Salim Arslan
Salim Arslan has a total of 66 co-authors. Here are the top co-authors by number of shared publications:
Daniel Rueckert
Biomedical Image Analysis Group, Imperial College London, 180 Queen's Gate, London SW7 2AZ, UK
7 shared publications- Andre Geraldes
Panakeia Technologies
4 shared publications - Cher Bass
Panakeia Technologies, Cambridge, UK
4 shared publications Sarah Parisot
Biomedical Image Analysis Group, Imperial College London, 180 Queen's Gate, London SW7 2AZ, UK
4 shared publicationsDebapriya Mehrotra
Panakeia Technologies, London, UK
3 shared publications- Jakob Nikolas Kather
Else Kroener Fresenius Center for Digital Health, Technical University Dresden, Dresden, Germany
3 shared publications
Latest Publications
- Open Access
- 16 June 2026
- Vol. 15, e76785
Deep-Learning Solution Providing Molecular Marker Subtyping of Breast Cancer Whole Slide Images: Protocol for a UK Clinical Service Evaluation Study.
As the histopathology workforce continues to struggle and service demand continues to increase, it has become prudent to consider viable avenues to try to alleviate diagnostic workload burden. One such avenue is computer-based technologies (CBTs). Breast cancer (BC) is the most common malignant neoplasm in the United Kingdom and requires additional testing for estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor-2 (HER2) status at the time of histological diagnosis. This makes BC diagnostics a promising candidate for the application of an efficient CBT. However, for clinical acceptance, these technologies must prove that they work within a real-life diagnostic environment. We present a study protocol for a prospective clinical service evaluation aimed to validate a UK Conformity Assessed-marked CBT's ability to provide ER, PR, and HER2 results for invasive BCs from scanned hematoxylin and eosin-stained whole slide images. This protocol has been designed to use and mimic a preexisting digital pathology workflow within a National Health Service tertiary referral cancer center without disrupting normal patient care. Eligible cases are identified prospectively through the laboratory information management system, and their whole slide images are extracted from the clinical digital workflow. After verification of national data opt-out status and the exclusion of appropriate cases (N=400 analyzable cases), these cases are analyzed on a dedicated computer in parallel to the existing clinical workflow by a UK Conformity Assessed-marked deep learning-based CBT in a separate environment, providing results for ER, PR, and HER2 status. These results are compared to the ER, PR, and HER2 status reported on the corresponding pathology report. To evaluate the CBT's performance, a range of accepted concordance measures will be applied, including specificity, sensitivity, false-positive rate, false-negative rate, positive predictive value, and negative predictive value. Moreover, time stamps representing the duration of image analysis will also be collected. This study started in April 2025. There are no results to present, as this paper focuses on study design, and results have yet to be generated. As of March 2026, overall, 366 potentially analyzable cases have been collected. The anticipated end date of the study is May 2026 (400-case target). Results will be presented in a separate publication. This design assesses a CBT within a clinical environment while effectively eliminating any unwanted effects on patient care. This type of service evaluation provides a useful step to establish confidence in a CBT before trialing its effect on patient care. It also offers the opportunity to support interventional randomized controlled trials, health economic evaluations, and usability studies. This protocol will hopefully prove useful to others who wish to conduct a similar service evaluation at their own institution. DERR1-10.2196/76785.
- 16 June 2026
- Vol. 15, e76785
Open AccessDeep-Learning Solution Providing Molecular Marker Subtyping of Breast Cancer Whole Slide Images: Protocol for a UK Clinical Service Evaluation Study.
As the histopathology workforce continues to struggle and service demand continues to increase, it has become prudent to consider viable avenues to try to alleviate diagnostic workload burden. One such avenue is computer-based technologies (CBTs). Breast cancer (BC) is the most common malignant neoplasm in the United Kingdom and requires additional testing for estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor-2 (HER2) status at the time of histological diagnosis. This makes BC diagnostics a promising candidate for the application of an efficient CBT. However, for clinical acceptance, these technologies must prove that they work within a real-life diagnostic environment. We present a study protocol for a prospective clinical service evaluation aimed to validate a UK Conformity Assessed-marked CBT's ability to provide ER, PR, and HER2 results for invasive BCs from scanned hematoxylin and eosin-stained whole slide images. This protocol has been designed to use and mimic a preexisting digital pathology workflow within a National Health Service tertiary referral cancer center without disrupting normal patient care. Eligible cases are identified prospectively through the laboratory information management system, and their whole slide images are extracted from the clinical digital workflow. After verification of national data opt-out status and the exclusion of appropriate cases (N=400 analyzable cases), these cases are analyzed on a dedicated computer in parallel to the existing clinical workflow by a UK Conformity Assessed-marked deep learning-based CBT in a separate environment, providing results for ER, PR, and HER2 status. These results are compared to the ER, PR, and HER2 status reported on the corresponding pathology report. To evaluate the CBT's performance, a range of accepted concordance measures will be applied, including specificity, sensitivity, false-positive rate, false-negative rate, positive predictive value, and negative predictive value. Moreover, time stamps representing the duration of image analysis will also be collected. This study started in April 2025. There are no results to present, as this paper focuses on study design, and results have yet to be generated. As of March 2026, overall, 366 potentially analyzable cases have been collected. The anticipated end date of the study is May 2026 (400-case target). Results will be presented in a separate publication. This design assesses a CBT within a clinical environment while effectively eliminating any unwanted effects on patient care. This type of service evaluation provides a useful step to establish confidence in a CBT before trialing its effect on patient care. It also offers the opportunity to support interventional randomized controlled trials, health economic evaluations, and usability studies. This protocol will hopefully prove useful to others who wish to conduct a similar service evaluation at their own institution. DERR1-10.2196/76785.
- Open Access
- 14 April 2026
- Published by Elsevier in ESMO Real World Data and Digital Oncology
- Vol. 12, 100702
Multi-site validation of an AI-based biomarker test for determining ER, PR, and HER2 status from H&E-stained breast cancer slides
- 14 April 2026
- Published by Elsevier in ESMO Real World Data and Digital Oncology
- Vol. 12, 100702
Open AccessMulti-site validation of an AI-based biomarker test for determining ER, PR, and HER2 status from H&E-stained breast cancer slides
- Open Access
- 15 March 2024
- Published by Springer Nature in Communications Medicine
- Vol. 4 (1) , 1-15
A systematic pan-cancer study on deep learning-based prediction of multi-omic biomarkers from routine pathology images
Background: The objective of this comprehensive pan-cancer study is to evaluate the potential of deep learning (DL) for molecular profiling of multi-omic biomarkers directly from hematoxylin and eosin (H&E)-stained whole slide images. Methods: A total of 12,093 DL models predicting 4031 multi-omic biomarkers across 32 cancer types were trained and validated. The study included a broad range of genetic, transcriptomic, and proteomic biomarkers, as well as established prognostic markers, molecular subtypes, and clinical outcomes. Results: Here we show that 50% of the models achieve an area under the curve (AUC) of 0.644 or higher. The observed AUC for 25% of the models is at least 0.719 and exceeds 0.834 for the top 5%. Molecular profiling with image-based histomorphological features is generally considered feasible for most of the investigated biomarkers and across different cancer types. The performance appears to be independent of tumor purity, sample size, and class ratio (prevalence), suggesting a degree of inherent predictability in histomorphology. Conclusions: The results demonstrate that DL holds promise to predict a wide range of biomarkers across the omics spectrum using only H&E-stained histological slides of solid tumors. This paves the way for accelerating diagnosis and developing more precise treatments for cancer patients.
- 15 March 2024
- Published by Springer Nature in Communications Medicine
- Vol. 4 (1) , 1-15
Open AccessA systematic pan-cancer study on deep learning-based prediction of multi-omic biomarkers from routine pathology images
Background: The objective of this comprehensive pan-cancer study is to evaluate the potential of deep learning (DL) for molecular profiling of multi-omic biomarkers directly from hematoxylin and eosin (H&E)-stained whole slide images. Methods: A total of 12,093 DL models predicting 4031 multi-omic biomarkers across 32 cancer types were trained and validated. The study included a broad range of genetic, transcriptomic, and proteomic biomarkers, as well as established prognostic markers, molecular subtypes, and clinical outcomes. Results: Here we show that 50% of the models achieve an area under the curve (AUC) of 0.644 or higher. The observed AUC for 25% of the models is at least 0.719 and exceeds 0.834 for the top 5%. Molecular profiling with image-based histomorphological features is generally considered feasible for most of the investigated biomarkers and across different cancer types. The performance appears to be independent of tumor purity, sample size, and class ratio (prevalence), suggesting a degree of inherent predictability in histomorphology. Conclusions: The results demonstrate that DL holds promise to predict a wide range of biomarkers across the omics spectrum using only H&E-stained histological slides of solid tumors. This paves the way for accelerating diagnosis and developing more precise treatments for cancer patients.
- Open Access
Deep learning can predict multi-omic biomarkers from routine pathology images: A systematic large-scale study
We assessed the pan-cancer predictability of multi-omic biomarkers from haematoxylin and eosin (H&E)-stained whole slide images (WSI) using deep learning (DL) throughout a systematic study. A total of 13,443 DL models predicting 4,481 multi-omic biomarkers across 32 cancer types were trained and validated. The investigated biomarkers included a broad range of genetic, transcriptomic, proteomic, and metabolic alterations, as well as established markers relevant for prognosis, molecular subtypes and clinical outcomes. Overall, we found that DL can predict multi-omic biomarkers directly from routine histology images across solid cancer types, with 50% of the models performing at an area under the curve (AUC) of more than 0.633 (with 25% of the models having an AUC larger than 0.711). A wide range of biomarkers were detectable from routine histology images across all investigated cancer types, with a mean AUC of at least 0.62 in almost all malignancies. Strikingly, we observed that biomarker predictability was mostly consistent and not dependent on sample size and class ratio, suggesting a degree of true predictability inherent in histomorphology. Together, the results of our study show the potential of DL to predict a multitude of biomarkers across the omics spectrum using only routine slides. This paves the way for accelerating diagnosis and developing more precise treatments for cancer patients.
Open AccessDeep learning can predict multi-omic biomarkers from routine pathology images: A systematic large-scale study
We assessed the pan-cancer predictability of multi-omic biomarkers from haematoxylin and eosin (H&E)-stained whole slide images (WSI) using deep learning (DL) throughout a systematic study. A total of 13,443 DL models predicting 4,481 multi-omic biomarkers across 32 cancer types were trained and validated. The investigated biomarkers included a broad range of genetic, transcriptomic, proteomic, and metabolic alterations, as well as established markers relevant for prognosis, molecular subtypes and clinical outcomes. Overall, we found that DL can predict multi-omic biomarkers directly from routine histology images across solid cancer types, with 50% of the models performing at an area under the curve (AUC) of more than 0.633 (with 25% of the models having an AUC larger than 0.711). A wide range of biomarkers were detectable from routine histology images across all investigated cancer types, with a mean AUC of at least 0.62 in almost all malignancies. Strikingly, we observed that biomarker predictability was mostly consistent and not dependent on sample size and class ratio, suggesting a degree of true predictability inherent in histomorphology. Together, the results of our study show the potential of DL to predict a multitude of biomarkers across the omics spectrum using only routine slides. This paves the way for accelerating diagnosis and developing more precise treatments for cancer patients.
- Open Access
Evaluation of a predictive method for the H&E-based molecular profiling of breast cancer with deep learning
We present a public validation of PANProfiler (ER, PR, HER2), an in-vitro medical device (IVD) that predicts the qualitative status of estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) by analysing the hematoxylin and eosin (H&E)-stained tissue scan. In public validation on 648 (ER), 648 (PR) and 560 (HER2) unseen cases with known biomarker status, the device achieves an accuracy of 87% (ER), 83% (PR) and 87% (HER2). The validation offers early evidence of the ability to predict clinically relevant breast biomarkers from an H&E slide in a relevant clinical setting.
Open AccessEvaluation of a predictive method for the H&E-based molecular profiling of breast cancer with deep learning
We present a public validation of PANProfiler (ER, PR, HER2), an in-vitro medical device (IVD) that predicts the qualitative status of estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) by analysing the hematoxylin and eosin (H&E)-stained tissue scan. In public validation on 648 (ER), 648 (PR) and 560 (HER2) unseen cases with known biomarker status, the device achieves an accuracy of 87% (ER), 83% (PR) and 87% (HER2). The validation offers early evidence of the ability to predict clinically relevant breast biomarkers from an H&E slide in a relevant clinical setting.
- Top Cited PapersOpen Access
- 1 April 2018
- Published by Elsevier in NeuroImage
- Vol. 170, 5-30
Human brain mapping: A systematic comparison of parcellation methods for the human cerebral cortex
- 1 April 2018
- Published by Elsevier in NeuroImage
- Vol. 170, 5-30
Top Cited PapersOpen AccessHuman brain mapping: A systematic comparison of parcellation methods for the human cerebral cortex
- Open Access
- 1 November 2017
- Published by Elsevier in NeuroImage
- Vol. 162, 226-248
A flexible graphical model for multi-modal parcellation of the cortex
- 1 November 2017
- Published by Elsevier in NeuroImage
- Vol. 162, 226-248
Open AccessA flexible graphical model for multi-modal parcellation of the cortex
- Open Access
- 1 April 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 354-357
Exploring heritability of functional brain networks with inexact graph matching
Data-driven brain parcellations aim to provide a more accurate representation of an individual's functional connectivity, since they are able to capture individual variability that arises due to development or disease. This renders comparisons between the emerging brain connectivity networks more challenging, since correspondences between their elements are not preserved. Unveiling these correspondences is of major importance to keep track of local functional connectivity changes. We propose a novel method based on graph edit distance for the comparison of brain graphs directly in their domain, that can accurately reflect similarities between individual networks while providing the network element correspondences. This method is validated on a dataset of 116 twin subjects provided by the Human Connectome Project.
- 1 April 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 354-357
Open AccessExploring heritability of functional brain networks with inexact graph matching
Data-driven brain parcellations aim to provide a more accurate representation of an individual's functional connectivity, since they are able to capture individual variability that arises due to development or disease. This renders comparisons between the emerging brain connectivity networks more challenging, since correspondences between their elements are not preserved. Unveiling these correspondences is of major importance to keep track of local functional connectivity changes. We propose a novel method based on graph edit distance for the comparison of brain graphs directly in their domain, that can accurately reflect similarities between individual networks while providing the network element correspondences. This method is validated on a dataset of 116 twin subjects provided by the Human Connectome Project.
- 2 October 2016
- Published by Springer Nature
- p. 115-122
Boundary Mapping Through Manifold Learning for Connectivity-Based Cortical Parcellation
- 2 October 2016
- Published by Springer Nature
- p. 115-122
Boundary Mapping Through Manifold Learning for Connectivity-Based Cortical Parcellation
- 1 August 2016
- Published by Elsevier in NeuroImage
- Vol. 136, 68-83
Group-wise parcellation of the cortex through multi-scale spectral clustering
- 1 August 2016
- Published by Elsevier in NeuroImage
- Vol. 136, 68-83
Group-wise parcellation of the cortex through multi-scale spectral clustering