Published online Sep 8, 2026. doi: 10.35713/aic.v7.i1.116464
Revised: December 4, 2025
Accepted: February 2, 2026
Published online: September 8, 2026
Processing time: 294 Days and 17.9 Hours
For pediatric patients with osteosarcoma, nutritional status is associated with treatment response and survival outcome.
To explore computed tomography (CT)-derived body composition using artificial intelligence (AI)-based tissue segmentation in pediatric patients with osteosar
In this study, the body composition of 133 patients aged ≤ 18 years with young osteosarcoma at the third lumbar vertebra level was analyzed on a retrospective dataset using an AI-based software tool (Visage Imaging). All patients received neoadjuvant chemotherapy and surgical resection who underwent abdominal CT scanning before and after chemotherapy. The primary outcome was OS, and the secondary outcome was response to neoadjuvant chemotherapy. The CT-derived factors, including skeletal muscle index, subcutaneous and visceral adipose tissue index, and skeletal muscle density (SMD), and other nutrition factor, such as systemic immune-inflammation index and prognostic nutritional index were collected before and after chemotherapy. Changes (Δ) in body composition pa
On our results age at diagnosis, post-chemotherapy systemic immune-inflammation index, and ΔSMD had prognostic value for classification of responders and non-responders to neoadjuvant chemotherapy at multi
AI-based analysis of third lumbar vertebra body composition on CT images is feasible in pediatric patients receiving neoadjuvant chemotherapy. ΔSMD is a prognostic indicator that a great value significantly predicts the poor treatment response and short OS. Using CT-derived and other nutritional predictors may conduct optimized nutrition support and appropriate selection of individualized therapeutic regimen.
Core Tip: Using artificial intelligence based automated computed tomography analysis of the third lumbar vertebra body composition in 133 pediatric osteosarcoma patients, early reductions in skeletal muscle density during neoadjuvant chemotherapy were independently associated with poor histologic response and shorter overall survival. An integrated index combining age, postchemotherapy inflammatory indicator and change of skeletal muscle density improved prediction of treatment response and prognosis, suggesting that early computed tomography-derived muscle metrics can guide timely nutritional and rehabilitative interventions to potentially improve outcomes.
- Citation: Yang YH, Li Y. Prognostic value of early changes in artificial intelligence-based computed tomography-measured body composition in pediatric osteosarcoma receiving neoadjuvant chemotherapy. Artif Intell Cancer 2026; 7(1): 116464
- URL: https://www.wjgnet.com/2644-3228/full/v7/i1/116464.htm
- DOI: https://dx.doi.org/10.35713/aic.v7.i1.116464
Osteosarcoma shows the highest incidence of malignant bone tumor in children and adolescents[1]. The prognosis of osteosarcoma after pure surgical resection is poor considering the high invasiveness[2]. Neoadjuvant chemotherapy combined with surgery shows abilities in controlling the primary tumor and its micro-metastatic lesions, which prolongs overall survival (OS) and optimize patients’ outcomes[3]. There have been some prognostic factors associated with histopathologic response to neoadjuvant chemotherapy, such as age, primary tumor volume, and anatomic location[4]. More advanced biomarkers are needed to identify patients at risk for poor responders to neoadjuvant chemotherapy, and the early intervention for these patients have greatly improved long-term survival.
Cancer cachexia, a metabolic disorder with loss of adipose and skeletal muscle mass, has recently been identified as a prognostic factor for osteosarcoma[5]. Cancer cachexia has been found association with survival outcomes by influencing dose-limiting toxicity and response to chemotherapy[6,7]. As for children, the estimation of body mass index reveals limited information about the real situation of cancer cachexia[8]. Routine radiological examination represents quantitative information of muscle and adipose tissue for diagnosis of cancer cachexia[9,10]. For patients with osteosarcoma receiving neoadjuvant chemotherapy, the pre- and post-chemotherapy whole-body imaging scanning represents the patient’s reflection for treatment, and also shows the changes of body composition during treatment which may have prognostic value for therapeutic response and OS.
The abdominal visceral tissue and skeletal muscle measured at the level of the third lumbar vertebra (LV3) have been found reduction during chemotherapy and association with long-term survival[11,12]. Artificial intelligence (AI)-based automated tissue segmentation on computed tomography (CT) imaging can increase delineation accuracy and decrease processing time compared with manual segmentation. It remains unclear whether AI-assisted CT quantified body composition features indicate prognostication for response to neoadjuvant chemotherapy and OS in osteosarcoma.
This study was intended as a proof-of-concept to determine whether early AI-quantified changes in LV3 body composition during neoadjuvant chemotherapy were associated with histopathological response and OS, and to inform the development of predictive models. The aims of present study were to establish robust associations between dynamic body composition metrics and outcomes, to assess the feasibility of integrating these metrics into prognostic indices, and to outline a validation pathway toward a generalizable predictive tool for clinical use.
We retrospectively selected patients ≤ 18 years old with histologically confirmed osteosarcoma who underwent neoadjuvant chemotherapy and post-chemotherapy surgery. All eligible patients were histologically diagnosed as osteosarcoma from January 2010 to December 2020, and underwent CT abdominal scanning before and after neoadjuvant chemotherapy. The study was approved by the institutional review board.
The inclusion criteria were as follows: (1) Patients were histologically diagnosed as osteosarcoma by pre-treatment biopsy samples; (2) Age at diagnosis was ≤ 18 years old; (3) Patients underwent neoadjuvant chemotherapy before surgical resection; (4) Radical resection was performed following neoadjuvant chemotherapy; and (5) Pre-chemotherapy and post-chemotherapy CT abdominal images were available. We excluded patients who died during the period of neoadjuvant chemotherapy or before the surgical procedure, or had received any other kind of treatment before neoadjuvant chemotherapy. Of 260 patients diagnosed with osteosarcoma from January 2010 to December 2020 were screened. Exclusion criteria were: (1) Age > 18 at diagnosis; (2) Lack of both pre- and post-neoadjuvant abdominal CT; (3) Death prior to surgery; (4) Receipt of other systemic therapy prior to neoadjuvant chemotherapy; and (5) Incomplete clinical follow-up. After applying these criteria, 133 patients remained and comprised the analytic cohort. A STROBE-style flowchart summarizing identification, exclusion and final cohort was provided in Figure 1.
We applied magnetic resonance imaging (MRI) examination to evaluate local invasion of the affected limb in pediatric osteosarcoma. The whole-body radiological examinations were performed, including the brain MRI and the CT of thorax, abdomen, and pelvis region for identification of the distant metastases. situation. Demographic characteristics were obtained at the time of diagnosis. The indicators from blood and biochemical tests were collected before and after neoadjuvant chemotherapy. The calculation of systemic immune-inflammation index (SII) was based on the pre-chemotherapy and post-chemotherapy indexes from peripheral blood tests with the formula: SII = platelet count × absolute neutrophil count/absolute lymphocyte count (1)[13]. The pre-chemotherapy or post-chemotherapy prognostic nutritional index (PNI) was calculated by the formula: PNI = albumin level (g/L) + 5 × absolute lymphocyte count (109/L) (2)[14]. The histological diagnosis was confirmed by biopsy samples via fine-needle aspiration biopsy or open biopsy. A neoadjuvant chemotherapy cycle included doxorubicin 75 mg/m2 and cisplatin 100 mg/m2. This combination lasted for 5-7 days with an interval of 3-4 weeks for 1-3 cycles.
The tumor size was measured by MRI scanning of the affected limb before and after neoadjuvant chemotherapy. Complete response, partial response (PR), progressive disease and stable disease were evaluated based on the pre-chemotherapy and post-chemotherapy data[15]. Patients with complete response or PR were considered as responders, and patients with PR or progressive disease were considered as non-responders.
The OS was defined as the duration from the date of surgical procedures to the date of death or the last follow-up time, December 2018. The survival status was determined by the follow-up which was accomplished every 1-3 months since discharge through telephone calls, e-mails or re-admission.
A 64-slice CT scanner was used for abdominal examination with slice thickness of 3 mm. The image acquisition parameter for reconstruction was 5 mm thickness with 5 mm intervals using bone or/and soft tissue algorithms. All patients underwent CT scanning prior to neoadjuvant chemotherapy, referred to as pre-chemotherapy CT, and after neoadjuvant chemotherapy for evaluation of treatment response, referred to as post-chemotherapy CT. All CT scans were performed using institutionally standardized protocols with slice thickness ranged with 5 mm. Both non-contrast and contrast-enhanced exams were included with recording contrast phase and scanner model for each exam and examined its effect on skeletal muscle density (SMD). Automated segmentation was performed using Visage Imaging v7.1 using algorithm based on U-Net architecture. Automated masks were visually inspected by two readers that scans with gross segmentation errors (< 10% of cases) were re-segmented manually.
Body composition measurements were performed at the level of the LV3. The cross-sectional areas and attenuation of tissues at L3 have been widely validated as representative of whole-body skeletal muscle and adipose tissue mass and correlate strongly with clinical outcomes across malignancies. Automated segmentation at L3 enables reproducible, objective quantification of skeletal muscle area/index and SMD, and therefore is an accepted reference standard for CT-based body composition analysis. For automated segmentation of body composition at LV3 CT images, we applied an AI-based automated software tool (Visage version 7.1., Visage Imaging GmbH, Berlin, Germany) based on a convolutional neural network. The U-net network was composed of nine blocks, including four down-sampling blocks, four up-sampling blocks, and one in-between block[16]. During the training process of our network, three hundred axial CT images at LV3 were applied with augmentation to increase generalization and decrease overfitting. Tissues were separated into skeletal muscle, visceral adipose and subcutaneous adipose, which was marked with different colors respectively. The internal organs, such as such as kidney, liver, spleen, intestine, and pancreas, were not annotated. The manual correction was used in cases with false tissue segmentation subsequently. All segmented classes were measured by the software with the area in square centimeters (cm2) and density in Hounsfield unit automatically on pre-chemotherapy and post-chemotherapy CT images at LV3. The body composition-associated indicators were calculated as follows: (1) SMD was defined as mean density of skeletal muscle; (2) Skeletal muscle index (SMI, cm2/m2) was measure as skeletal muscle area (cm2)/body surface area (m2); (3) Visceral adipose tissue index (cm2/m2) was measure as visceral adipose tissue area (cm2)/body surface area (m2); and (4) Subcutaneous adipose tissue index (SATI, cm2/m2) was measure as subcutaneous adipose tissue area (cm2)/body surface area (m2). An example of AI-based automated segmentation on pre-chemotherapy and post-chemotherapy CT images was shown in Figure 1. Otherwise, we calculated the difference (Δ) between pre-chemotherapy and post-chemotherapy CTs, normalized to a period of 30 days. Positive values of these differences indicated an area loss or a density reduction for each body composition. The calculation formula was listed as follows: ΔSMD/ΔSMI/Δvisceral adipose tissue index/ΔSATI = (3).
In this study, all continuous characteristics were evaluated by median value and their interquartile rage (IQR), and all categorical characteristics were evaluated by n (%). The primary outcome was OS, and the secondary outcome was response to neoadjuvant chemotherapy. The association of variables with response to neoadjuvant chemotherapy was measured via uni- and multi-variable logistic proportional hazard regression models. The selected variables with significant p value in the multi-variable logistic model were combined to generate a integrated index for classification of responders and non-responders. The integrated index was applied for prediction of OS in the following step. The association between variables and OS was assessed by uni- and multivariable Cox proportional hazard regression models. A backward step-down selection process was used to select the prognostic factors independently in the multivariable Cox regression analysis by minimizing the Akaike information criteria. These factors for OS were estimated by Kaplan-Meier analysis, and shown by covariate-adjusted survival curves. The optimal cut-off points for the continuous predictors were calculated by X-tile (version 3.6.1; Yale University School of Medicine, New Haven, CT, United States)[17], which maximized the survival difference between the dichotomized groups. Primary analysis used univariable and multivariable Cox proportional hazards regression and logistic regression for response. To account for longitudinal nature, Δ variables were standardized per 30 days. Internal validation was performed that predictive performance of multivariable models was estimated using bootstrap resampling (1000 replications) to derive optimism-corrected C-index with 5-fold cross-validation to evaluate prediction stability. As an exploratory predictive approach, we trained a random survival forest (RSF) using the randomForestSRC package with hyperparameters tuned by grid search. The variable importance was derived from RSF. For dichotomization of continuous predictors we used X-tile. Missing data were handled by multiple imputation (m = 5) using chained equations that sensitivity analyses used complete cases. All tests were two-sided. Statistical significance was defined as P < 0.05. The statistical analysis was completed by R version 3.6.1.
Demographic and body composition-associated variables were summarized and listed in Table 1. The time interval between pre-chemotherapy and post-chemotherapy CT examinations was measured with the median value of 81 days and the IQR of 22 days. Anatomic primary tumor locations were further categorized as follows: Distal femur, proximal tibia, proximal humerus, pelvis, spine, proximal femur and other sites. The distribution of tumor locations was: Distal femur: 52 (39.1%), proximal tibia: 38 (28.6%), proximal humerus: 14 (10.5%), pelvis: 14 (10.5%), and spine: 7 (5.3%). Overall, 112 patients (84.2%) had limb tumors and 21 (15.8%) had axial tumors. The median OS was 32 months with the IQR of 43 months. The 3-year OS rate was 81.20% (108/133), and the 5-year OS rate was 78.95% (105/133).
| Variable | Overall (n = 133) |
| Age, years | 15 (4) |
| Gender | |
| Male | 72 (54.1) |
| Female | 61 (45.9) |
| Location | |
| Axial | 21 (15.8) |
| Pelvis | -14 (10.5) |
| Spine | -7 (5.3) |
| Limb | 112 (84.2) |
| Distal femur | -52 (39.1) |
| Proximal tibia | -38 (28.6) |
| Proximal humerus | -14 (10.5) |
| Others | -8 (6.0) |
| Maxium tumor length, cm | 10.0 (7.0) |
| BMI, kg/m2 | 18.36 (4.59) |
| Pre-chemotherapy indexes | |
| VATI, cm2/m2 | 10.13 (11.80) |
| SATI, cm2/m2 | 16.57 (24.05) |
| SMI, cm2/m2 | 34.20 (12.75) |
| SMD, HU | 45.66 (7.72) |
| Albumin, g/L | 43.3 (5.4) |
| Platelet count, 109/L | 238 (141) |
| Neutrophil count, 109/L | 4.07 (3.20) |
| Lymphocyte count, 109/L | 1.57 (0.91) |
| ALP, g/L | 155 (153) |
| SII | 642.6 (689.6) |
| PNI | 51.15 (5.48) |
| Post-chemotherapy indexes | |
| VATI, cm2/m2 | 10.41 (14.15) |
| SATI, cm2/m2 | 17.03 (25.44) |
| SMI, cm2/m2 | 34.05 (11.72) |
| SMD, HU | 44.75 (8.06) |
| Albumin, g/L | 39.9 (7.7) |
| Platelet count, 109/L | 227 (145) |
| Neutrophil count, 109/L | 5.96 (5.13) |
| Lymphocyte count, 109/L | 1.33 (0.90) |
| ALP, g/L | 125 (123) |
| SII | 1240.1 (1341.0) |
| PNI | 46.03 (10.55) |
| ΔVATI, cm2/m2/30 days | 1.29 (5.57) |
| 1.27 (7.94) | |
| ΔSMI, cm2/m2/30 days | 0.68 (3.85) |
| ΔSMD, HU/30 days | 1.01 (3.36) |
| ΔSII, score/30 days | 754.1 (1416.4) |
| ΔPNI, score/30 days | 6.38 (9.16) |
| Histopathological response to neoadjuvant chemotherapy | |
| Non-responder | 20 (15.0) |
| Responder | 113 (85.0) |
Variables including baseline characteristics and CT-derived indexes were measured by univariable and multivariable logistic regression analysis for classification of responders and non-responders to neoadjuvant chemotherapy. Four factors, including age at diagnosis, tumor size, post-chemotherapy SII, and ΔSMD were selected with P < 0.05 in the univariable analysis (Table 2). In the multivariable regression model, age at diagnosis [hazard ratio (HR): 1.289, 95% confidence interval (CI): 1.023-1.625], post-chemotherapy SII [HR (95%CI): 0.999 (0.999-1.000)], and ΔSMD [HR (95%CI): 1.165 (1.028-1.321)] were identified as independent factors for predicting response to neoadjuvant chemotherapy. These predictors were combined to generated an integrated index with the formula of 0.254 × age - 0.001 × post-chemotherapy SII + 0.153 × ΔSMD - 5.993 (4).
As for OS, there were six variables, including pre-chemotherapy SMI, pre-chemotherapy alkaline phosphatase (ALP), pre-chemotherapy PNI, post-chemotherapy ALP, ΔSATI, and ΔSMD, selected in the univariate analysis with P < 0.10. The pre-chemotherapy SMI [HR (95%CI): 1.041 (1.003-1.081)], pre-chemotherapy PNI [HR (95%CI): 0.929 (0.872-0.989)], post-chemotherapy ALP [HR (95%CI): 1.003 (1.001-1.005)], and ΔSMD [HR (95%CI): 1.113 (1.035-1.197)] were considered as significant prognostic factors for OS in the multivariable Cox regression model with P < 0.05 (Table 3). Patients with lower scores of pre-chemotherapy PNI were found at risk compared with those with higher scores (P = 0.022). The reduction of SMD after neoadjuvant chemotherapy was associated with poor OS significantly (P = 0.004). No significant association was found between OS and other changes (Δ) of body composition (P > 0.05).
| Variable | Univariable analysis | Multivariable analysis | ||
| HR (95%CI) | P value | HR (95%CI) | P value | |
| Age | 1.094 (0.949-1.261) | 0.214 | ||
| Gender | 0.528 | |||
| Male | Reference | |||
| Female | 0.788 (0.375-1.654) | |||
| Location | 0.885 | |||
| Trunk | Reference | |||
| Limb | 0.931 (0.354-2.449) | |||
| Maxium tumor length | 1.020 (0.966-1.076) | 0.478 | ||
| BMI | 1.038 (0.948-1.137) | 0.421 | ||
| Pre-chemotherapy indexes | ||||
| VATI | 1.004 (0.974-1.035) | 0.783 | ||
| SATI | 1.000 (0.982-1.019) | 0.966 | ||
| SMI | 1.033 (0.999-1.069) | 0.057 | 1.041 (1.003-1.081) | 0.033 |
| SMD | 1.035 (0.976-1.097) | 0.250 | ||
| ALP | 1.001 (1.000-1.002) | 0.043 | 0.999 (0.997-1.001) | 0.172 |
| SII | 1.000 (1.000-1.001) | 0.138 | ||
| PNI | 0.922 (0.872-0.976) | 0.005 | 0.929 (0.872-0.989) | 0.022 |
| Post-chemotherapy indexes | ||||
| VATI | 1.000 (0.968-1.033) | 0.996 | ||
| SATI | 0.999 (0.980-1.017) | 0.874 | ||
| SMI | 1.018 (0.975-1.062) | 0.418 | ||
| SMD | 1.026 (0.962-1.094) | 0.441 | ||
| ALP | 1.002 (1.001-1.003) | 0.001 | 1.003 (1.001-1.005) | 0.007 |
| SII | 1.000 (1.000-1.000) | 0.385 | ||
| PNI | 1.001 (0.955-1.050) | 0.953 | ||
| ΔVATI | 1.005 (0.940-1.075) | 0.881 | ||
| ΔSATI | 1.036 (0.998-1.076) | 0.063 | 1.010 (0.970-1.051) | 0.634 |
| ΔSMI | 1.032 (0.979-1.088) | 0.240 | ||
| ΔSMD | 1.115 (1.042-1.193) | 0.002 | 1.113 (1.035-1.197) | 0.004 |
| ΔSII | 1.000 (1.000-1.001) | 0.181 | ||
| ΔPNI | 0.970 (0.914-1.029) | 0.313 | ||
Confounder-adjusted survival curves of independent predictors for OS were calculated based on the multivariable Cox regression model (Figure 2, Table 4). For internal validation, the optimism-corrected C-index of the multivariable Cox model was 0.824 with 95%CI of 0.768-0.880. The C-index using RSF was 0.820 (95%CI: 0.752-0.902), which did not materially improved discrimination. According to the results of sensitivity analyses, excluding patients with > 120-day interval between scans did not significantly change ΔSMD association with HR of 1.258 (95%CI: 0.872-1.689, P = 0.325).
| Variable | Cut-off point | Group | Median survival time (95%CI) | Log rank test |
| Pre-chemotherapy SMI | 48.86 cm2/m2 | Low | 88.297 (77.206-99.388) | 0.066 |
| High | 64.739 (43.620-85.858) | |||
| Pre-chemotherapy PNI | 50.25 | Low | 72.654 (57.202-88.107) | 0.002 |
| High | 82.702 (73.927-91.478) | |||
| Post-chemotherapy ALP | 380 g/L | Low | 89.401 (78.820-99.981) | 0.012 |
| High | 44.397 (28.647-60.148) | |||
| ΔSMD | 6.27 HU/30 days | Low | 94.657 (86.277-103.037) | 0.001 |
| High | 43.356 (29.130-57.581) | |||
| Integrated index | -1.09 | Low | 94.319 (85.826-102.813) | 0.010 |
| High | 46.988 (33.710-60.267) |
Using X-tile software, an optimal cut-off value of 6.27 Hounsfield unit/30 days was measured for ΔSMD (Figure 3). The OS was significantly shorter in patients with high ΔSMD (median: 43.356 months, 95%CI: 29.130-57.581 months) vs low ΔSMD (median: 94.657 months, 95%CI: 86.277-103.037) (P = 0.001). As for the integrated index from treatment response, all patients were dichotomized by the optimal cut-off point of -1.09. Patients with higher scores (median: 46.988 months, 95%CI: 33.710-60.267 months) were associated with shorter OS compared with those with lower scores (median: 94.319 months, 95%CI: 85.826-102.813) (P = 0.010).
In this study, we demonstrated that a greater reduction of ΔSMD during the neoadjuvant chemotherapy was associated with both poor treatment response and short OS for pediatric patients with osteosarcoma. The integrated index generated for discrimination of response to neoadjuvant chemotherapy was found significantly prognostic value to predict OS in pediatric patients with osteosarcoma.
To our knowledge, this was the first study to assess and evaluate the direct correlation of body composition-associated parameters, both before and after neoadjuvant chemotherapy, with response to neoadjuvant chemotherapy and long-term survival in young patients with osteosarcoma. In the present study, we considered that a poor chemotherapy response had significant relevance with decreased survival rates. Additionally, the treatment response might attribute to the selection of the postoperative regimens. Our main results showed that age at diagnosis (P = 0.031), post-chemotherapy SII (P = 0.032), and ΔSMD (P = 0.017) were significantly associated with therapeutic response in multivariable analysis. Moreover, we combined these variables to generate a novel integrated index to predict OS. On the survival analysis of potential predictors for OS, we demonstrated the prognostication of the pre-chemotherapy PNI (P = 0.002), post-chemotherapy ALP (P = 0.012), ΔSMD (P = 0.001), and the integrated index (P = 0.010). Our results provided supporting evidence about the relevance between response to neoadjuvant chemotherapy and long-term survival in pediatric patients. By applying internal validation and sensitivity analyses, we confirmed that ΔSMD and prechemotherapy SMI remained independent prognostic signals after correction for optimism. However, our study remained retrospective single center design that a multiinstitutional collaboration should be proposed to validate the integrated index with consideration of federated learning approaches to combine data across centers while protecting patient privacy.
In this study, nutritional indicators on blood tests were found associated with treatment response or OS. Cancer patients with high SII are often observed with thrombocytosis, neutrophilia, or lymphopenia. Our results suggested that patients with high SII on post-chemotherapy examinations preferred to obtain poor response to neoadjuvant che
Automated segmentation of axial CT images at the LV3 level has been applied as an established reference method for body composition analysis, which is a useful indicator for prediction of prognosis in malignancy[25]. Both sarcopenia and sarcopenic obesity are risk factors in association with unfavorable outcomes in malignancies after operation procedures[26,27]. However, the influence of body composition in pediatric osteosarcoma has not been explored before. The area and density of abdominal skeletal muscle are closely connected with individuals’ activity level and physical fitness[28,29]. In present study, patients with worse OS had a significantly higher ΔSMD compared with those with better OS. The large value of ΔSMD indicated a great reduction of physical activity and fitness, which resulted in poor survival outcome. The pre-chemotherapy SMI trended to be low in patients with poor OS. On our main results, the multivariate regression analysis identified ΔSMD and pre-chemotherapy SMI, two CT-derived parameters of muscle related body composition, as significant prognostic factors of overall death.
Body composition analysis on axial LV3 level performs objective assessment of physical activity in patients receiving abdominal scanning[30,31]. This is the first research to apply AI-based automated quantification and qualification of L3 body composition in patients with pediatric osteosarcoma who received neoadjuvant chemotherapy. The pre-chemotherapy, post-chemotherapy and change of body composition are evaluated comprehensively to instruct further lifestyle and recommend appropriate treatment. Limitations include retrospective singlecenter design, heterogeneity in CT contrast and scanner settings that could affect SMD, limited sample size for machine-learning modeling, and lack of external validation. Additionally, there is a selection bias that patients with osteosarcoma ≤ 18 years old are less likely to receive standardized therapeutic regimen compared with adult patients. Finally, the automated calculation of SMD on CT images may be interfered by the influence of contrast medium uptake. The cohort in present study had a relatively high 3- and 5-year OS (81.2% and 78.9%, respectively), which might may reflect selection bias inherent to our retrospective single-center dataset, the exclusion of patients who died during neoadjuvant chemotherapy or prior to surgery, variability in treatment regimens, and referral patterns. These factors might limit generalizability and underscore the need for validation in larger multicenter cohorts.
In conclusion, AI-based analysis of LV3 body composition on CT images is feasible in pediatric patients receiving neoadjuvant chemotherapy. ΔSMD is a prognostic indicator of physical activity and training that a great value significantly predicts the poor response to neoadjuvant chemotherapy and short OS. As a promising prognostic biomarker, ΔSMD needs further work to externally validate and transform these findings into a robust predictive tool for clinical decisionmaking. Using CT-derived and other nutritional predictors may conduct optimized nutrition support and appropriate selection of individualized therapeutic regimen.
| 1. | Smith MA, Altekruse SF, Adamson PC, Reaman GH, Seibel NL. Declining childhood and adolescent cancer mortality. Cancer. 2014;120:2497-2506. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 452] [Cited by in RCA: 385] [Article Influence: 32.1] [Reference Citation Analysis (0)] |
| 2. | Marina N, Gebhardt M, Teot L, Gorlick R. Biology and therapeutic advances for pediatric osteosarcoma. Oncologist. 2004;9:422-441. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 475] [Cited by in RCA: 488] [Article Influence: 22.2] [Reference Citation Analysis (0)] |
| 3. | Zhu W, Zhu L, Bao Y, Zhong X, Chen Y, Wu Q. Clinical evaluation of neoadjuvant chemotherapy for osteosarcoma. J BUON. 2019;24:1181-1185. [PubMed] |
| 4. | Duchman KR, Gao Y, Miller BJ. Prognostic factors for survival in patients with high-grade osteosarcoma using the Surveillance, Epidemiology, and End Results (SEER) Program database. Cancer Epidemiol. 2015;39:593-599. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 129] [Cited by in RCA: 181] [Article Influence: 16.5] [Reference Citation Analysis (0)] |
| 5. | Mu X, Agarwal R, March D, Rothenberg A, Voigt C, Tebbets J, Huard J, Weiss K. Notch Signaling Mediates Skeletal Muscle Atrophy in Cancer Cachexia Caused by Osteosarcoma. Sarcoma. 2016;2016:3758162. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 22] [Cited by in RCA: 33] [Article Influence: 3.3] [Reference Citation Analysis (0)] |
| 6. | Mitsunaga S, Kasamatsu E, Machii K. Incidence and frequency of cancer cachexia during chemotherapy for advanced pancreatic ductal adenocarcinoma. Support Care Cancer. 2020;28:5271-5279. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 25] [Cited by in RCA: 43] [Article Influence: 7.2] [Reference Citation Analysis (0)] |
| 7. | Chan MY, Chok KSH. Sarcopenia in pancreatic cancer - effects on surgical outcomes and chemotherapy. World J Gastrointest Oncol. 2019;11:527-537. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in CrossRef: 66] [Cited by in RCA: 69] [Article Influence: 9.9] [Reference Citation Analysis (0)] |
| 8. | Vanderwall C, Randall Clark R, Eickhoff J, Carrel AL. BMI is a poor predictor of adiposity in young overweight and obese children. BMC Pediatr. 2017;17:135. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 80] [Cited by in RCA: 162] [Article Influence: 18.0] [Reference Citation Analysis (0)] |
| 9. | Fosbøl MØ, Zerahn B. Contemporary methods of body composition measurement. Clin Physiol Funct Imaging. 2015;35:81-97. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 101] [Cited by in RCA: 181] [Article Influence: 15.1] [Reference Citation Analysis (0)] |
| 10. | Beaudart C, McCloskey E, Bruyère O, Cesari M, Rolland Y, Rizzoli R, Araujo de Carvalho I, Amuthavalli Thiyagarajan J, Bautmans I, Bertière MC, Brandi ML, Al-Daghri NM, Burlet N, Cavalier E, Cerreta F, Cherubini A, Fielding R, Gielen E, Landi F, Petermans J, Reginster JY, Visser M, Kanis J, Cooper C. Sarcopenia in daily practice: assessment and management. BMC Geriatr. 2016;16:170. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 689] [Cited by in RCA: 596] [Article Influence: 59.6] [Reference Citation Analysis (0)] |
| 11. | Kays JK, Shahda S, Stanley M, Bell TM, O'Neill BH, Kohli MD, Couch ME, Koniaris LG, Zimmers TA. Three cachexia phenotypes and the impact of fat-only loss on survival in FOLFIRINOX therapy for pancreatic cancer. J Cachexia Sarcopenia Muscle. 2018;9:673-684. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 140] [Cited by in RCA: 128] [Article Influence: 16.0] [Reference Citation Analysis (0)] |
| 12. | Griffin OM, Duggan SN, Ryan R, McDermott R, Geoghegan J, Conlon KC. Characterising the impact of body composition change during neoadjuvant chemotherapy for pancreatic cancer. Pancreatology. 2019;19:850-857. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 31] [Cited by in RCA: 61] [Article Influence: 8.7] [Reference Citation Analysis (1)] |
| 13. | Yang S, Wu C, Wang L, Shan D, Chen B. Pretreatment inflammatory indexes as prognostic predictors for survival in osteosarcoma patients. Int J Clin Exp Pathol. 2020;13:515-524. [PubMed] |
| 14. | Huang X, Hu H, Zhang W, Shao Z. Prognostic value of prognostic nutritional index and systemic immune-inflammation index in patients with osteosarcoma. J Cell Physiol. 2019;234:18408-18414. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 23] [Cited by in RCA: 49] [Article Influence: 7.0] [Reference Citation Analysis (0)] |
| 15. | Guenther LM, Rowe RG, Acharya PT, Swenson DW, Meyer SC, Clinton CM, Guo D, Sridharan M, London WB, Grier HE, Ecklund K, Janeway KA. Response Evaluation Criteria in Solid Tumors (RECIST) following neoadjuvant chemotherapy in osteosarcoma. Pediatr Blood Cancer. 2018;65. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 30] [Cited by in RCA: 56] [Article Influence: 7.0] [Reference Citation Analysis (0)] |
| 16. | Ronneberger O, Fischer P, Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation. In: Navab N, Hornegger J, Wells W, Frangi A. Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015. Cham: Springer, 2015. [RCA] [DOI] [Full Text] [Cited by in Crossref: 13000] [Cited by in RCA: 11480] [Article Influence: 1043.6] [Reference Citation Analysis (1)] |
| 17. | Camp RL, Dolled-Filhart M, Rimm DL. X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Clin Cancer Res. 2004;10:7252-7259. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 3094] [Cited by in RCA: 3152] [Article Influence: 143.3] [Reference Citation Analysis (3)] |
| 18. | Jain S, Harris J, Ware J. Platelets: linking hemostasis and cancer. Arterioscler Thromb Vasc Biol. 2010;30:2362-2367. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 199] [Cited by in RCA: 190] [Article Influence: 11.9] [Reference Citation Analysis (0)] |
| 19. | Wojtukiewicz MZ, Sierko E, Hempel D, Tucker SC, Honn KV. Platelets and cancer angiogenesis nexus. Cancer Metastasis Rev. 2017;36:249-262. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 210] [Cited by in RCA: 196] [Article Influence: 21.8] [Reference Citation Analysis (0)] |
| 20. | Mantovani A, Cassatella MA, Costantini C, Jaillon S. Neutrophils in the activation and regulation of innate and adaptive immunity. Nat Rev Immunol. 2011;11:519-531. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 2271] [Cited by in RCA: 2266] [Article Influence: 151.1] [Reference Citation Analysis (7)] |
| 21. | De Giorgi U, Mego M, Scarpi E, Giuliano M, Giordano A, Reuben JM, Valero V, Ueno NT, Hortobagyi GN, Cristofanilli M. Relationship between lymphocytopenia and circulating tumor cells as prognostic factors for overall survival in metastatic breast cancer. Clin Breast Cancer. 2012;12:264-269. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 84] [Cited by in RCA: 90] [Article Influence: 6.4] [Reference Citation Analysis (0)] |
| 22. | Yao ZH, Tian GY, Wan YY, Kang YM, Guo HS, Liu QH, Lin DJ. Prognostic nutritional index predicts outcomes of malignant pleural mesothelioma. J Cancer Res Clin Oncol. 2013;139:2117-2123. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 63] [Cited by in RCA: 65] [Article Influence: 5.0] [Reference Citation Analysis (0)] |
| 23. | Gao Y, Zhou S, Jiang W, Huang M, Dai X. Effects of ganopoly (a Ganoderma lucidum polysaccharide extract) on the immune functions in advanced-stage cancer patients. Immunol Invest. 2003;32:201-215. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 237] [Cited by in RCA: 183] [Article Influence: 8.0] [Reference Citation Analysis (3)] |
| 24. | Tenardi RD, Frühwald MC, Jürgens H, Hertroijs D, Bauer J. Nutritional status of children and young adults with Ewing sarcoma or osteosarcoma at diagnosis and during multimodality therapy. Pediatr Blood Cancer. 2012;59:621-626. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 13] [Cited by in RCA: 21] [Article Influence: 1.5] [Reference Citation Analysis (0)] |
| 25. | Cespedes Feliciano EM, Popuri K, Cobzas D, Baracos VE, Beg MF, Khan AD, Ma C, Chow V, Prado CM, Xiao J, Liu V, Chen WY, Meyerhardt J, Albers KB, Caan BJ. Evaluation of automated computed tomography segmentation to assess body composition and mortality associations in cancer patients. J Cachexia Sarcopenia Muscle. 2020;11:1258-1269. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 122] [Cited by in RCA: 120] [Article Influence: 20.0] [Reference Citation Analysis (0)] |
| 26. | Britton KA, Massaro JM, Murabito JM, Kreger BE, Hoffmann U, Fox CS. Body fat distribution, incident cardiovascular disease, cancer, and all-cause mortality. J Am Coll Cardiol. 2013;62:921-925. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 557] [Cited by in RCA: 520] [Article Influence: 40.0] [Reference Citation Analysis (4)] |
| 27. | Weerink LBM, van der Hoorn A, van Leeuwen BL, de Bock GH. Low skeletal muscle mass and postoperative morbidity in surgical oncology: a systematic review and meta-analysis. J Cachexia Sarcopenia Muscle. 2020;11:636-649. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 63] [Cited by in RCA: 96] [Article Influence: 16.0] [Reference Citation Analysis (0)] |
| 28. | Hsu KJ, Liao CD, Tsai MW, Chen CN. Effects of Exercise and Nutritional Intervention on Body Composition, Metabolic Health, and Physical Performance in Adults with Sarcopenic Obesity: A Meta-Analysis. Nutrients. 2019;11:2163. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 136] [Cited by in RCA: 194] [Article Influence: 27.7] [Reference Citation Analysis (0)] |
| 29. | Evans WJ, Morley JE, Argilés J, Bales C, Baracos V, Guttridge D, Jatoi A, Kalantar-Zadeh K, Lochs H, Mantovani G, Marks D, Mitch WE, Muscaritoli M, Najand A, Ponikowski P, Rossi Fanelli F, Schambelan M, Schols A, Schuster M, Thomas D, Wolfe R, Anker SD. Cachexia: a new definition. Clin Nutr. 2008;27:793-799. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 1976] [Cited by in RCA: 1688] [Article Influence: 93.8] [Reference Citation Analysis (1)] |
| 30. | Mourtzakis M, Prado CM, Lieffers JR, Reiman T, McCargar LJ, Baracos VE. A practical and precise approach to quantification of body composition in cancer patients using computed tomography images acquired during routine care. Appl Physiol Nutr Metab. 2008;33:997-1006. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 1879] [Cited by in RCA: 1794] [Article Influence: 99.7] [Reference Citation Analysis (4)] |
| 31. | Fearon K, Strasser F, Anker SD, Bosaeus I, Bruera E, Fainsinger RL, Jatoi A, Loprinzi C, MacDonald N, Mantovani G, Davis M, Muscaritoli M, Ottery F, Radbruch L, Ravasco P, Walsh D, Wilcock A, Kaasa S, Baracos VE. Definition and classification of cancer cachexia: an international consensus. Lancet Oncol. 2011;12:489-495. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 4755] [Cited by in RCA: 4253] [Article Influence: 283.5] [Reference Citation Analysis (4)] |