categorized by KDOQI and KDIGO (P above 5.5 mg/dL), and cardiovascular survival (p < 0.01). Likewise, a compelling association was found between lower levels of PTH, categorized by guidelines (KDOQI and KDIGO - PTH less than 150 pg/mL, p < 0.01), and cardiovascular survival. CONCLUSION: In conclusion, levels of P above and PTH below the values proposed by KDOQI and KDIGO were associated with cardiovascular mortality in PD patients.

Effect of parathyroidectomy on bone tissue biomarkers and body composition in patients with chronic kidney disease and secondary hyperparathyroidism.

BACKGROUND/OBJECTIVE: Loss of renal function may induce secondary hyperparathyroidism (s-HPT), which triggers several complications leading to an extreme decline in quality of life and increased mortality in affected patients. We evaluated whether parathyroidectomy (PTx), as surgical treatment for s-HPT, modifies body composition, and hormones involved in the protein-energy metabolism of affected patients. SUBJECTS/METHODS: Overall, 30 s-HPT patients were evaluated at two times, before PTx (pre PTx) and 6 months after PTx (post PTx). Patients were evaluated by biochemistry analysis, anthropometry, electrical bioimpedance (BIA), food intake diary, handgrip strength, and modified global subjective nutritional assessment (SGA). RESULTS: After PTx, patients showed decreased serum levels of total and ionic calcium, as well as decreased alkaline phosphatase and PTH, and increased 25 (OH) vitamin D. These results demonstrate that PTx was efficient to correct part of the mineral disorder. We also observed an increase in caloric intake, body weight, body mass index (BMI), phase angle, handgrip strength, SGA score, and a decreasing in the percentage of weight loss. The osteocalcin concentration of both carboxylated (cOC) and undercarboxylated form was diminished post PTx. The cOC correlated with bone metabolism markers and SGA score. CONCLUSIONS: PTx modified body composition improving nutritional status and preventing the progression of weight loss with increased of energy intake, BMI, handgrip strength, phase angle of BIA, and SGA score. The present study also suggests an association of cOC with bone markers and SGA score. Further studies are needed to better clarify these associations with larger sample size.

The effect of vitamin D and zoledronic acid in bone marrow adiposity in kidney transplant patients: A post hoc analysis.

PURPOSE: Osteoblasts and adipocytes are derived from mesenchymal stem cells. An imbalance in the differentiation of these lineages could affect the preservation of bone integrity. Several studies have suggested the importance of this imbalance in the pathogenesis of osteoporosis after kidney transplant (KT), but the role of bone marrow adiposity in this process is not well known, and if the treatment with the anti-absorptive (zoledronic acid-ZA) drugs could attenuate bone loss. Thus, our objective was compare bone marrow adiposity, osteoblasts and osteocytes before and after KT, verify an association between bone remodeling process (Turnover, Volume, and Mineralization-TMV classification), the osteocyte sclerostin expression to evaluate if there is a role of Wnt pathway, as well as the effect of ZA on these cells. METHODS: We studied 29 new living-donor KT patients. One group received ZA at the time of KT plus cholecalciferol for twelve months, and the other group received only cholecalciferol. Bone biopsies were performed at baseline and after 12 months of treatment. Histomorphometric evaluation was performed in bone and bone marrow adipocytes. Sclerostin (Scl) expression in osteocytes was evaluated by immunohistochemistry. RESULTS: Some bone marrow adiposity parameters were increased before KT. After KT, some of them remained increased and they worsened with the use of ZA. In the baseline, lower bone Volume and Turnover, were associated with increased bone marrow adiposity parameters (some of them). After KT, both groups showed the same associations. Osteocyte Scl expression after KT decreased with the use of ZA. We observed also an inverse association between bone adiposity parameters and lower osteocyte sclerostin expression 12 months after KT. CONCLUSION: In conclusion, the present study suggests that KT fails to normalize bone marrow adiposity, and it even gets worse with the use of ZA. Moreover, bone marrow adiposity is inversely associated with bone Volume and Turnover, which seems to be accentuated by the antiresorptive therapy.