End-to-end automated body composition analyses with integrated quality control for opportunistic assessment of sarcopenia in CT

Abstract Objectives To develop a pipeline for automated body composition analysis and skeletal muscle assessment with integrated quality control for large-scale application in opportunistic imaging. Methods First, a convolutional neural network for extraction of a single slice at the L3/L4 lumbar level was developed on CT scans of 240 patients applying the nnU-Net framework. Second, a 2D competitive dense fully convolutional U-Net for segmentation of visceral and subcutaneous adipose tissue (VAT, SAT), skeletal muscle (SM), and subsequent determination of fatty muscle fraction (FMF) was developed on single CT slices of 1143 patients. For both steps, automated quality control was integrated by a logistic regression model classifying the presence of L3/L4 and a linear regression model predicting the segmentation quality in terms of Dice score. To evaluate the performance of the entire pipeline end-to-end, body composition metrics, and FMF were compared to manual analyses including 364 patients from two centers. Results Excellent results were observed for slice extraction (z-deviation = 2.46 ± 6.20 mm) and segmentation (Dice score for SM = 0.95 ± 0.04, VAT = 0.98 ± 0.02, SAT = 0.97 ± 0.04) on the dual-center test set excluding cases with artifacts due to metallic implants. No data were excluded for end-to-end performance analyses. With a restrictive setting of the integrated segmentation quality control, 39 of 364 patients were excluded containing 8 cases with metallic implants. This setting ensured a high agreement between manual and fully automated analyses with mean relative area deviations of ΔSM = 3.3 ± 4.1%, ΔVAT = 3.0 ± 4.7%, ΔSAT = 2.7 ± 4.3%, and ΔFMF = 4.3 ± 4.4%. Conclusions This study presents an end-to-end automated deep learning pipeline for large-scale opportunistic assessment of body composition metrics and sarcopenia biomarkers in clinical routine. Key Points • Body composition metrics and skeletal muscle quality can be opportunistically determined from routine abdominal CT scans. • A pipeline consisting of two convolutional neural networks allows an end-to-end automated analysis. • Machine-learning-based quality control ensures high agreement between manual and automatic analysis.

Download Full-text

CACHEXIO: Evaluation of body composition changes and immunotherapy in patients with metastatic cancer.

Journal of Clinical Oncology ◽

10.1200/jco.2018.36.34_suppl.209 ◽

2018 ◽

Vol 36 (34_suppl) ◽

pp. 209-209

Author(s):

Eric Roeland ◽

Areej El-Jawahri ◽

Nora Horick ◽

Sandahl H Nelson ◽

Andrea Gallivan ◽

...

Keyword(s):

Skeletal Muscle ◽

Health Care ◽

Body Composition ◽

Adipose Tissue ◽

Health Care Utilization ◽

Metastatic Cancer ◽

Ct Scans ◽

Care Utilization ◽

Mixed Effect ◽

Composition Changes

209 Background: Given body composition predicts toxicity for patients receiving cytotoxic chemotherapy, we explored changes in body composition and biomarkers as predictors of immune-related adverse events (irAEs) and health care utilization. Methods: We conducted a longitudinal study of patients with metastatic solid tumor receiving immunotherapy (07/2014-10/2017). Eligible patients had a computed tomography (CT) scan prior to first-line immunotherapy with at least two additional CT scans at three, six or nine months after immunotherapy initiation. We analyzed body composition using cross-sectional CT scans at the third lumbar vertebra. We utilized mixed effect linear regression models to assess longitudinal changes in body composition (weight, skeletal muscle, total adipose). We examined associations of baseline body composition and biomarkers (albumin, neutrophil-lymphocyte ratio (NLR)) with the incidence of irAEs and healthcare utilization (hospitalizations/ED visits) using logistic regression. Results: Of 140 patients treated with immunotherapy, 61 met inclusion criteria. The majority (80%) received prior chemotherapy and the most common malignancies included lung (26%), head and neck (20%), and melanoma (20%). We found that one-third (n=19) experienced an irAE and colitis (53%) was the most common irAE. Patients experienced substantial weight loss over time (B= -1.88, p<0.001) with a decrease both in skeletal muscle (B= -3.08, p=0.001) and total adipose tissue (B =-50.44, p<0.001). We found greater skeletal muscle at baseline was associated with lower risk of health care utilization (OR 0.98, 95% CI: 0.965-0.998, p=0.03). We observed no association with biomarkers and/or body composition variables with irAEs or health care utilization. Conclusions: Patients with metastatic cancer receiving immunotherapy lose weight including skeletal muscle and adipose tissue. Aside from higher baseline skeletal muscle predicting less health care utilization, we did not observe any other associations between body composition changes and irAEs or health care utilization.

Download Full-text

Longitudinal Changes in CT Body Composition in Patients Undergoing Surgery for Colorectal Cancer and Associations With Peri-Operative Clinicopathological Characteristics

Frontiers in Nutrition ◽

10.3389/fnut.2021.678410 ◽

2021 ◽

Vol 8 ◽

Author(s):

Ross D. Dolan ◽

Tanvir Abbass ◽

Wei M. J. Sim ◽

Arwa S. Almasaudi ◽

Ly B. Dieu ◽

...

Keyword(s):

Colorectal Cancer ◽

Skeletal Muscle ◽

Body Composition ◽

Inflammatory Response ◽

Muscle Mass ◽

Skeletal Muscle Mass ◽

Systemic Inflammatory Response ◽

Ct Scans ◽

Lower Percentage

There is evidence for the direct association between body composition, the magnitude of the systemic inflammatory response, and outcomes in patients with colorectal cancer. Patients with a primary operable disease with and without follow-up CT scans were examined in this study. CT scans were used to define the presence and changes in subcutaneous fat, visceral fat, skeletal muscle mass, and skeletal muscle density (SMD). In total, 804 patients had follow-up scans and 83 patients did not. Furthermore, 783 (97%) patients with follow-up scans and 60 (72%) patients without follow-up scans were alive at 1 year. Patients with follow-up scans were younger (p < 0.001), had a lower American Society of Anaesthesiology Grade (p < 0.01), underwent a laparoscopic surgery (p < 0.05), had a higher BMI (p < 0.05), a higher skeletal muscle index (SMI) (p < 0.01), a higher SMD (p < 0.01), and a better 1-year survival (p < 0.001). Overall only 20% of the patients showed changes in their SMI (n = 161) and an even lower percentage of patients showed relative changes of 10% (n = 82) or more. In conclusion, over the period of ~12 months, a low–skeletal muscle mass was associated with a systemic inflammatory response and was largely maintained following surgical resection.

Download Full-text

Body composition and treatment-related toxicity in pediatric lymphoma.

Journal of Clinical Oncology ◽

10.1200/jco.2020.38.15_suppl.e22503 ◽

2020 ◽

Vol 38 (15_suppl) ◽

pp. e22503-e22503

Author(s):

Aman Wadhwa ◽

Kandice Barnett ◽

Chen Dai ◽

Joshua Richman ◽

Andrew Michael McDonald ◽

...

Keyword(s):

Skeletal Muscle ◽

Body Composition ◽

Adipose Tissue ◽

Age At Diagnosis ◽

Muscle Quality ◽

Ct Scans ◽

Hematologic Toxicity ◽

Cancer Type ◽

Skeletal Muscle Index ◽

Grade 3

e22503 Background: Body composition is an emerging predictor of toxicity and survival in older adults with cancer ( Shachar, Eur J Can, 2016); however, its role in pediatric cancer is not known. We examined body composition (using computed-tomography [CT] scans at the 3rd lumbar level) in children with lymphoma (Hodgkin [HL] and non-Hodgkin [NHL]) at cancer diagnosis and examined its association with treatment-related toxicities. Methods: We constructed a retrospective cohort of 87 consecutive children (HL: n = 45; NHL: n = 42) diagnosed between 2000 and 2015 (2-21y at diagnosis) with pretreatment abdominal CT scans. Body composition was assessed using sliceOmatic (TomoVision) and included skeletal muscle index (SMI, cm2/m2), skeletal muscle density (SMD: Hounsfield units [HU]), and height-adjusted total adipose tissue (hTAT: sum of visceral, intramuscular and subcutaneous adipose tissue, cm2/m2). For the analysis, we used skeletal muscle gauge (SMG = SMI x SMD, expressed per 1000 in arbitrary units [AU]) and hTAT. Sociodemographics, disease and treatment details, as well as toxicities (CTCAE v5) were abstracted from medical records. Proportion of chemotherapy cycles with grade 4 hematologic or grade 3-4 non-hematologic toxicities were calculated (percent toxicity). Generalized linear regression models were constructed to examine associations between body composition metrics and toxicities, adjusting for age at diagnosis, gender, race/ethnicity and lymphoma subtype. Results: Median age at diagnosis was 12.9y (range, 2-18.5y); 60.9% males; 60.4% non-Hispanic white. Median BMI%ile was 62 (0-99), median SMG was 2.2AU (0.9-3.7) and median hTAT was 20.1 cm2/m2 (0.04-226.7). Overall, the mean percent toxicity for grade 4 hematologic and grade 3-4 non-hematologic toxicity was 38.9% (±32.6) and 31.4% (±32.6) respectively. Correlation was poor between SMG and BMI%ile ( R2= 0.04), SMG and hTAT ( R2= -0.01) and moderate between hTAT and BMI%ile ( R2= 0.4). SMG was significantly associated with grade 4 hematologic percent toxicity ( β= -18, P= 0.007) after adjusting for hTAT and cancer type. BMI%ile was not associated with grade 4 hematologic percent toxicity ( β= -0.09, P= 0.5). Non-hematologic percent toxicity was not associated with BMI%ile, hTAT or SMG. Conclusions: In this first study of its kind, we find that children with poorer muscle quality are more likely to experience grade 4 hematologic toxicities. These findings form the basis for larger studies to incorporate body composition when developing prediction models for chemotherapy-related toxicity and disease outcomes.

Download Full-text

Quality Control of Data in a Large-Scale Cancer Register Program

Methods of Information in Medicine ◽

10.1055/s-0038-1636339 ◽

1966 ◽

Vol 05 (02) ◽

pp. 67-74 ◽

Cited By ~ 5

Author(s):

W. I. Lourie ◽

W. Haenszeland

Keyword(s):

United States ◽

Quality Control ◽

Cancer Patients ◽

Large Scale ◽

The United States ◽

Newly Diagnosed ◽

Related Data ◽

Cancer Register ◽

Independent Review ◽

End Results

Quality control of data collected in the United States by the Cancer End Results Program utilizing punchcards prepared by participating registries in accordance with a Uniform Punchcard Code is discussed. Existing arrangements decentralize responsibility for editing and related data processing to the local registries with centralization of tabulating and statistical services in the End Results Section, National Cancer Institute. The most recent deck of punchcards represented over 600,000 cancer patients; approximately 50,000 newly diagnosed cases are added annually.Mechanical editing and inspection of punchcards and field audits are the principal tools for quality control. Mechanical editing of the punchcards includes testing for blank entries and detection of in-admissable or inconsistent codes. Highly improbable codes are subjected to special scrutiny. Field audits include the drawing of a 1-10 percent random sample of punchcards submitted by a registry; the charts are .then reabstracted and recoded by a NCI staff member and differences between the punchcard and the results of independent review are noted.

Download Full-text

Phoneme-to-Grapheme Conversion Based Large-Scale Pre-Training for End-to-End Automatic Speech Recognition

10.21437/interspeech.2020-1930 ◽

2020 ◽

Author(s):

Ryo Masumura ◽

Naoki Makishima ◽

Mana Ihori ◽

Akihiko Takashima ◽

Tomohiro Tanaka ◽

...

Keyword(s):

Speech Recognition ◽

Automatic Speech Recognition ◽

Large Scale ◽

End To End

Download Full-text

Role of the Appendicular Skeletal Muscle Index for Predicting the Recurrence-Free Survival of Head and Neck Cancer

Diagnostics ◽

10.3390/diagnostics11020309 ◽

2021 ◽

Vol 11 (2) ◽

pp. 309

Author(s):

Kun-Yun Yeh ◽

Hang Huong Ling ◽

Shu-Hang Ng ◽

Cheng-Hsu Wang ◽

Pei-Hung Chang ◽

...

Keyword(s):

Skeletal Muscle ◽

Body Composition ◽

Head And Neck Cancer ◽

Mass Loss ◽

Lean Mass ◽

Recurrence Free Survival ◽

Skeletal Muscle Index ◽

Free Survival ◽

Prognostic Predictor ◽

Appendicular Skeletal Muscle

Background: This study investigates whether the appendicular skeletal muscle index (ASMI) was an independent prognostic predictor for patients with locally advanced head and neck cancer (LAHNC) receiving concurrent chemoradiotherapy (CCRT) and whether there were any differences in lean mass loss in different body regions during CCRT. Methods: In this prospective study, we analyzed the clinicopathological variables and the total body composition data before and after treatment. The factors associated with the 2-year recurrence-free survival rate (RFSR) were analyzed via logistic regression analysis. Results: A total of 98 patients were eligible for analysis. The body weight, body mass index, and all parameters of body composition significantly decreased after CCRT. The pretreatment ASMI was the only independent prognostic factor for predicting the 2-year RFSR (hazard ratio, 0.235; 95% confidence interval, 0.062–0.885; p = 0.030). There was at least 5% reduction in total lean and fat mass (p < 0.001); however, the highest lean mass loss was observed in the arms (9.5%), followed by the legs (7.2%), hips (7.1%), waist (4.7%), and trunk (3.6%). Conclusions: The pretreatment ASMI was the only independent prognostic predictor for the 2-year RFSR of LAHNC patients undergoing CCRT. Asynchronous loss of lean mass may be observed in different body parts after CCRT.

Download Full-text

Fully automated preoperative segmentation of temporal bone structures from clinical CT scans

Scientific Reports ◽

10.1038/s41598-020-80619-0 ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

C. A. Neves ◽

E. D. Tran ◽

I. M. Kessler ◽

N. H. Blevins

Keyword(s):

Facial Nerve ◽

Inner Ear ◽

Temporal Bone ◽

Hausdorff Distance ◽

Sigmoid Sinus ◽

Ct Scans ◽

Automated Segmentation ◽

Dice Coefficient ◽

Lateral Skull Base ◽

End To End

AbstractMiddle- and inner-ear surgery is a vital treatment option in hearing loss, infections, and tumors of the lateral skull base. Segmentation of otologic structures from computed tomography (CT) has many potential applications for improving surgical planning but can be an arduous and time-consuming task. We propose an end-to-end solution for the automated segmentation of temporal bone CT using convolutional neural networks (CNN). Using 150 manually segmented CT scans, a comparison of 3 CNN models (AH-Net, U-Net, ResNet) was conducted to compare Dice coefficient, Hausdorff distance, and speed of segmentation of the inner ear, ossicles, facial nerve and sigmoid sinus. Using AH-Net, the Dice coefficient was 0.91 for the inner ear; 0.85 for the ossicles; 0.75 for the facial nerve; and 0.86 for the sigmoid sinus. The average Hausdorff distance was 0.25, 0.21, 0.24 and 0.45 mm, respectively. Blinded experts assessed the accuracy of both techniques, and there was no statistical difference between the ratings for the two methods (p = 0.93). Objective and subjective assessment confirm good correlation between automated segmentation of otologic structures and manual segmentation performed by a specialist. This end-to-end automated segmentation pipeline can help to advance the systematic application of augmented reality, simulation, and automation in otologic procedures.

Download Full-text

Magnetization Transfer Ratio of Peripheral Nerve and Skeletal Muscle

Clinical Neuroradiology ◽

10.1007/s00062-021-01067-5 ◽

2021 ◽

Author(s):

Olivia Fösleitner ◽

Véronique Schwehr ◽

Tim Godel ◽

Fabian Preisner ◽

Philipp Bäumer ◽

...

Keyword(s):

Skeletal Muscle ◽

Body Weight ◽

Sciatic Nerve ◽

Linear Regression ◽

Peripheral Nerve ◽

Regression Model ◽

Multiple Linear Regression ◽

Linear Regression Model ◽

Magnetization Transfer ◽

Multiple Linear Regression Model

Abstract Purpose To assess the correlation of peripheral nerve and skeletal muscle magnetization transfer ratio (MTR) with demographic variables. Methods In this study 59 healthy adults evenly distributed across 6 decades (mean age 50.5 years ±17.1, 29 women) underwent magnetization transfer imaging and high-resolution T2-weighted imaging of the sciatic nerve at 3 T. Mean sciatic nerve MTR as well as MTR of biceps femoris and vastus lateralis muscles were calculated based on manual segmentation on six representative slices. Correlations of MTR with age, body height, body weight, and body mass index (BMI) were expressed by Pearson coefficients. Best predictors for nerve and muscle MTR were determined using a multiple linear regression model with forward variable selection and fivefold cross-validation. Results Sciatic nerve MTR showed significant negative correlations with age (r = −0.47, p < 0.001), BMI (r = −0.44, p < 0.001), and body weight (r = −0.36, p = 0.006) but not with body height (p = 0.55). The multiple linear regression model determined age and BMI as best predictors for nerve MTR (R2 = 0.40). The MTR values were different between nerve and muscle tissue (p < 0.0001), but similar between muscles. Muscle MTR was associated with BMI (r = −0.46, p < 0.001 and r = −0.40, p = 0.002) and body weight (r = −0.36, p = 0.005 and r = −0.28, p = 0.035). The BMI was selected as best predictor for mean muscle MTR in the multiple linear regression model (R2 = 0.26). Conclusion Peripheral nerve MTR decreases with higher age and BMI. Studies that assess peripheral nerve MTR should consider age and BMI effects. Skeletal muscle MTR is primarily associated with BMI but overall less dependent on demographic variables.

Download Full-text