Background:

Diffuse Large B-cell Lymphoma (DLBCL), the most prevalent aggressive subtype of Non-Hodgkin Lymphoma, constitutes approximately 35% of all lymphomas. While standardized treatments and novel therapeutics have improved DLBCL survival, prognostic assessment remains suboptimal. Body composition parameters are key indicators of nutritional status in cancer patients, with established prognostic value. Recent advances in Artificial Intelligence applied to medical imaging have significantly enhanced diagnostic accuracy and efficiency by mitigating subjectivity and technical variability. This study utilizes an AI deep learning model to precisely segment chest CT images from newly diagnosed DLBCL patients, extracting body composition parameters (subcutaneous fat, pectoralis major, pectoralis minor, mediastinal fat). The prognostic significance of these parameters will be examined to inform personalized DLBCL treatment strategies.

Purpose:

Utilizing a deep learning model, this study performs automated segmentation of chest CT images from newly diagnosed DLBCL patients to derive body composition parameters at the T4 vertebral level and subsequently examines their correlation with clinical characteristics.

Methods:

A retrospective study was conducted on patients definitively diagnosed with DLBCL at the Department of Oncology Hematology, China-Japan Union Hospital of Jilin University between January 2018 and April 2025. Patient demographics (age, sex, height, weight, BMI, ECOG performance status) and clinical characteristics (Ann Arbor stage, B symptoms, bone marrow involvement, molecular subtype, albumin, maximum standardized uptake value [SUVmax]) were collected. Body composition components (subcutaneous fat, pectoralis major, pectoralis minor, mediastinal fat) at the T4 vertebral level were segmented from the initial chest CT scans of over a hundred treatment-compliant DLBCL patients using a deep learning model and quantified using ImageJ. Associations between body composition parameters and clinical characteristics were analyzed using SPSS 27.0. Survival analysis employed Kaplan-Meier curves with Log-rank testing, and prognostic factors were assessed via univariate and multivariate Cox proportional hazards regression.

Results:

  • BMI < 24 kg/m² was significantly associated with reduced pectoralis minor muscle area (HR = 11.904, 95% CI 1.579-89.724, P = 0.016), reduced subcutaneous fat area (HR = 4.194, 95% CI 1.244-14.131, P = 0.021), and reduced total fat area (HR = 0.294, 95% CI 0.117-0.738, P = 0.009), compared to BMI ≥ 24 kg/m².

  • Patients with low pectoralis minor muscle area exhibited significantly shorter overall survival (OS) than those with high pectoralis minor area (HR = 0.310, 95% CI 0.125-0.769, P = 0.012). Patients aged >60 years had significantly shorter OS than those ≤60 years (HR = 0.252, 95% CI 0.087-0.731, P = 0.011). Patients with Ann Arbor stage III/IV disease had significantly shorter OS than those with stage I/II disease (HR = 3.689, 95% CI 1.395-9.757, P = 0.009).

  • Multivariate analysis identified low pectoralis minor muscle area as an independent predictor of shorter survival time (HR = 0.269, 95% CI 0.078-0.891, P = 0.032). Ann Arbor stage III/IV was also an independent predictor of shorter survival time (HR = 4.782, 95% CI 1.155-17.390, P = 0.030).

  • Kaplan-Meier analysis revealed significantly different OS between DLBCL patient groups stratified by pectoralis minor area (Log-rank P = 0.007; n=88). Subgroup analysis showed no significant OS difference by pectoralis minor area among males (Log-rank P = 0.098), but a significant difference among females (Log-rank P = 0.043).Significantly different OS was also observed between patients with Ann Arbor stage I/II versus III/IV disease (Log-rank P = 0.005) and between younger (≤60 years) and older (>60 years) patients (Log-rank P = 0.006).

Conclusion:

  1. Body composition parameters measured at the T4 level on chest CT can serve as indicators for assessing nutritional status.

  2. Low pectoralis minor muscle area is associated with shorter overall survival and represents an independent risk factor for poor prognosis in DLBCL patients.

  3. BMI, subcutaneous fat area, mediastinal fat area, and pectoralis major muscle area showed no significant association with OS in DLBCL and lack predictive value for patient prognosis.

  4. Ann Arbor stage III-IV is associated with shorter overall survival and is an independent risk factor for poor prognosis in DLBCL patients.

This content is only available as a PDF.
2025
Sign in via your Institution