Deep learning-aided 3D proxy-bridged region-growing framework for multi-organ segmentation.

Accurate multi-organ segmentation in 3D CT images is imperative for enhancing computer-aided diagnosis and radiotherapy planning. However, current deep learning-based methods for 3D multi-organ segmentation face challenges such as the need for labor-intensive manual pixel-level annotations and high hardware resource demands, especially regarding GPU resources. To address these issues, we propose a 3D proxy-bridged region-growing framework specifically designed for the segmentation of the liver and spleen. Specifically, a key slice is selected from each 3D volume according to the corresponding intensity histogram. Subsequently, a deep learning model is employed to pinpoint the semantic central patch on this key slice, to calculate the growing seed. To counteract the impact of noise, segmentation of the liver and spleen is conducted on superpixel images created through proxy-bridging strategy. The segmentation process is then extended to adjacent slices by applying the same methodology iteratively, culminating in the comprehensive segmentation results. Experimental results demonstrate that the proposed framework accomplishes segmentation of the liver and spleen with an average Dice Similarity Coefficient of approximately 0.93 and a Jaccard Similarity Coefficient of around 0.88. These outcomes substantiate the framework's capability to achieve performance on par with that of deep learning methods, albeit requiring less guidance information and lower GPU resources.

Exploration of a noninvasive radiomics classifier for breast cancer tumor microenvironment categorization and prognostic outcome prediction.

RATIONALE AND OBJECTIVES: Breast cancer progression and treatment response are significantly influenced by the tumor microenvironment (TME). Traditional methods for assessing TME are invasive, posing a challenge for patient care. This study introduces a non-invasive approach to TME classification by integrating radiomics and machine learning, aiming to predict the TME status using imaging data, thereby aiding in prognostic outcome prediction. MATERIALS AND METHODS: Utilizing multi-omics data from The Cancer Genome Atlas (TCGA) and The Cancer Imaging Archive (TCIA), this study employed CIBERSORT and MCP-counter algorithms analyze immune infiltration in breast cancer. A radiomics classifier was developed using a random forest algorithm, leveraging quantitative features extracted from intratumoral and peritumoral regions of Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) scans. The classifer's ability to predict diverse TME states were and their prognostic implications were evaluated using Kaplan-Meier survival curves. RESULTS: Three distinct TME states were identified using RNA-Seq data, each displaying unique prognostic and biological characteristics. Notably, patients with increased immune cell infiltration showed significantly improved prognoses (P < 0.05). The classifier, comprising 24 radiomic features, demonstrated high predictive accuracy (AUC of training set = 0.960, 95 % CI: 0.922, 0.997; AUC of testing set = 0.853, 95 % CI: 0.687, 1.000) in differentiating these TME states. Predictions from the classifier also correlated significantly with overall patient survival (P < 0.05). CONCLUSION: This study offers a detailed analysis of the complex TME states in breast cancer and presents a reliable, noninvasive radiomics classifier for TME assessment. The classifer's accurate prediction of TME status and its correlation with prognosis highlight its potential as a tool in personalized breast cancer treatment, paving the way for more individualized and less invasive therapeutic strategies.