Harnessing Latent Attributes: A Data-Centric Approach to Feature Extraction and Selection in Lung Cancer Analysis

DOI

10.2991/978-94-6239-756-9_14How to use a DOI?

Keywords

Lung Cancer; Computer-aided Diagnosis (CAD); Feature Engineering; Hyperparameter Optimization; RobustScore; Distributed Training; Cross-Validation

Abstract

Lung cancer is the leading cancer killer and as computer-assisted systems enter the clinic, it requires the creation of powerful systems across the scanners, locations and workflows. The study examines the feature engineering strategies of CT/CXR, pathology, and EMR/omics and summarize them in a manner mindful of stability and consistent with data characteristics-Harnessing Latent Attributes (HLA). The study obtains the result of the literature analysis and comparing the datasets presented in the several datasets that proper basic dataset design (stratified splits, class balance, cross-scanner normalization, targeted augmentation) is one of the most significant elements affecting the generalization. Several key concepts in feature learning are (a) to discover low dimensional subspace or feature maps that do not know anything about the actual values of the variables and (b) the advantages of the feature selection are sparsity, bottleneck, attention mechanism, and posterior stability. Given the proximity of the data, intrinsic hyperparameter optimization of federated/distributed training sacrifices privacy to gain higher accuracy. No external evidence can be better than calibration, Jaccard overlap, empirical external validation and temporary empirical validation (i.e., in cross-validation). The study states our problem as follows: On the contrary, end-to-end deep/federated models are commonly trained to attain ideal image performance and systematic choice and interpretable multi-view fusion can produce reproducible and interpretable output. Finally, under institutional constraints, the study offers model-choice (RobustScore) solutions that take into account accuracy/stability, without compromising the accuracy of the calibration and robustness to shift: these solutions represent but a step toward robust prediction and detection.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Sneha Satpute
AU  - Rupali Rajendra Gangarde
PY  - 2026
DA  - 2026/08/31
TI  - Harnessing Latent Attributes: A Data-Centric Approach to Feature Extraction and Selection in Lung Cancer Analysis
BT  - Proceedings of the Conference on Bioengineering for Global Health (SYMRESEARCH 2.0 2025)
PB  - Atlantis Press
SP  - 183
EP  - 193
SN  - 2468-5747
UR  - https://doi.org/10.2991/978-94-6239-756-9_14
DO  - 10.2991/978-94-6239-756-9_14
ID  - Satpute2026
ER  -