Pankaj Kumar

Pankaj Kumar

Auto-generated profile, based on publicly available information

Pankaj Kumar

School of Computer Science, University of Petroleum and Energy Studies, Dehradun, Uttarakhand 248007, India

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 January 2024 in International Journal of Biomedical Imaging
Verified
Journal Article
Open access
UOUrvi Oza
Urvi Oza
BGBakul Gohel
Bakul Gohel
PKPankaj Kumar
Pankaj Kumar
POParita Oza
Parita Oza
Accurate segmentation of breast masses in mammogram images is essential for early cancer diagnosis and treatment planning. Several deep learning (DL) models have been proposed for whole mammogram segmentation and mass patch/crop segmentation. However, current DL models for breast mammogram mass segmentation face several limitations, including false positives (FPs), false negatives (FNs), and challenges with the end‐to‐end approach. This paper presents a novel two‐stage end‐to‐end cascaded breast mass segmentation framework that incorporates a saliency map of potential mass regions to guide the DL models for breast mass segmentation. The first‐stage segmentation model of the cascade framework is used to generate a saliency map to establish a coarse region of interest (ROI), effectively narrowing the focus to probable mass regions. The proposed presegmenter attention (PSA) blocks are introduced in the second‐stage segmentation model to enable dynamic adaptation to the most informative regions within the mammogram images based on the generated saliency map. Comparative analysis of the Attention U‐net model with and without the cascade framework is provided in terms of dice scores, precision, recall, FP rates (FPRs), and FN outcomes. Experimental results consistently demonstrate enhanced breast mass segmentation performance by the proposed cascade framework across all three datasets: INbreast, CSAW‐S, and DMID. The cascade framework shows superior segmentation performance by improving the dice score by about 6% for the INbreast dataset, 3% for the CSAW‐S dataset, and 2% for the DMID dataset. Similarly, the FN outcomes were reduced by 10% for the INbreast dataset, 19% for the CSAW‐S dataset, and 4% for the DMID dataset. Moreover, the proposed cascade framework’s performance is validated with varying state‐of‐the‐art segmentation models such as DeepLabV3+ and Swin transformer U‐net. The presegmenter cascade framework has the potential to improve segmentation performance and mitigate FNs when integrated with any medical image segmentation framework, irrespective of the choice of the model.

3 December 2022
Verified
Book Chapter
Closed access
PDF file
SDSuyash Dhondkar
Suyash Dhondkar
MKManish Khare
Manish Khare
PKPankaj Kumar
Pankaj Kumar


6 October 2025 in Biomedical Engineering: Applications, Basis and Communications
Verified
Journal Article
Closed access
UOUrvi Oza
Urvi Oza
BGBakul Gohel
Bakul Gohel
PKPankaj Kumar
Pankaj Kumar
Breast mammogram mass detection and segmentation play a critical role in early breast cancer diagnosis, with deep learning (DL)-based models becoming increasingly prominent in computer-assisted screening. However, most existing systems are trained exclusively on abnormal mammograms, limiting their clinical applicability, as real-world screenings also include normal cases. To address this limitation, we propose a comparative analysis of DL-based segmentation models within an end-to-end diagnostic framework that incorporates both normal and abnormal mammograms. We considered INbreast and DMID datasets for our study. We begin with a preliminary analysis of two end-to-end approaches: one involving initial classification of normal and abnormal mammograms, followed by mass segmentation, and another relying solely on segmentation with thresholding for classification. Various segmentation models, including patch-based and whole mammogram models, were trained using popular DL architectures such as Unet, DeepLabV3[ ], and YOLOv8. Both pixel-level (e.g. Dice score, recall) and lesion-level (false/true positive/negative rates) metrics were used for comprehensive evaluation. To enhance performance, we propose a fusion framework combining YOLO outputs and whole mammogram segmentation models for mass localization, followed by patch-based segmentation for refinement. Our findings highlight the value of including normal mammograms during training and demonstrate that lesion-level evaluation provides a more clinically relevant assessment than pixel-level metrics alone. Additionally, we show that breast density and tumor size significantly impact segmentation outcomes, emphasizing the need for robust evaluation. Within the scope of our comparative study, which includes both normal and abnormal mammograms, the proposed end-to-end fusion framework demonstrates improved detection sensitivity and segmentation accuracy over individual models and frameworks.