Sparse Dynamic Deep Neural Network in Marginal Shape Space for Accurate COVID-19 Lung Tissue Segmentation from Chest CT Images
Bibliographic record
Abstract
In medical diagnosis systems, predicting and quantifying COVID-19 lung abnormalities from chest Computed Tomography (CT) images is essential for early identification of infected lesions and accurate diagnosis.Visual assessment and quantification of COVID-19 lung tissues by expert radiologists can be costly and error-prone.Consequently, numerous deep learning (DL)-based segmentation models have been developed for the automatic segmentation and prediction of infected lung tissues.Among these models, the Multi-Scale Attention-based UNet (MS-AUNet) can extract complex geometric features from CT images and segment small boundary areas infected by COVID-19.However, it may introduce errors by misclassifying normal tissues that resemble infected tissues.To address this issue, this study proposes a Marginal Space Deep Learning (MSDL) model in conjunction with the MS-AUNet to accurately segment normal and COVID-19-infected tissues from chest CT images.Initially, the MS-AUNet is applied to obtain Region-Of-Interests (ROIs) from the chest CT images.Subsequently, these ROIs are refined using the MSDL model, which consists of a Sparse Dynamic Deep Neural Network (SDeepNet) and an Active Shape Model (ASM) for non-rigid tissue segmentation of COVID-19 CT images.The SDeepNet acts as a boundary detector, automatically learning dynamic sparse features from the given ROI in each marginal shape space and detecting bounding boxes to localize target tissues.The ASM is employed to learn shape deformation and accurately segment infected lung tissues.Experimental results demonstrate that the MS-AUNet-MSDL model using a CT image dataset achieves 89.7% dice score, 88% recall, 89.4% precision, 11.42mm Hausdorff distance, and 22.8% Root Mean Square Error (RMSE).
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".