Abhir Bhalerao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Abhir Bhalerao has a total of 53 co-authors. Here are the top co-authors by number of shared publications:
- Jun Wang
Department of Computer Science, University of Warwick, Coventry, U.K.
7 shared publications - Yulan He
King's College London
4 shared publications - Kai Liang
Department of Computer Science University of Warwick UK
2 shared publications - Nasir Rajpoot
Tissue Image Analytics Centre, University of Warwick, Coventry, UK
2 shared publications - Shan E. Ahmed Raza
University of Warwick
2 shared publications - Simon Graham
Department of Computer Science, University of Warwick, Coventry, U.K.
2 shared publications
Latest Publications
- 23 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 30  (5) , 4214-4225
Improving Medical Visual Representation Learning With Pathological-Level Cross-Modal Alignment and Correlation Exploration
Learning medical visual representations from image-report pairs through joint learning has garnered increasing research attention due to its potential for transferring acquired knowledge to various downstream medical tasks. Previous works have predominantly focused on instance-wise or token-wise cross-modal alignment, often neglecting the importance of pathological-level consistency. This paper presents a novel framework PLACE that promotes the Pathological-Level Alignment and enriches the fine-grained details via Correlation Exploration without additional human annotations. Specifically, we propose a novel pathological-level cross-modal alignment (PCMA) approach to maximize the consistency of pathology observations from both images and reports. To facilitate this, a Visual Pathology Observation Extractor is introduced to extract visual pathological observation representations from localized tokens. The PCMA module operates independently of any external disease annotations, enhancing the generalizability and robustness of our methods. Furthermore, we design a proxy task that enforces the model to identify correlations among image patches, thereby enriching the fine-grained details crucial for various downstream tasks. Experimental results demonstrate that our proposed framework achieves new state-of-the-art performance on multiple downstream tasks, including classification, image-to-text retrieval, semantic segmentation, object detection and report generation.
- 23 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 30  (5) , 4214-4225
Improving Medical Visual Representation Learning With Pathological-Level Cross-Modal Alignment and Correlation Exploration
Learning medical visual representations from image-report pairs through joint learning has garnered increasing research attention due to its potential for transferring acquired knowledge to various downstream medical tasks. Previous works have predominantly focused on instance-wise or token-wise cross-modal alignment, often neglecting the importance of pathological-level consistency. This paper presents a novel framework PLACE that promotes the Pathological-Level Alignment and enriches the fine-grained details via Correlation Exploration without additional human annotations. Specifically, we propose a novel pathological-level cross-modal alignment (PCMA) approach to maximize the consistency of pathology observations from both images and reports. To facilitate this, a Visual Pathology Observation Extractor is introduced to extract visual pathological observation representations from localized tokens. The PCMA module operates independently of any external disease annotations, enhancing the generalizability and robustness of our methods. Furthermore, we design a proxy task that enforces the model to identify correlations among image patches, thereby enriching the fine-grained details crucial for various downstream tasks. Experimental results demonstrate that our proposed framework achieves new state-of-the-art performance on multiple downstream tasks, including classification, image-to-text retrieval, semantic segmentation, object detection and report generation.
- Open Access
- 16 September 2025
- Published by Elsevier in Computerized Medical Imaging and Graphics
- Vol. 125, 102644
SGRRG: Leveraging radiology scene graphs for improved and abnormality-aware radiology report generation
- 16 September 2025
- Published by Elsevier in Computerized Medical Imaging and Graphics
- Vol. 125, 102644
Open AccessSGRRG: Leveraging radiology scene graphs for improved and abnormality-aware radiology report generation
- Open Access
Characterization of Breast Cancer Intra-Tumor Heterogeneity Using Artificial Intelligence
Intra-tumor heterogeneity (ITH) is a fundamental characteristic of breast cancer (BC), influencing tumor progression, prognosis, and therapeutic responses. However, the complexity of ITH in BC makes its accurate characterization challenging. This study leverages deep learning (DL) techniques to comprehensively evaluate ITH in early-stage luminal BC and provide a nuanced understanding of its impact on tumor behavior and patient outcomes. A large cohort (n = 2561) of early-stage luminal BC was evaluated using whole slide images (WSIs) of hematoxylin and eosin-stained slides of excision specimens. Morphological features of both the tumor and stromal components were meticulously annotated by a panel of pathologists in a subset of cases. A DL model was applied to develop an algorithm to assess the degree of heterogeneity of various morphological features per individual case utilizing defined patches. The results of extracted features were used to generate an overall heterogeneity score that was correlated with the clinicopathological features and outcome. Overall, 162 features were quantified and a significant positive correlation between these features was identified. Specifically, there was a significant association between a high degree of intra-tumor heterogeneity and larger tumor size, poorly differentiated tumors, highly proliferative tumors, tumors of no special type (NST), and those with low estrogen receptor (ER) expression. When all features are considered in combination, a high overall heterogeneity score was significantly associated with parameters characteristic of aggressive tumor behavior, and it was an independent predictor of poor patient outcome. In conclusion, DL models can be used to accurately decipher the complexity of ITH and provide extra information for outcome prediction.
Open AccessCharacterization of Breast Cancer Intra-Tumor Heterogeneity Using Artificial Intelligence
Intra-tumor heterogeneity (ITH) is a fundamental characteristic of breast cancer (BC), influencing tumor progression, prognosis, and therapeutic responses. However, the complexity of ITH in BC makes its accurate characterization challenging. This study leverages deep learning (DL) techniques to comprehensively evaluate ITH in early-stage luminal BC and provide a nuanced understanding of its impact on tumor behavior and patient outcomes. A large cohort (n = 2561) of early-stage luminal BC was evaluated using whole slide images (WSIs) of hematoxylin and eosin-stained slides of excision specimens. Morphological features of both the tumor and stromal components were meticulously annotated by a panel of pathologists in a subset of cases. A DL model was applied to develop an algorithm to assess the degree of heterogeneity of various morphological features per individual case utilizing defined patches. The results of extracted features were used to generate an overall heterogeneity score that was correlated with the clinicopathological features and outcome. Overall, 162 features were quantified and a significant positive correlation between these features was identified. Specifically, there was a significant association between a high degree of intra-tumor heterogeneity and larger tumor size, poorly differentiated tumors, highly proliferative tumors, tumors of no special type (NST), and those with low estrogen receptor (ER) expression. When all features are considered in combination, a high overall heterogeneity score was significantly associated with parameters characteristic of aggressive tumor behavior, and it was an independent predictor of poor patient outcome. In conclusion, DL models can be used to accurately decipher the complexity of ITH and provide extra information for outcome prediction.
- Open Access
Lane Change Classification and Prediction with Action Recognition Networks
Anticipating lane change intentions of surrounding vehicles is crucial for efficient and safe driving decision making in an autonomous driving system. Previous works often adopt physical variables such as driving speed, acceleration and so forth for lane change classification. However, physical variables do not contain semantic information. Although 3D CNNs have been developing rapidly, the number of methods utilising action recognition models and appearance feature for lane change recognition is low, and they all require additional information to pre-process data. In this work, we propose an end-to-end framework including two action recognition methods for lane change recognition, using video data collected by cameras. Our method achieves the best lane change classification results using only the RGB video data of the PREVENTION dataset. Class activation maps demonstrate that action recognition models can efficiently extract lane change motions. A method to better extract motion clues is also proposed in this paper.
Open AccessLane Change Classification and Prediction with Action Recognition Networks
Anticipating lane change intentions of surrounding vehicles is crucial for efficient and safe driving decision making in an autonomous driving system. Previous works often adopt physical variables such as driving speed, acceleration and so forth for lane change classification. However, physical variables do not contain semantic information. Although 3D CNNs have been developing rapidly, the number of methods utilising action recognition models and appearance feature for lane change recognition is low, and they all require additional information to pre-process data. In this work, we propose an end-to-end framework including two action recognition methods for lane change recognition, using video data collected by cameras. Our method achieves the best lane change classification results using only the RGB video data of the PREVENTION dataset. Class activation maps demonstrate that action recognition models can efficiently extract lane change motions. A method to better extract motion clues is also proposed in this paper.
- Open Access
CAMANet: Class Activation Map Guided Attention Network for Radiology Report Generation
Radiology report generation (RRG) has gained increasing research attention because of its huge potential to mitigate medical resource shortages and aid the process of disease decision making by radiologists. Recent advancements in RRG are largely driven by improving a model's capabilities in encoding single-modal feature representations, while few studies explicitly explore the cross-modal alignment between image regions and words. Radiologists typically focus first on abnormal image regions before composing the corresponding text descriptions, thus cross-modal alignment is of great importance to learn a RRG model which is aware of abnormalities in the image. Motivated by this, we propose a Class Activation Map guided Attention Network (CAMANet) which explicitly promotes crossmodal alignment by employing aggregated class activation maps to supervise cross-modal attention learning, and simultaneously enrich the discriminative information. CAMANet contains three complementary modules: a Visual Discriminative Map Generation module to generate the importance/contribution of each visual token; Visual Discriminative Map Assisted Encoder to learn the discriminative representation and enrich the discriminative information; and a Visual Textual Attention Consistency module to ensure the attention consistency between the visual and textual tokens, to achieve the cross-modal alignment. Experimental results demonstrate that CAMANet outperforms previous SOTA methods on two commonly used RRG benchmarks.
Open AccessCAMANet: Class Activation Map Guided Attention Network for Radiology Report Generation
Radiology report generation (RRG) has gained increasing research attention because of its huge potential to mitigate medical resource shortages and aid the process of disease decision making by radiologists. Recent advancements in RRG are largely driven by improving a model's capabilities in encoding single-modal feature representations, while few studies explicitly explore the cross-modal alignment between image regions and words. Radiologists typically focus first on abnormal image regions before composing the corresponding text descriptions, thus cross-modal alignment is of great importance to learn a RRG model which is aware of abnormalities in the image. Motivated by this, we propose a Class Activation Map guided Attention Network (CAMANet) which explicitly promotes crossmodal alignment by employing aggregated class activation maps to supervise cross-modal attention learning, and simultaneously enrich the discriminative information. CAMANet contains three complementary modules: a Visual Discriminative Map Generation module to generate the importance/contribution of each visual token; Visual Discriminative Map Assisted Encoder to learn the discriminative representation and enrich the discriminative information; and a Visual Textual Attention Consistency module to ensure the attention consistency between the visual and textual tokens, to achieve the cross-modal alignment. Experimental results demonstrate that CAMANet outperforms previous SOTA methods on two commonly used RRG benchmarks.
- 16 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 28  (4) , 2199-2210
CAMANet: Class Activation Map Guided Attention Network for Radiology Report Generation
Radiology report generation (RRG) has gained increasing research attention because of its huge potential to mitigate medical resource shortages and aid the process of disease decision making by radiologists. Recent advancements in Radiology Report Generation (RRG) are largely driven by improving a model's capabilities in encoding single-modal feature representations, while few studies explicitly explore the cross-modal alignment between image regions and words. Radiologists typically focus first on abnormal image regions before composing the corresponding text descriptions, thus cross-modal alignment is of great importance to learn a RRG model which is aware of abnormalities in the image. Motivated by this, we propose a C lass A ctivation M ap guided A ttention Net work (CAMANet) which explicitly promotes cross-modal alignment by employing aggregated class activation maps to supervise cross-modal attention learning, and simultaneously enrich the discriminative information. Experimental results demonstrate that CAMANet outperforms previous SOTA methods on two commonly used RRG benchmarks.
- 16 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 28  (4) , 2199-2210
CAMANet: Class Activation Map Guided Attention Network for Radiology Report Generation
Radiology report generation (RRG) has gained increasing research attention because of its huge potential to mitigate medical resource shortages and aid the process of disease decision making by radiologists. Recent advancements in Radiology Report Generation (RRG) are largely driven by improving a model's capabilities in encoding single-modal feature representations, while few studies explicitly explore the cross-modal alignment between image regions and words. Radiologists typically focus first on abnormal image regions before composing the corresponding text descriptions, thus cross-modal alignment is of great importance to learn a RRG model which is aware of abnormalities in the image. Motivated by this, we propose a C lass A ctivation M ap guided A ttention Net work (CAMANet) which explicitly promotes cross-modal alignment by employing aggregated class activation maps to supervise cross-modal attention learning, and simultaneously enrich the discriminative information. Experimental results demonstrate that CAMANet outperforms previous SOTA methods on two commonly used RRG benchmarks.
- Open Access
- 25 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 28  (3) , 1161-1172
LYSTO: The Lymphocyte Assessment Hackathon and Benchmark Dataset
We introduce LYSTO, the Lymphocyte Assessment Hackathon, which was held in conjunction with the MICCAI 2019 Conference in Shenzhen (China). The competition required participants to automatically assess the number of lymphocytes, in particular T-cells, in images of colon, breast, and prostate cancer stained with CD3 and CD8 immunohistochemistry. Differently from other challenges setup in medical image analysis, LYSTO participants were solely given a few hours to address this problem. In this paper, we describe the goal and the multi-phase organization of the hackathon; we describe the proposed methods and the on-site results. Additionally, we present post-competition results where we show how the presented methods perform on an independent set of lung cancer slides, which was not part of the initial competition, as well as a comparison on lymphocyte assessment between presented methods and a panel of pathologists. We show that some of the participants were capable to achieve pathologist-level performance at lymphocyte assessment. After the hackathon, LYSTO was left as a lightweight plug-and-play benchmark dataset on grand-challenge website, together with an automatic evaluation platform. LYSTO has supported a number of research in lymphocyte assessment in oncology. LYSTO will be a long-lasting educational challenge for deep learning and digital pathology, it is available at https://lysto.grand-challenge.org/ .
- 25 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 28  (3) , 1161-1172
Open AccessLYSTO: The Lymphocyte Assessment Hackathon and Benchmark Dataset
We introduce LYSTO, the Lymphocyte Assessment Hackathon, which was held in conjunction with the MICCAI 2019 Conference in Shenzhen (China). The competition required participants to automatically assess the number of lymphocytes, in particular T-cells, in images of colon, breast, and prostate cancer stained with CD3 and CD8 immunohistochemistry. Differently from other challenges setup in medical image analysis, LYSTO participants were solely given a few hours to address this problem. In this paper, we describe the goal and the multi-phase organization of the hackathon; we describe the proposed methods and the on-site results. Additionally, we present post-competition results where we show how the presented methods perform on an independent set of lung cancer slides, which was not part of the initial competition, as well as a comparison on lymphocyte assessment between presented methods and a panel of pathologists. We show that some of the participants were capable to achieve pathologist-level performance at lymphocyte assessment. After the hackathon, LYSTO was left as a lightweight plug-and-play benchmark dataset on grand-challenge website, together with an automatic evaluation platform. LYSTO has supported a number of research in lymphocyte assessment in oncology. LYSTO will be a long-lasting educational challenge for deep learning and digital pathology, it is available at https://lysto.grand-challenge.org/ .
- 15 February 2023
- Published by Springer Nature
- p. 617-632
Lane Change Classification and Prediction with Action Recognition Networks
- 15 February 2023
- Published by Springer Nature
- p. 617-632
Lane Change Classification and Prediction with Action Recognition Networks
- Top Cited Papers
- 4 November 2022
- Published by Springer Nature
- p. 563-579
Cross-Modal Prototype Driven Network for Radiology Report Generation
- 4 November 2022
- Published by Springer Nature
- p. 563-579
Top Cited PapersCross-Modal Prototype Driven Network for Radiology Report Generation
- Open Access
- 15 March 2016
- Published by Taylor & Francis in Applied Artificial Intelligence
- Vol. 30  (3) , 233-256
Data Mining for Vehicle Telemetry
This article presents a data mining methodology for driving-condition monitoring via CAN-bus data that is based on the general data mining process. The approach is applicable to many driving condition problems, and the example of road type classification without the use of location information is investigated. Location information from Global Positioning Satellites and related map data are often not available (for business reasons), or cannot represent the full dynamics of road conditions. In this work, Controller Area Network (CAN)-bus signals are used instead as inputs to models produced by machine learning algorithms. Road type classification is formulated as two related labeling problems: Road Type (A, B, C, and Motorway) and Carriageway Type (Single or Dual). An investigation is presented into preprocessing steps required prior to applying machine learning algorithms, that is, signal selection, feature extraction, and feature selection. The selection methods used include principal components analysis (PCA) and mutual information (MI), which are used to determine the relevance and redundancy of extracted features and are performed in various combinations. Finally, because there is an inherent bias toward certain road and carriageway labelings, the issue of class imbalance in classification is explained and investigated. A system is produced, which is demonstrated to successfully ascertain road type from CAN-bus data, and it is shown that the classification correlates well with input signals such as vehicle speed, steering wheel angle, and suspension height.
- 15 March 2016
- Published by Taylor & Francis in Applied Artificial Intelligence
- Vol. 30  (3) , 233-256
Open AccessData Mining for Vehicle Telemetry
This article presents a data mining methodology for driving-condition monitoring via CAN-bus data that is based on the general data mining process. The approach is applicable to many driving condition problems, and the example of road type classification without the use of location information is investigated. Location information from Global Positioning Satellites and related map data are often not available (for business reasons), or cannot represent the full dynamics of road conditions. In this work, Controller Area Network (CAN)-bus signals are used instead as inputs to models produced by machine learning algorithms. Road type classification is formulated as two related labeling problems: Road Type (A, B, C, and Motorway) and Carriageway Type (Single or Dual). An investigation is presented into preprocessing steps required prior to applying machine learning algorithms, that is, signal selection, feature extraction, and feature selection. The selection methods used include principal components analysis (PCA) and mutual information (MI), which are used to determine the relevance and redundancy of extracted features and are performed in various combinations. Finally, because there is an inherent bias toward certain road and carriageway labelings, the issue of class imbalance in classification is explained and investigated. A system is produced, which is demonstrated to successfully ascertain road type from CAN-bus data, and it is shown that the classification correlates well with input signals such as vehicle speed, steering wheel angle, and suspension height.