Enhanced Predictive Analytics for Early Malignancy Discovery in Routine Screening

• Dr. Latha Kiran Krishna Rajendran

  Author

Optimizing the effectiveness of population-level cancer screening programs hinges on precise and timely identification of subtle anomalies. This paper introduces an advanced computational methodology, rooted in deep learning principles, to revolutionize the interpretation of prophylactic radiographic examinations. Our system employs a two-stage deep learning framework that combines patch-based convolutional neural networks with full-image classification, autonomously detecting potential malignant indicators while streamlining the diagnostic workflow. Evaluated on two public mammography datasets (CBIS-DDSM and INbreast), the proposed ResNet-based architecture with multi-patch sampling (S10) achieves an area under the ROC curve (AUC) of 95.0\% on CBIS-DDSM and generalizes to INbreast with an AUC of 91.7\%, representing a substantial enhancement over single-patch baselines. This AI-driven approach marks a significant stride towards more accurate and scalable early disease detection within comprehensive cancer screening initiatives.

• PDF

Copyright (c) 2026 International Journal of Clinical Research and Medical Sciences

This work is licensed under a Creative Commons Attribution 4.0 International License.