The Radiogenomics and Quantitative Imaging Group led by Prof Evis Sala is a multi-disciplinary team of radiologists, physicists, oncologists and computational scientists. One of the most active areas of research is in ovarian cancer. The group has developed novel computational methods for data integration and prediction of treatment response in the setting of neoadjuvant chemotherapy.
Research Team
Principal Investigator:
Research Associates:
PhD Students:
- Thomas Buddenkotte
- María Delgado Ortet
- David Hulse
- Paula Martin Gonzalez
- Cathal McCague
- Syafiq Ramlee
- Marika Reinus
- Tanvi Rao
- Ian Selby
- Vencel Somai
- Stephan Ursprung
Visitors:
- Amandine Crombe
- Stephanie Nougaret
- Roxana Pitican
Projects
Real-time habitat-guided multimodal fusion biopsies
In collaboration with Canon® Medical Systems, the group has recently developed and has completed piloting a new ultrasound-guided fusion biopsy technique, which allows for selective tissue sampling of tumour areas with distinct or similar radiomic/texture characteristics. The group has managed to overlay texture-based clustering maps onto diagnostic CT images and PET/MRI habitat maps onto MRI images, respectively and then fuse them with the US during biopsies. This technique will enable biological profiling and longitudinal tracking of tumour habitats in the clinical settings.

Figure 1. CT habitat-guided US fusion biopsies. On the left side of the figure, the CT with the CT texture habitat overlay is shown. The tumour habitats are highlighted in green, red and blue, respectively. On the right side, the fused US image is shown.
The research led by Professor Evis Sala and funded through the MFICM and CRUK shows that combining CT with US images creates a visual guide for radiologists to ensure they sample the full complexity of a tumour with fewer targeted biopsies.
The publication, with Co-first authors Dr Lucian Beer and Paula Martin-Gonzalez from Radiogenomics and Quantitative Imaging group “Ultrasound-guided targeted biopsies of CT-based radiomic tumour habitats” published in European Radiology, 14 Dec 2020, has attracted national media attention resulting from a press release sent out by the University of Cambridge press office on 6 January 2021, the MFICM news page and other pages on the MFICM website, in the CRUK Cambridge Centre website.
Media coverage for this article: Daily Mail | iNews | Health Europa | Science Times | Cancer World | Imaging Wire | Diagnostic | World News
Ovarian cancer segmentation and tissue-specific sub-segmentation on CT images
Currently, abdominal CT scans are evaluated visually and response is assessed based on unidimensional tumour measurements at different time points. Segmentation of the whole tumour burden is not routinely performed, as it is labour-intensive and also operator-dependent. Despite the recent advances in machine learning techniques, automated image segmentation for ovarian cancer is not yet available for clinical implementation owing to challenges including low tumour-to-tissue contrast and a lack of large-scale datasets for training and testing. This group has developed a prototype deep learning automated segmentation method for ovarian cancer on CT scans.

Segmentation examples for a pelvic/ovarian lesions and an omental lesion. The automated and manual segmentations from two observers are displayed. The colour legend is shown at the bottom of the figures.
This group has also proposed the first approach based on unsupervised clustering techniques to provide tissue-specific sub-segmentation and quantify solid and cystic components, which are commonly found in ovarian cancer lesions. Exploiting these novel techniques, we aim at providing robust tools for future radiomic biomarker validation. This work, entitled “Tissue-specific and interpretable sub-segmentation of whole tumour burden on CT images by unsupervised fuzzy clustering”, was published in Computers in Biology and Medicine (2020).
All in One cancer imaging optimisation using an integrated mathematical and deep learning approach
This project is funded by Wellcome Trust Innovator Award and led by Prof. Evis Sala and Prof. Carola Bibiane Schönlieb from Department of Applied Mathematics and Theoretical Physics (DAMTP).

Overall workflow of the proposed end-to-end deep learning approach: starting from computed tomography (CT) image reconstruction jointly performed with automated segmentation, the downstream radiomics analyses are independent of scanner manufacturers and can be integrated into clinical decision-making tasks for robust biomarker development.
Recent developments in mathematical image analysis, in particular deep neural networks, have unravelled the potential of robust image quantification which is essential to cancer care. Despite these achievements, the high variability of image acquisition and reconstruction and the multiple individual optimisation steps of the imaging pipeline (acquisition, reconstruction, segmentation and classification) remain two major challenges. This leads to inherent biases in image quantification and to the development of suboptimal quantitative imaging biomarkers that lack reproducibility. As a consequence, the analysis of these biomarkers has limited use in the clinic which becomes even more challenging for multi-centre clinical trials. Our project therefore aims to:
- Apply novel deep learning algorithms, recently developed by our group, to reconstruct medical images in a fully unbiased manner.
- Perform the reconstruction, segmentation, and prediction steps jointly, thereby creating an end-to-end pipeline optimised for the task at hand.
- Develop robust evaluation metrics that will be used to analyse large heterogeneous retrospective imaging datasets.
Integration of Radiomics and Circulating Tumour DNA (ctDNA) for the Development of Predictive and Prognostic Biomarkers on Patients with Ovarian Cancer
Patients with suspected or confirmed ovarian cancer usually undergo computed tomography (CT) scans as their standard-of-care imaging to assess the extent of their disease. We are collaborating with molecular cancer biology and computational groups to extract quantitative data from these images (radiomics) and combine them with information about circulating tumour DNA (ctDNA) from blood samples, to develop tools that predict response to chemotherapy as well as survival in patients with ovarian cancer. These studies also help us understand which features of ovarian cancer affect the extent of ctDNA shedding into circulation and how molecular features of ovarian cancer affect tumour appearance on imaging.
Integration of Proteomics and Radiomics in Ovarian Cancer
In collaboration with colleagues from the Walter Reed National Military Medical Center, the Inova Schar Cancer Institute and Frederick National Laboratory for Cancer Research our group is working on the integration of proteomic- and radiomic data into clinical decision models for patients with ovarian cancer. Over the last years, our group has shown that radiomic metrics correlate with the response to chemotherapy and predict clinical endpoints such as progression-free survival and overall survival. In addition, it is well known that certain protein signatures of the tumour are associated with outcome. We aim to build risk models that integrate both data streams to stratify patients according to risk for progression, thereby allowing for tailored therapy. In an exploratory analysis, we showed for the first time the association between proteomics and CT-based qualitative and texture features in ovarian cancer.

Figure 2. Schematic workflow for the integration of proteomics with CT imaging traits and radiomics in patients with ovarian cancer (adapted from Beer et al. 2020 European Radiology)
Co-registration of imaging and tumour tissue using 3-dimensional moulds
We have demonstrated proof-of-concept of the use of custom-printed 3D moulds to allow accurate spatial co-registration between tumour tissue samples and their corresponding imaging features in both ovarian (published in JCO Precision Oncology, 2019) and kidney cancer (JCO Clinical Informatics, 2020). We are collaborating with genomics and computational groups to expand on this technique, in order to further investigate the relationship between cellular and molecular features of these cancers with imaging and radiomic features. This will improve our ability to use non-invasive imaging features to understand how individual tumours respond to standard treatment as well as novel therapeutic approaches in clinical trials.
Reference:
https://ascopubs.org/doi/full/10.1200/PO.18.00410
https://ascopubs.org/doi/full/10.1200/CCI.20.00026
Hyperpolarised Carbon-13 MRI
Our group is assessing the feasibility of multiparametric multinuclear MRI (including hyperpolarised 13C MRI) for assessing early and heterogeneous response to treatment in NACT setting. The metabolic imaging might be able to assess treatment response earlier compared to clinically used techniques. It may also help detect heterogeneous treatment response, often a cause of treatment failure.
Targeted Molecular Imaging of CA125 in High Grade Serous Ovarian Cancer
High grade serous ovarian cancer cells over express the Cancer Antigen 125 (CA125), that is used clinically as a tumour marker measured in the blood serum. Negative serum CA-125 assays do not always indicate absence of recurrent ovarian malignancy, while elevation of serum CA125 levels can also be observed in non-gynaecologic malignancies or benign conditions. Our team is performing a physiological study to assess CA125 expression and heterogeneity in patients with ovarian cancer. This study will establish the first application of immuno-PET imaging in Cambridge, creating a platform for the wider application of this powerful technique for tracking and quantification of mAbs in vivo.
Renal Cancer
Imaging in renal cancer is another area of active research, which is pursued, both in independent radiological research and as part of clinical trials. Being one of the medical services with the highest volume of patients with kidney cancer undergoing surgical or oncological treatment, Addenbrooke’s Hospital and the University of Cambridge are ideally placed for renal cancer research.
Far reaching alterations in cellular energy metabolism are a hallmark of renal cancer. Hyperpolarised [1-13C]-pyruvate MRI is a non-ionising, quantitative imaging technique which allows our group to investigate the core of these metabolic changes. Our interest covers both clinical and translational aspects of this technique.
While hyperpolarised [1-13C]-pyruvate MRI enables unknown insights into patient tumour metabolism, it is currently reserved for a small number of centres worldwide which have the relevant infrastructure and expertise. We also investigate the ability of radiomics of standard- of- care imaging in renal cell carcinoma (RCC) to support clinical decision making. Radiomics models may help address clinical questions which are difficult to assess with subjective image interpretation. The main questions, which are of particular interest to our group, are the imaging-based prediction of tumour subtype, aggressiveness and early response to therapy.
One aspect we are working on is the analysis of the robustness of RCC radiomic features extracted from CT images of heterogeneous cohorts, and what is the impact of different acquisition parameters (e.g. vendor, reconstruction kernel), as well as perturbations of regions of interest, on the values of these features. These studies can also be connected to measure their expected effect in Radiomics models such as those for tumour classification. Imaging plays an important role in oncological treatment response assessment for many cancer types. With the availability of an increasing number of targeted and non-targeted anti-cancer drugs for RCC, the rapid detection of treatment response and relapse become of increasing clinical relevance. In collaboration with colleagues from the Departments of Urology and Oncology, we investigate how multiparametric MRI can be utilised for treatment stratification in current and future clinical trials.
Pancreatic cancer
Pancreatic cancer (PCa) is the 4th leading cause of cancer-related deaths, with approximately equal rates of annual incidence and mortality. It is estimated that it will become the second leading cause of cancer-related deaths by 2030. There is still a critical unmet need for better ways to non-invasively assess PCa at the different stages of diagnosis. Finding new imaging approaches to better clinically evaluate pancreatic cancer is a field of active research both by radiology led research and also as part of ongoing Oncology clinical trials.
Liver cancer
Liver cancer is the 6th most common cancer in the world, with hepatocellular carcinoma (HCC) being the most common type. Our research is focused on HCC in patients to whom liver transplant was offered as a curative treatment. As one of the main transplant units in the country, the Addenbrooke’s Hospital and the University of Cambridge provide unique research opportunities in this field. We work in close collaboration with the clinical Hepatology team and the Transplant Unit, with most projects using a quantitative imaging approach.
A first study on HCC we conducted, entitled “Robustness of radiomic features in CT images with different slice thickness, investigated the radiomic feature robustness of liver cancer comparing liver tumour and muscle” has been published in Scientific Reports (2021). In this study, we analysed the robustness of CT radiomic features extracted from images of the same tumours with different reconstructed slice thickness, and we provided guidelines for radiomics studies in heterogeneous cohorts.

Illustration of a 2D CT slice of a case used to compare the values of the radiomic features with reconstructed slice thickness of 2mm (left) and 5mm (right) for HCC tumour (red) and muscle tissue (blue).

Example of one radiomic feature (first order Energy) values for 5mm vs 2mm before (left) and after (right) correcting its volume dependency. Adapted from Nature Scientific Reports 2021
Metastatic Breast Cancer
Advanced breast cancer frequently metastasises to the bones. Computed tomography (CT) is the most commonly used imaging technique to assess treatment response in these patients. With metastases frequently presenting as a mixture of lytic and sclerotic metastases at diagnosis, it can be very challenging to distinguish between disease progression and response to treatment if only sclerotic lesions increase over time. We are currently working on the development of quantitative methods for the monitoring of metastatic disease and are collaborating with other groups in Cambridge working on molecular oncology to correlate our imaging findings in metastatic breast cancer with liquid biopsies.
From a clinical perspective, deformable image registration of longitudinal datasets is a necessary step towards automated quantitative and objective response assessment. Together with the Centrum Wiskunde & Informatica, Amsterdam, The Netherlands, the Department of Biomedical Engineering, Eindhoven University of Technology, The Netherlands, and the Department of Applied Mathematics and Theoretical Physics, University of Cambridge, UK, we devised a novel deep learning based deformable approach for abdominal CT scans. The developed method achieves accurate registration results and is two orders of magnitude faster than classic iterative registration methods, thus allowing for real-time registration.
NCITA Imaging Repository
Prof Sala is the Cambridge Principal Investigator of the CRUK funded National Cancer Imaging Translational Accelerator (NCITA) award. She is also the chair of its Repository Unit, whose aim is to build a common image data repository and establish a fast and robust pipeline across seven different UK institutions (University of Cambridge, ICR, University College London, Oxford, Manchester, King’s College London and Imperial College London). This collaboration aims to expand our capabilities in cancer research by building a unified infrastructure for discovery, validation and adoption of cancer imaging biomarkers into clinical trials and the NHS.
Our group has built one of the NCITA repository nodes to store de-identified images and related metadata, that will be connected with other institutions via an NCITA XNAT federation. Our repository is a web-based database using the XNAT software (developed by the Neuroinformatics Research Group at the Washington University), and it is deployed using cloud services from the High Performance Computing of the University of Cambridge. It already has over 50 users working on 30 active projects containing over 30k imaging sessions of CT, MR, MG, X-rays for cancer and COVID-19 research.
In collaboration with the developers of the OHIF viewer plugin to XNAT (ICR, Imperial College) and NVIDIA developers, we have successfully integrated the NVIDIA Clara tools into the viewer plugin, allowing for DL-based automated segmentation models to be run directly from the viewer in XNAT.
COVID-19
AIX-COVNET Collaboration:
The group are key members of the AIX-COVNET collaboration, a multidisciplinary team of radiologists and other clinicians working alongside image-processing and machine learning specialists to develop AI tools to support front-line practitioners in the COVID-19 pandemic and beyond. This growing collaboration includes partners from across the globe in academia, healthcare and industry.
By utilising multi-source data, including chest X-rays, CTs, automated coronary calcium scoring and electronic patient records, the collaboration expects to increase accuracy of diagnosis and prognostication by developing AI solutions that allow for more accurate triage and personalised treatment regimes.

Annotated examples of COVID-19 scans: (a.) a chest radiograph demonstrating hazy ground glass opacification throughout both lungs, with more opaque and localised consolidation (outlined in orange) in the lower areas; (b.) a CT scan of a patient with moderately severe COVID-19 showing ground glass opacification (green) and consolidation (orange); and (c.) a CT scan of a patient with highly severe COVID-19 showing a crazy paving pattern within the ground glass opacification, which is a result of superimposed inflammation of lung tissue and gets its name from its similar appearance to a pathway made up of disorganised paving stones (annotations courtesy of Dr Ian Selby; public domain photograph of paving stones, credit to Liam Quin).

Our collaboration has identified five promising applications of machine learning in the COVID-19 pandemic. The AIX-COVNET collaboration’s vision for a multi-stream model incorporates multiple imaging segmentation methods (a., b. and c.) with flow cytometry (d.) and clinical data. (a. A saliency map on a radiograph, b. Segmented parenchymal disease on a CT scan, c. Segmentation of calcified atherosclerotic disease, d. A projection of a flow cytometry scatter plot of side-scattered light (SSC) versus side fluorescence light (SFL), giving insight into cell structures (analysis performed on a Sysmex UK (1) flow cytometer). (Radiological images from NCCID/NHSX (2). Used with permission.) (Image originally created by Dr Ian Selby for our piece in RSNA Radiology AI).
To inform development, the collaboration published a systematic review of the literature. None of the currently published models are of potential clinical use due to a number of systemic issues, including methodological flaws and/or underlying biases. Based upon the knowledge gained from the review, the team wrote an opinion piece to discuss opportunities for machine learning to assist front-line workers during the COVID-19 pandemic and the steps we take now will leave us better in the future.
Many of the team have been working on the project in addition to their usual duties and everyone involved is quick to stress the importance of the unique working relationship between partners. The tools will first be deployed at Addenbrooke’s and later released open source. Prof. Sala is a co-lead this project for University of Cambridge together with Prof. Carola Bibiane Schönlieb from Department of Applied Mathematics and Theoretical Physics.
AIX-COVNET: Latest News & Results | Contact or Join Us
Media coverage of the AIX-COVNET Collaboration: Addenbrooke’s Press Release | gov.uk | NHSx | inews | Cambridge Independent
DRAGON Consortium:
The AIX-COVNET collaboration is a member of the DRAGON consortium, which officially began in October 2020. The project is funded by the Innovative Medicines Initiative (IMI) and consists of high-tech small and medium sized enterprises (SMEs), academic research institutes, biotechnology and pharmacological partners from across Europe.
A key goal of the research project is to apply artificial intelligence and machine learning to deliver a decision support system for precise coronavirus diagnosis and prognosis. Existing as well as new imaging and clinical data will be used to develop and validate new AI tools for diagnosis and prognosis and integrate them into clinical decision-making. Furthermore, the consortium aim to ensure preparedness should we be faced with a similar pandemic in the future.
DRAGON have already developed a prototype app to inform health professionals about existing predictive models for COVID-19 related risk assessment, whilst work is also progressing on development of diagnostic and prognostic tools for thoracic CT.
Press Release: ERS
Immuno-PET/MR imaging
We are pleased to be collaborating on multi-centre trial investigating CD8-PET/MR imaging in patient with metastatic cancer undergoing treatment with checkpoint inhibition, funded by the Immune Image Consortium supported by the European Commission (https://www.immune-image.eu).
The project aims to demonstrate and validate the use of PET/MR with a 89Zr-labelled anti-CD8 minibody, to serially visualize/track the presence of CD8 expressing T-lymphocytes in cancer patients undergoing treatment in a dedicated clinical trial with a novel Immune Checkpoint Inhibition agent. We aim to confirm safety of new immunotracers and to demonstrate their clinical utility in combination with anatomic imaging (MRI or CT) for patients with cancer. Clinical application of the novel immunotracers will focus on diagnostics and patient stratification based on immune status as well as prediction of response or long-term outcome of therapeutic interventions, early detection of immune system activation, target engagement within the tissue of interest, both systemically and locally.
We are part of a multi-centre consortium with partner institutions in Groningen (UMCG), Amsterdam (VUmc), and Barcelona (VHIO/ VHIR), and locally in collaboration with Prof. Brindle’s lab in CRUK –CI. Prof. Sala and her team will bring to the consortium the a wealth of expertise in the field of Radiomics to execute this research successfully, and will be working closely with Dr. Luigi Aloj (Nuclear Medicine) on this project
Technologies / Methods
- Advanced MRI (DKI-DWI, IVIM-DWI, DCE-analysis, MT)
- Radiomics
- Metabolic imaging (Hyperpolarised Carbon-13 MRI, FDG-PET/MRI, FDG-PET/CT)
- Ultrasound guided fusion biopsies
- Data Integration (Radiomics, Proteomics, Genomics, ctDNA, clinical data)
Recent Awards
- Evis Sala
- Prestigious award by the British Institute of Radiology/Canon Mayneord Memorial, November 2020
- RSNA Honored Educator Award of 2020, June 2020
- Appointed Senior Consulting Editor for Radiology: Artificial Intelligence, 2020
Funding
- Cambridge Cancer Center (CRUK)
- Mark Foundation for Cancer Research
- Wellcome Trust
- National Cancer Imaging Translational Accelerator (NCITA)
- Armstrong Foundation
- GlaxoSmithKline
- AstraZeneca (cancer trials)
- Amazon HealthLake
- OncoQuest (radiopharmaceuticals)
Collaborators
- Apollo
- Canon Medical Systems
- Centrum Wiskunde & Informatica, Amsterdam, The Netherlands – Computational Imaging Group
- CRUK-CI groups:
- DAMTP – Cambridge Image Analysis group
- GE Healthcare
- MSKCC – The Jason Lewis Lab
- NCITA (Twitter)
- TCGA
- TCIA
- VUMC
Outreach
- Cambridge Science Festival 2020 (postponed until 2021)
Initiatives
- Cambridge Imaging Festival 2019
- Workshop: Collaborative Approaches to Cancer Imaging. Applying known methods in new ways
- Cambridge Imaging Festival 2020 (postponed until May 2021)
Latest Publications (01/01/2018 – Present)
- Wang G, Liu X, Shen J, Deng R, Yang J, Ye L, Wu X, Zhou Z, Zhang X, Wang C, Li Z, Liang W, Zheng L, Sang Y, Yu T, Gao M, Wang J, Yang Z, Cai H, Lu G, Zhang L, Yang L, Xu W, Wang W, Olevera A, Ziyar I, Zhang C, Li O, Liao W, Liu J, Chen W, Chen W, Shi J, Zhang L, Yan Z, Zhou X, Lin G, Cao G, Lau LL, Mo L, Roberts M, Sala E, Schönlieb CB, Fok M, Lau JYN, He J, Li W, Chen T, Zhang K, Lin T. A deep-learning pipeline for the diagnosis and discrimination of viral, non-viral and COVID-19 pneumonia from chest X-ray images. Nature Biomedical Engineering 2021 (in press).
- Le E, Rundo L, Evans NR, Tarkin J M, Chowdhury MM, Coughlin P, Pavey H, Wall C, Zaccagna F, Gallagher FA, Huang Y, Sriranjan R, Le A, Weir-Mccall JR, Roberts M, Gilbert F, Waburton EA, Schonlieb CB, Sala E, Rudd JHF. Assessing robustness of carotid artery CT angiography radiomics in the identification of culprit lesions in cerebrovascular events. Nature Scientific Reports, 2021, DOI:https://doi.org/10.1038/s41598-021-82760-w
- Wibmer AG, Chaim J, Lakhman Y, Lefkowitz R, Nincevic J, Nikolovski I, Sala E, Gonen M, Carlsson SV, Fine SW, Zelefsky M, Scardino P, Hricak H, Vargas AH. Oncologic outcomes after localized prostate cancer treatment: associations with pretreatment prostate magnetic resonance imaging findings. J Urology (in press).
- Sahin H, Panico C, Ursprung S, Simeon V, Chiodini P, Frary A, Carmo B, Smith J, Freeman S, Jimenez-Linan M, Bolton H, Haldar K, Ang JE, Reinhold C, Sala E*, Addley HA*. Non-contrast MRI can accurately characterize adnexal masses: a retrospective study. European Radiology 2021 (in press).
- Rundo L, Tangherloni A, Cazzaniga P, Mitri M, Galimberti S, Woitek R, Sala E, Mauri G, Nobile MS. A CUDA‑powered method for the feature extraction and unsupervised analysis of medical images. The Journal of Supercomputing 2021, https://doi.org/10.1007/s11227-020-03565-8.
- Veeraraghavan H*, Vargas HA*, Sanchez A, Micco M, Mema E, Lakhman Y, Crispin-Ortuzar M, Huang EP, Levine DA, Grisham RN, Abu-Rustum N, Deasy JO, Snyder A, Miller ML, Brenton JD*, Sala E*. Integrated multi-tumor radiogenomic marker of outcomes in patients with high serous ovarian carcinoma. Cancers 2020, 12(11), 3403; https://doi.org/10.3390/cancers12113403.
- Gill A, Rundo L, Wan J, Lau D, Zawaideh J, Woitek R, Zaccagna F, Beer L, Gale D, Sala E, Couturier D, Corrie P, Rosenfeld N, Gallagher F. Correlating radiomic features of heterogeneity on CT with circulating tumor DNA in metastatic melanoma. Cancers 2020,12(12), 3493; https://doi.org/10.3390/cancers12123493.
- Beer L, Martin-Gonzales P, Delgado-Ortet M, Reinus M, Rundo L, Woitek R, Ursprung S, Escudero L, Sahin H, Lawton T, Phadke G, Davey S, Funingana G, Ang JE, Jimenez-Lina M, Markowetz F, Nguyen NQ, Benton JD, Addley H, Sala E. Ultrasound-guided targeted biopsies of CT based radiomic tumour habitats: Technical Development and Initial Experience in metastatic ovarian cancer. Eur Radiol, 2020, https://doi.org/10.1007/s00330-020-07560-8.
- Serrao EM, Kessler DA, Carmo C, Beer L, Brindle KM, Buonincontri G, Gallagher FA, Gilbert FJ, Godfrey E, Graves MJ, McLean M, Sala E, Schulte RF, Kaggie J. Magnetic Resonance fingerprinting of the pancreas at 1.5T and 3.0T. Scientific Reports, 2020, https://doi.org/10.1038/s41598-020-74462-6.
- Han C, Rundo L, Murao K, Milacski Z.Á, Umemoto K, Sala E, Nakayama H, Satoh S. GAN-based multiple adjacent brain MRI slice reconstruction for unsupervised Alzheimer’s disease diagnosis”. In: Proc. Computational Intelligence methods for Bioinformatics and Biostatistics (CIBB) 2019, LNBI Springer Series, CIBB 2019, LNBI 12313, pp. 44–54, 2020. https://doi.org/10.1007/978-3-030-63061-4_5..
- Crispin-Ortuzar M, Gehrung M, Ursprung S, Gill AB, Warren AY, Beer L, Gallagher FA, Mitchell TJ, Mendichovszky IA, Priest AN, Stewart GD*, Sala E*, Markowetz F*. Three-dimensional printed molds for image-guided surgical biopsies: an open source computational platform. JCO Clin Cancer Inform 2020; 4:736-748.
- Caglic I, Sushentsev N, Gnanapragasam VJ, Sala E, Shaida N, Koo BC, Kozlov V, Warren AY, Kastner C, Barrett B. MRI-derived PRECISE scores for predicting pathologically-confirmed radiological progression in prostate cancer patients on active surveillance. Eur Radiol, 2020; https://doi.org/10.17863/CAM.55752.
- Jiménez-Sánchez A, Cybulska P, LaVigne K, Koplev S, Cast O, Couturier DL, Memon D, Selenica P, Nikolovski I, Mazaheri Y, Bykov Y, Geyer FC, Macintyre G, Gavarró LM, Drews RM, Gill MB, Papanastasiou AD, Sosa ER, Soslow RA, Walther T, Shen R, Chi DS, Park KJ, Hollmann T, Reis-Filho JS, Markowetz F, Beltrao P, Vargas HA, Zamarin D, Brenton JD, Snyder A, Weigelt B, Sala E, Miller M. Unraveling tumor-immune heterogeneity in advanced ovarian cancer uncovers immunogenic effect of chemotherapy. Nat Genet, 2020; Jun;52(6):582-593. doi: 1038/s41588-020-0630-5.
- Woitek R, McLean MA, Gill AB, Grist JT, Provenzano E, Patterson AJ, Ursprung S, Torheim T, Zaccagna F, Locke M, Laurent M-C, Hilborne S, Frary A, Beer L, Rundo L, Patterson I, Slough R, Kane J, Biggs H, Harrison E, Lanz T, Basu B, Baird R, Sala E, Graves MJ, Gilbert FJ, Abraham JE, Caldas C, Brindle KM, Gallagher FA. Hyperpolarized 13c-MRI of tumor metabolism demonstrates early metabolic response to neoadjuvant chemotherapy in breast cancer. Radiology: Imaging Cancer, 2020; https://doi.org/10.1148/rycan.2020200017.