Fully Automated Deep Learning Tool for Sarcopenia Assessment on CT: L1 Versus L3 Vertebral Level Muscle Measurements for Opportunistic Prediction of Adverse Clinical Outcomes

Pickhardt, Perry J.; Pérez, Alejandro; Garrett, John W.; Graffy, Peter M; Zea, Ryan; Summers, Ronald M.

doi:10.2214/ajr.21.26486

Cited by 30 publications

(38 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…By decomposing the C3 detection and SM segmentation problem into two separate tasks, we ensure that accurate representations of patient anatomy are identified by the models (C3 region) and subsequently maximize performance for SM segmentation. While previous SM auto-segmentation studies often required specific slices as model inputs ( 21 , 26 ) or utilized separate pre-processing software ( 23 , 25 ), multi-stage deep learning methods have recently been adapted in this domain as well ( 22 , 24 ). Both the 2D and 3D ResUNet models that make up our segmentation pipeline had high performance, with mean DSC values in the test set above 0.95.…”

Section: Discussionmentioning

confidence: 99%

Deep learning auto-segmentation of cervical skeletal muscle for sarcopenia analysis in patients with head and neck cancer

et al. 2022

View full text Add to dashboard Cite

Background/PurposeSarcopenia is a prognostic factor in patients with head and neck cancer (HNC). Sarcopenia can be determined using the skeletal muscle index (SMI) calculated from cervical neck skeletal muscle (SM) segmentations. However, SM segmentation requires manual input, which is time-consuming and variable. Therefore, we developed a fully-automated approach to segment cervical vertebra SM.Materials/Methods390 HNC patients with contrast-enhanced CT scans were utilized (300-training, 90-testing). Ground-truth single-slice SM segmentations at the C3 vertebra were manually generated. A multi-stage deep learning pipeline was developed, where a 3D ResUNet auto-segmented the C3 section (33 mm window), the middle slice of the section was auto-selected, and a 2D ResUNet auto-segmented the auto-selected slice. Both the 3D and 2D approaches trained five sub-models (5-fold cross-validation) and combined sub-model predictions on the test set using majority vote ensembling. Model performance was primarily determined using the Dice similarity coefficient (DSC). Predicted SMI was calculated using the auto-segmented SM cross-sectional area. Finally, using established SMI cutoffs, we performed a Kaplan-Meier analysis to determine associations with overall survival.ResultsMean test set DSC of the 3D and 2D models were 0.96 and 0.95, respectively. Predicted SMI had high correlation to the ground-truth SMI in males and females (r>0.96). Predicted SMI stratified patients for overall survival in males (log-rank p = 0.01) but not females (log-rank p = 0.07), consistent with ground-truth SMI.ConclusionWe developed a high-performance, multi-stage, fully-automated approach to segment cervical vertebra SM. Our study is an essential step towards fully-automated sarcopenia-related decision-making in patients with HNC.

show abstract

Section: Discussionmentioning

confidence: 99%

Deep learning auto-segmentation of cervical skeletal muscle for sarcopenia analysis in patients with head and neck cancer

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Bei der Geschlechterverteilung werden je nach verwendeter Definition und Population teils bei Männern, teils aber auch bei Frauen höhere Prävalenzen angegeben oder geschätzt [6]. Es existieren bereits einige Studien mit großen Fallzahlen, die sich mit der Prävalenz, den Risikofaktoren und dem Screening der Sarkopenie beschäftigt haben [7][8][9][10]. So ergab die "UK Biobank"-Studie mit 168 682 Teilnehmern ein erhöhtes Osteoporoserisiko bei präsarkopenen Männern und sarkopenen Frauen [10].…”

Section: Definitionunclassified

“…In some studies, muscle and fat tissue in CT and MRI datasets has already been analyzed using artificial intelligence (AI) 9 54 55 56 57 . In their study including 1143 CT datasets, Nowak et al used two neuronal networks 57 .…”

Section: Artificial Intelligence In Sarcopenia Diagnosismentioning

confidence: 99%

“…In einigen Studien wurden die Muskulatur und das Fettgewebe in CT- und MRT-Datensätzen bereits mittels künstlicher Intelligenz (KI) analysiert 9 54 55 56 57 . Nowak et al verwendeten in ihrer Studie mit 1143 CT-Datensätzen 2 neuronale Netze 57 .…”

Section: Künstliche Intelligenz In Der Sarkopenie-diagnostikunclassified

“…With respect to gender distribution, the specified or estimated prevalence is higher in men or in women depending on the definition and population [6]. There are already some studies with large case numbers addressing the prevalence, risk factors, and screening of sarcopenia [7][8][9][10]. According to the "UK Biobank" study including 168 682 participants, pre-sarcopenic men and sarcopenic women have an elevated risk of osteoporosis [10].…”

Section: Definitionmentioning

confidence: 99%

See 2 more Smart Citations

Sarcopenia – Definition, Radiological Diagnosis, Clinical Significance

Vogele

Otto

Sollmann

et al. 2023

Rofo

View full text Add to dashboard Cite

Background Sarcopenia is an age-related syndrome characterized by a loss of muscle mass and strength. As a result, the independence of the elderly is reduced and the hospitalization rate and mortality increase. The onset of sarcopenia often begins in middle age due to an unbalanced diet or malnutrition in association with a lack of physical activity. This effect is intensified by concomitant diseases such as obesity or metabolic diseases including diabetes mellitus. Method With effective preventative diagnostic procedures and specific therapeutic treatment of sarcopenia, the negative effects on the individual can be reduced and the negative impact on health as well as socioeconomic effects can be prevented. Various diagnostic options are available for this purpose. In addition to basic clinical methods such as measuring muscle strength, sarcopenia can also be detected using imaging techniques like dual X-ray absorptiometry (DXA), computed tomography (CT), magnetic resonance imaging (MRI), and sonography. DXA, as a simple and cost-effective method, offers a low-dose option for assessing body composition. With cross-sectional imaging techniques such as CT and MRI, further diagnostic possibilities are available, including MR spectroscopy (MRS) for noninvasive molecular analysis of muscle tissue. CT can also be used in the context of examinations performed for other indications to acquire additional parameters of the skeletal muscles (opportunistic secondary use of CT data), such as abdominal muscle mass (total abdominal muscle area – TAMA) or the psoas as well as the pectoralis muscle index. The importance of sarcopenia is already well studied for patients with various tumor entities and also infections such as SARS-COV2. Results and Conclusion Sarcopenia will become increasingly important, not least due to demographic changes in the population. In this review, the possibilities for the diagnosis of sarcopenia, the clinical significance, and therapeutic options are described. In particular, CT examinations, which are repeatedly performed on tumor patients, can be used for diagnostics. This opportunistic use can be supported by the use of artificial intelligence. Key Points: Citation Format

show abstract

Development and Validation of an Automated Image-Based Deep Learning Platform for Sarcopenia Assessment in Head and Neck Cancer

Saraf

Ravipati

et al. 2023

JAMA Netw Open

View full text Add to dashboard Cite

ImportanceSarcopenia is an established prognostic factor in patients with head and neck squamous cell carcinoma (HNSCC); the quantification of sarcopenia assessed by imaging is typically achieved through the skeletal muscle index (SMI), which can be derived from cervical skeletal muscle segmentation and cross-sectional area. However, manual muscle segmentation is labor intensive, prone to interobserver variability, and impractical for large-scale clinical use.ObjectiveTo develop and externally validate a fully automated image-based deep learning platform for cervical vertebral muscle segmentation and SMI calculation and evaluate associations with survival and treatment toxicity outcomes.Design, Setting, and ParticipantsFor this prognostic study, a model development data set was curated from publicly available and deidentified data from patients with HNSCC treated at MD Anderson Cancer Center between January 1, 2003, and December 31, 2013. A total of 899 patients undergoing primary radiation for HNSCC with abdominal computed tomography scans and complete clinical information were selected. An external validation data set was retrospectively collected from patients undergoing primary radiation therapy between January 1, 1996, and December 31, 2013, at Brigham and Women’s Hospital. The data analysis was performed between May 1, 2022, and March 31, 2023.ExposureC3 vertebral skeletal muscle segmentation during radiation therapy for HNSCC.Main Outcomes and MeasuresOverall survival and treatment toxicity outcomes of HNSCC.ResultsThe total patient cohort comprised 899 patients with HNSCC (median [range] age, 58 [24-90] years; 140 female [15.6%] and 755 male [84.0%]). Dice similarity coefficients for the validation set (n = 96) and internal test set (n = 48) were 0.90 (95% CI, 0.90-0.91) and 0.90 (95% CI, 0.89-0.91), respectively, with a mean 96.2% acceptable rate between 2 reviewers on external clinical testing (n = 377). Estimated cross-sectional area and SMI values were associated with manually annotated values (Pearson r = 0.99; P &lt; .001) across data sets. On multivariable Cox proportional hazards regression, SMI-derived sarcopenia was associated with worse overall survival (hazard ratio, 2.05; 95% CI, 1.04-4.04; P = .04) and longer feeding tube duration (median [range], 162 [6-1477] vs 134 [15-1255] days; hazard ratio, 0.66; 95% CI, 0.48-0.89; P = .006) than no sarcopenia.Conclusions and RelevanceThis prognostic study’s findings show external validation of a fully automated deep learning pipeline to accurately measure sarcopenia in HNSCC and an association with important disease outcomes. The pipeline could enable the integration of sarcopenia assessment into clinical decision making for individuals with HNSCC.

show abstract

Fully Automated Deep Learning Tool for Sarcopenia Assessment on CT: L1 Versus L3 Vertebral Level Muscle Measurements for Opportunistic Prediction of Adverse Clinical Outcomes

Cited by 30 publications

References 33 publications

Deep learning auto-segmentation of cervical skeletal muscle for sarcopenia analysis in patients with head and neck cancer

Deep learning auto-segmentation of cervical skeletal muscle for sarcopenia analysis in patients with head and neck cancer

Sarcopenia – Definition, Radiological Diagnosis, Clinical Significance

Development and Validation of an Automated Image-Based Deep Learning Platform for Sarcopenia Assessment in Head and Neck Cancer

Contact Info

Product

Resources

About