Santosh Sanjeev
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Santosh Sanjeev has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Ibrahim Almakky
Mohamed bin Zayed University of Artificial Intelligence
5 shared publications - Mohammad Yaqub
Mohamed bin Zayed University of Artificial Intelligence
5 shared publications - Anees Ur Rehman Hashmi
Mohamed bin Zayed University of Artificial Intelligence
3 shared publications - Fadillah Adamsyah Maani
Mohamed bin Zayed University of Artificial Intelligence
3 shared publications - Mohammad Areeb Qazi
Mohamed bin Zayed University of Artificial Intelligence
2 shared publications - Muhammad Haris Khan
Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, United Arab Emirates
2 shared publications
Latest Publications
- Top Cited PapersOpen Access
- 4 February 2026
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 58 (8) , 1-25
Continual Learning in Medical Imaging: A Survey and Practical Analysis
Deep Learning has shown great success in reshaping medical imaging, yet it faces several challenges hindering widespread application. Distribution shifts in the continuously evolving data stream and catastrophic forgetting are two barriers that considerably increase the performance gap between research and applications. Continual Learning offers promise in addressing these hurdles by enabling the sequential acquisition of new knowledge without forgetting previous information. In this survey, we comprehensively review the recent literature on continual learning in the medical domain, highlight recent trends, and point out several practical issues. Specifically, we survey the continual learning studies on classification, segmentation, detection, and other related tasks in the medical domain and develop a taxonomy for the reviewed studies. We also critically discuss the current state of continual learning in medical imaging, including identifying open problems and outlining promising future directions. We hope that this survey provides researchers with a useful overview of the developments in the field and further increases engagement with this topic within the community. To keep up with the fast-paced advancements in the field, we will routinely update the repository with the latest relevant articles at https://github.com/BioMedIA-MBZUAI/awesome-cl-in-medical .
- 4 February 2026
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 58 (8) , 1-25
Top Cited PapersOpen AccessContinual Learning in Medical Imaging: A Survey and Practical Analysis
Deep Learning has shown great success in reshaping medical imaging, yet it faces several challenges hindering widespread application. Distribution shifts in the continuously evolving data stream and catastrophic forgetting are two barriers that considerably increase the performance gap between research and applications. Continual Learning offers promise in addressing these hurdles by enabling the sequential acquisition of new knowledge without forgetting previous information. In this survey, we comprehensively review the recent literature on continual learning in the medical domain, highlight recent trends, and point out several practical issues. Specifically, we survey the continual learning studies on classification, segmentation, detection, and other related tasks in the medical domain and develop a taxonomy for the reviewed studies. We also critically discuss the current state of continual learning in medical imaging, including identifying open problems and outlining promising future directions. We hope that this survey provides researchers with a useful overview of the developments in the field and further increases engagement with this topic within the community. To keep up with the fast-paced advancements in the field, we will routinely update the repository with the latest relevant articles at https://github.com/BioMedIA-MBZUAI/awesome-cl-in-medical .
- Open Access
- 18 March 2025
- Published by Copernicus GmbH
Deep learning-based approach to Levelling Airborne Magnetic Data
Airborne magnetic surveys provide valuable insights into subsurface structures but often suffer from levelling errors due to inconsistencies between flight lines. These errors, such as striping patterns caused by sensor variations and magnetic field fluctuations, can obscure anomalies and distort interpretations. Traditional corrections like tie-line or micro levelling address these issues but rely on time and frequency domain analyses, making the process labor-intensive, costly, and reliant on expert intervention. Automating and enhancing these workflows is crucial for efficient and accurate levelling across large-scale airborne magnetic datasets. In this work, we propose a deep learning framework for levelling airborne magnetic data by leveraging a U-Net-based architecture. The model is trained in a supervised manner. We use a combination of perceptual loss and mean squared error (MSE) loss to capture fine-grained details while maintaining global consistency in the levelled data. Once trained, the proposed method demonstrates computational efficiency during inference, enabling automatic and robust levelling corrections for large datasets without requiring manual intervention or additional tie-line constraints. The model's performance was evaluated on an independent survey data from the Geological Survey of Brazil database, as well as on an out-of-distribution (OOD) dataset consisting of magnetic field data acquired by Geotech Limited, demonstrating its generalizability and robustness. Our approach demonstrates performance on par with traditional levelling methods, as validated by both quantitative and qualitative metrics, while introducing significant advantages in efficiency and automation. This deep learning-based solution simplifies the levelling process and provides a scalable, adaptive framework designed to meet the demands of modern geophysical surveys.
- 18 March 2025
- Published by Copernicus GmbH
Open AccessDeep learning-based approach to Levelling Airborne Magnetic Data
Airborne magnetic surveys provide valuable insights into subsurface structures but often suffer from levelling errors due to inconsistencies between flight lines. These errors, such as striping patterns caused by sensor variations and magnetic field fluctuations, can obscure anomalies and distort interpretations. Traditional corrections like tie-line or micro levelling address these issues but rely on time and frequency domain analyses, making the process labor-intensive, costly, and reliant on expert intervention. Automating and enhancing these workflows is crucial for efficient and accurate levelling across large-scale airborne magnetic datasets. In this work, we propose a deep learning framework for levelling airborne magnetic data by leveraging a U-Net-based architecture. The model is trained in a supervised manner. We use a combination of perceptual loss and mean squared error (MSE) loss to capture fine-grained details while maintaining global consistency in the levelled data. Once trained, the proposed method demonstrates computational efficiency during inference, enabling automatic and robust levelling corrections for large datasets without requiring manual intervention or additional tie-line constraints. The model's performance was evaluated on an independent survey data from the Geological Survey of Brazil database, as well as on an out-of-distribution (OOD) dataset consisting of magnetic field data acquired by Geotech Limited, demonstrating its generalizability and robustness. Our approach demonstrates performance on par with traditional levelling methods, as validated by both quantitative and qualitative metrics, while introducing significant advantages in efficiency and automation. This deep learning-based solution simplifies the levelling process and provides a scalable, adaptive framework designed to meet the demands of modern geophysical surveys.
- 23 October 2024
- Published by Springer Nature
- p. 131-141
FissionFusion: Fast Geometric Generation and Hierarchical Souping for Medical Image Analysis
- 23 October 2024
- Published by Springer Nature
- p. 131-141
FissionFusion: Fast Geometric Generation and Hierarchical Souping for Medical Image Analysis
- 9 October 2024
- Published by Springer Nature
- p. 169-179
TiBiX: Leveraging Temporal Information for Bidirectional X-Ray and Report Generation
- 9 October 2024
- Published by Springer Nature
- p. 169-179
TiBiX: Leveraging Temporal Information for Bidirectional X-Ray and Report Generation
- 24 July 2024
- Published by Springer Nature
- p. 245-257
DynaMMo: Dynamic Model Merging for Efficient Class Incremental Learning for Medical Images
- 24 July 2024
- Published by Springer Nature
- p. 245-257
DynaMMo: Dynamic Model Merging for Efficient Class Incremental Learning for Medical Images
- 15 October 2023
- Published by Springer Nature
- p. 434-443
PECon: Contrastive Pretraining to Enhance Feature Alignment Between CT and EHR Data for Improved Pulmonary Embolism Diagnosis
- 15 October 2023
- Published by Springer Nature
- p. 434-443
PECon: Contrastive Pretraining to Enhance Feature Alignment Between CT and EHR Data for Improved Pulmonary Embolism Diagnosis
- 15 October 2023
- Published by Springer Nature
- p. 444-453
Exploring the Transfer Learning Capabilities of CLIP in Domain Generalization for Diabetic Retinopathy
- 15 October 2023
- Published by Springer Nature
- p. 444-453
Exploring the Transfer Learning Capabilities of CLIP in Domain Generalization for Diabetic Retinopathy
- Open Access
Exploring the Transfer Learning Capabilities of CLIP in Domain Generalization for Diabetic Retinopathy
Diabetic Retinopathy (DR), a leading cause of vision impairment, requires early detection and treatment. Developing robust AI models for DR classification holds substantial potential, but a key challenge is ensuring their generalization in unfamiliar domains with varying data distributions. To address this, our paper investigates cross-domain generalization, also known as domain generalization (DG), within the context of DR classification. DG, a challenging problem in the medical domain, is complicated by the difficulty of gathering labeled data across different domains, such as patient demographics and disease stages. Some recent studies have shown the effectiveness of using CLIP to handle the DG problem in natural images. In this study, we investigate CLIP's transfer learning capabilities and its potential for cross-domain generalization in diabetic retinopathy (DR) classification. We carry out comprehensive experiments to assess the efficacy and potential of CLIP in addressing DG for DR classification. Further, we introduce a multi-modal fine-tuning strategy named Context Optimization with Learnable Visual Tokens (CoOpLVT), which enhances context optimization by conditioning on visual features. Our findings demonstrate that the proposed method increases the F1-score by 1.8% over the baseline, thus underlining its promise for effective DG in DR classification. Our code is publicly available at https://github.com/Sanoojan/CLIP-DRDG.
Open AccessExploring the Transfer Learning Capabilities of CLIP in Domain Generalization for Diabetic Retinopathy
Diabetic Retinopathy (DR), a leading cause of vision impairment, requires early detection and treatment. Developing robust AI models for DR classification holds substantial potential, but a key challenge is ensuring their generalization in unfamiliar domains with varying data distributions. To address this, our paper investigates cross-domain generalization, also known as domain generalization (DG), within the context of DR classification. DG, a challenging problem in the medical domain, is complicated by the difficulty of gathering labeled data across different domains, such as patient demographics and disease stages. Some recent studies have shown the effectiveness of using CLIP to handle the DG problem in natural images. In this study, we investigate CLIP's transfer learning capabilities and its potential for cross-domain generalization in diabetic retinopathy (DR) classification. We carry out comprehensive experiments to assess the efficacy and potential of CLIP in addressing DG for DR classification. Further, we introduce a multi-modal fine-tuning strategy named Context Optimization with Learnable Visual Tokens (CoOpLVT), which enhances context optimization by conditioning on visual features. Our findings demonstrate that the proposed method increases the F1-score by 1.8% over the baseline, thus underlining its promise for effective DG in DR classification. Our code is publicly available at https://github.com/Sanoojan/CLIP-DRDG.