Deep Learning-Based Body Composition Analysis for Cancer Patients Using Computed Tomographic Imaging DOI
İlkay Yıldız Potter,

Maria Virginia Velasquez-Hammerle,

Ara Nazarian

et al.

Deleted Journal, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 11, 2024

Malnutrition is a commonly observed side effect in cancer patients, with 30-85% worldwide prevalence this population. Existing malnutrition screening tools miss ~ 20% of at-risk patients at initial and do not capture the abnormal body composition phenotype. Meanwhile, gold-standard clinical criteria to diagnose use changes as key parameters, particularly fat skeletal muscle mass loss. Diagnostic imaging, such computed tomography (CT), analyzing typically accessible part standard care. In study, we developed deep learning-based analysis approach over diverse dataset 200 abdominal/pelvic CT scans from patients. The proposed segments adipose tissue using Swin UNEt TRansformers (Swin UNETR) third lumbar vertebrae (L3) level automatically localizes L3 before segmentation. involves first transformer-based learning model for heatmap regression-based vertebra localization UNETR attained 0.92 Dice score 0.87 segmentation, significantly outperforming convolutional benchmarks including 2D U-Net by 2-12% (p-values < 0.033). Moreover, predictions showed high agreement ground-truth areas 0.7-0.93 R

Language: Английский

In vivo X-ray based imaging methods to assess bone quality DOI
Klaus Engelke

Elsevier eBooks, Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 1, 2024

Language: Английский

Citations

0
Vitamin K intake levels are associated with bone health in people aged over 50 years: a NHANES-based survey DOI Creative Commons

Jiankui Guo,

Ziqi Zhou, Jie Gong

et al.

Frontiers in Medicine, Journal Year: 2024, Volume and Issue: 11

Published: Nov. 25, 2024

Background Bone health is important for older adults, and vitamin K (VK) central to regulating bone formation promoting health. However, whether VK can reduce the risk of osteoporosis loss unclear. This study hypothesized that different levels intake exert varying effects on in people aged over 50 years. Methods Individuals above years were recruited from National Health Nutrition Examination Survey. intake, based 24-h dietary recall, was divided into three groups, namely high, medium, low by sex tertile. Weighted multiple logistic regression used investigate at femoral neck, trochanter, intertrochanter, total femur, lumbar spine, overall. Results included 5,075 individuals. Of them, 1,001 (18%) had (808 women, 83%) 2,226 (46%) osteopenia (1,076 54%). Overall, a medium level associated with reduced loss. In medium- [odds ratio, OR (95% confidence interval, CI): 0.66(0.47, 0.93)] high-level [OR 0.71(0.52, 0.98)] decreased osteoporosis. contrast, only medium-level 0.58(0.41, 0.81)]. Similar results obtained spine. men, neck 0.66(0.48, 0.90)], whereas corresponded spine 0.68(0.47, 0.99)]. Nonetheless, did not affect Conclusion demonstrates sex- bone-site-specific variations associations between individuals Further large-scale cohort studies or randomized controlled trials are warranted explore regardless their site.

Language: Английский

Citations

0
Deep Learning-Based Body Composition Analysis for Cancer Patients Using Computed Tomographic Imaging DOI
İlkay Yıldız Potter,

Maria Virginia Velasquez-Hammerle,

Ara Nazarian

et al.

Deleted Journal, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 11, 2024

Malnutrition is a commonly observed side effect in cancer patients, with 30-85% worldwide prevalence this population. Existing malnutrition screening tools miss ~ 20% of at-risk patients at initial and do not capture the abnormal body composition phenotype. Meanwhile, gold-standard clinical criteria to diagnose use changes as key parameters, particularly fat skeletal muscle mass loss. Diagnostic imaging, such computed tomography (CT), analyzing typically accessible part standard care. In study, we developed deep learning-based analysis approach over diverse dataset 200 abdominal/pelvic CT scans from patients. The proposed segments adipose tissue using Swin UNEt TRansformers (Swin UNETR) third lumbar vertebrae (L3) level automatically localizes L3 before segmentation. involves first transformer-based learning model for heatmap regression-based vertebra localization UNETR attained 0.92 Dice score 0.87 segmentation, significantly outperforming convolutional benchmarks including 2D U-Net by 2-12% (p-values < 0.033). Moreover, predictions showed high agreement ground-truth areas 0.7-0.93 R

Language: Английский

Citations

0