Ovarian Cancer

Deep learning-based segmentation of multisite disease in ovarian cancer

Buddenkotte, T., et al., European Radiology Experimental, Dec 2023

Abstract

Read Publication

Lesion-specific 3D-printed moulds for image-guided tissue multi-sampling of ovarian tumours: A prospective pilot study

Delgardo-Ortet, M., et al, Frontiers in Oncology, Feb 2023

Background: High-Grade Serous Ovarian Carcinoma (HGSOC) is the most prevalent and lethal subtype of ovarian cancer, but has a paucity of clinically-actionable biomarkers due to high degrees of multi-level heterogeneity. Radiogenomics markers have the potential to improve prediction of patient outcome and treatment response, but require accurate multimodal spatial registration between radiological imaging and histopathological tissue samples. Previously published co-registration work has not taken into account the anatomical, biological and clinical diversity of ovarian tumours.

Methods: In this work, we developed a research pathway and an automated computational pipeline to produce lesion-specific three-dimensional (3D) printed moulds based on preoperative cross-sectional CT or MRI of pelvic lesions. Moulds were designed to allow tumour slicing in the anatomical axial plane to facilitate detailed spatial correlation of imaging and tissue-derived data. Code and design adaptations were made following each pilot case through an iterative refinement process.

Results: Five patients with confirmed or suspected HGSOC who underwent debulking surgery between April and December 2021 were included in this prospective study. Tumour moulds were designed and 3D-printed for seven pelvic lesions, covering a range of tumour volumes (7 to 133 cm³) and compositions (cystic and solid proportions). The pilot cases informed innovations to improve specimen and subsequent slice orientation, through the use of 3D-printed tumour replicas and incorporation of a slice orientation slit in the mould design, respectively. The overall research pathway was compatible with implementation within the clinically determined timeframe and treatment pathway for each case, involving multidisciplinary clinical professionals from Radiology, Surgery, Oncology and Histopathology Departments.

Conclusions: We developed and refined a computational pipeline that can model lesion-specific 3D-printed moulds from preoperative imaging for a variety of pelvic tumours. This framework can be used to guide comprehensive multi-sampling of tumour resection specimens.

Read Publication

Calibrating ensembles for scalable uncertainty quantification in deep learning-based medical image segmentation

Buddenkotte, T., et al., Computers in Biology and Medicine, Sept 2023

Uncertainty quantification in automated image analysis is highly desired in many applications. Typically, machine learning models in classification or segmentation are only developed to provide binary answers; however, quantifying the uncertainty of the models can play a critical role for example in active learning or machine human interaction. Uncertainty quantification is especially difficult when using deep learning-based models, which are the state-of-the-art in many imaging applications. The current uncertainty quantification approaches do not scale well in high-dimensional real-world problems. Scalable solutions often rely on classical techniques, such as dropout, during inference or training ensembles of identical models with different random seeds to obtain a posterior distribution. In this paper, we present the following contributions. First, we show that the classical approaches fail to approximate the classification probability. Second, we propose a scalable and intuitive framework for uncertainty quantification in medical image segmentation that yields measurements that approximate the classification probability. Third, we suggest the usage of k-fold cross-validation to overcome the need for held out calibration data. Lastly, we motivate the adoption of our method in active learning, creating pseudo-labels to learn from unlabeled images and human–machine collaboration.

Read Publication

Integrated radiogenomics models predict response to neoadjuvant chemotherapy in high grade serous ovarian cancer

Crispin-Ortuzar, M., et al Nature Communications, Oct 2023

High grade serous ovarian carcinoma (HGSOC) is a highly heterogeneous disease that typically presents at an advanced, metastatic state. The multi-scale complexity of HGSOC is a major obstacle to predicting response to neoadjuvant chemotherapy (NACT) and understanding critical determinants of response. Here we present a framework to predict the response of HGSOC patients to NACT integrating baseline clinical, blood-based, and radiomic biomarkers extracted from all primary and metastatic lesions. We use an ensemble machine learning model trained to predict the change in total disease volume using data obtained at diagnosis (n = 72). The model is validated in an internal hold-out cohort (n = 20) and an independent external patient cohort (n = 42). In the external cohort the integrated radiomics model reduces the prediction error by 8% with respect to the clinical model, achieving an AUC of 0.78 for RECIST 1.1 classification compared to 0.47 for the clinical model. Our results emphasize the value of including radiomics data in integrative models of treatment response and provide methods for developing new biomarker-based clinical trials of NACT in HGSOC.

Read Publication

Multiparameter single-cell proteomic technologies give new insights into the biology of ovarian tumors

Funingana, I. G., et al, Seminars in Immunopathology, Jan 23

High-grade serous ovarian cancer (HGSOC) is the most lethal gynecological malignancy. Its diagnosis at advanced stage compounded with its excessive genomic and cellular heterogeneity make curative treatment challenging. Two critical therapeutic challenges to overcome are carboplatin resistance and lack of response to immunotherapy. Carboplatin resistance results from diverse cell autonomous mechanisms which operate in different combinations within and across tumors. The lack of response to immunotherapy is highly likely to be related to an immunosuppressive HGSOC tumor microenvironment which overrides any clinical benefit. Results from a number of studies, mainly using transcriptomics, indicate that the immune tumor microenvironment (iTME) plays a role in carboplatin response. However, in patients receiving treatment, the exact mechanistic details are unclear. During the past decade, multiplex single-cell proteomic technologies have come to the forefront of biomedical research. Mass cytometry or cytometry by time-of-flight, measures up to 60 parameters in single cells that are in suspension. Multiplex cellular imaging technologies allow simultaneous measurement of up to 60 proteins in single cells with spatial resolution and interrogation of cell–cell interactions. This review suggests that functional interplay between cell autonomous responses to carboplatin and the HGSOC immune tumor microenvironment could be clarified through the application of multiplex single-cell proteomic technologies. We conclude that for better clinical care, multiplex single-cell proteomic technologies could be an integral component of multimodal biomarker development that also includes genomics and radiomics. Collection of matched samples from patients before and on treatment will be critical to the success of these efforts.

Read publication

Automatic whole tumour burden segmentation and tissue-specific sub segmentation

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 Computed Tomography (CT) 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. A prototype deep learning automated segmentation method for ovarian cancer has been developed relying upon cutting-edge deep neural network architectures. The main goal is to develop an automated platform to track tumour dynamics. The computational end-to-end pipeline will include CT image reconstruction, segmentation and radiomics analyses. These steps will be jointly performed and the overall analysis is independent of scanner manufacturers and acquisition protocols. Moreover, the first approach based on unsupervised techniques has been proposed to provide tissue-specific sub-segmentation and quantify solid, hypo-dense (i.e. cystic and necrotic regions) and hyper-dense (i.e., calcifications) components on CT, which are commonly found in ovarian cancer lesions. Exploiting these novel techniques, we aim to provide robust tools for future radiomic biomarker validation.

Further information on the Tissue-Specific Sub Segmentation project.

Computational Biomedicine - 3D tumour mould

Molecular stratification of patients based on integration of data from heterogeneous habitats derived from imaging and digital pathology:

To address intra-tumoural heterogeneity, we have developed a novel method to link imaging (MRI) data to digital pathology data by developing an image guided 3D-printed tumour mould which allows us to establish and validate the relationship between imaging habitats and tissue-based measurements.

Patients potentially benefit in a number of ways:

• Improved representation of heterogeneous tumour profile based on multiple targeted samples
• Improved stratification and informed decision for precision therapeutics based on actionable targets detected through targeted sampling – before and after chemotherapy treatment
  • Combination of immunotherapy and genomically targeted therapies

Further information on the 3D Tumour Mould project.

A further publication:

Radiogenomics Analysis of Intratumor Heterogeneity in a Patient With High-Grade Serous Ovarian Cancer. Weigelt, B, et al., JCO Precision Oncology, June 2019.

In this study we use lesion-specific three-dimensional (3D) molds for phenotypic image-guided tumor sampling to ensure spatial colocation of imaging, histology, and genomic data, critical for understanding tumor biology. Phenotypic imaging maps of heterogeneity (ie, imaging habitats) of two HGSOC sites were obtained by combining perfusion, diffusion, and metabolic maps derived from multiparametric imaging. We evaluated if this phenotypic imaging-based heterogeneity reflects the underlying histologic and/or genetic heterogeneity of the tumor.

Read Publication

Integrating radiomics and ctDNA data

Patients with findings suspicious for ovarian cancer or with a definite diagnosis of the disease usually undergo computed tomography (CT) scans as their standard-of-care imaging to assess the extent of the disease. We are collaborating with molecular oncology and computational groups to extract quantitative data from these images (radiomics) and combine them with ctDNA isolated from blood samples of these patients to develop tools that predict response to chemotherapy as well as survival in these patients. 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.

Further information on the Integrating Radiomics and ctDNA data

Integration of Proteomics and Radiomics

In collaboration with colleagues from the Walter Reed National Military Medical Center, the Inova Schar Cancer Institute and Frederick National Laboratory for Cancer Research the Radiomics group is working on the integration of proteomic and radiomic data into clinical decision models for patients with ovarian cancer. Over the last few years, the 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. The group aims 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, they showed for the first time the association between proteomics and CT-based qualitative and texture features in ovarian cancer.

Integration of proteomics with CT-based qualitative and radiomic features in high-grade serous ovarian cancer patients: an exploratory analysis. Beer, L., et al., European Radiology, April 2020.

Summary: This study investigated the association between CT imaging traits and texture metrics with proteomic data in patients with high-grade serous ovarian cancer (HGSOC).  It provides the first insights into the potential associations between standard-of-care CT imaging traits and texture measures of intra- and inter-site heterogeneity, and the abundance of several proteins.

Further information on the Integration of Proteomics and Radiomics project.

Ovarian Hyperpolarised 13C MRI

The imaging group is assessing the feasibility of multiparametric multinuclear MRI (including hyperpolarised 13C MRI) for testing early and heterogeneous response to treatment in NACT setting. The metabolic imaging might be able to indicate treatment response earlier compared to clinically used techniques. It may also help detect heterogeneous treatment response, often a cause of treatment failure.

Further information on the Hyperpolarised Carbon-13 MRI project.

Real time habitat guided multimodal fusion biopsies

In collaboration with Canon® Medical Systems, the radiomics group has recently developed and is currently 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 achieved overlaying texture-based clustering maps onto diagnostic CT images and PET/MRI habitat maps onto MRI images, respectively and then has fused them with the US during biopsies. This technique will enable biological profiling and longitudinal tracking of tumour habitats in clinical settings.

Martin-Gonzalez, P. et al., Integrative radiogenomics for virtual biopsy and treatment monitoring in ovarian cancer. Insights Into Imaging, 17 Aug 2020.

In this review, we describe how these challenges might be overcome by integrating quantitative features extracted from medical imaging with the analysis of paired genomic profiles, a combined approach called radiogenomics, to generate virtual biopsies. Radiomic studies have been used to model different imaging phenotypes, and some radiomic signatures have been associated with paired molecular profiles to monitor spatiotemporal changes in the heterogeneity of tumours. We describe different strategies to integrate radiogenomic information in a global and local manner, the latter by targeted sampling of tumour habitats, defined as regions with distinct radiomic phenotypes.

Further information on the Real Time Habitat Guided Multimodal Fusion Biopsies project.

Read Publication

Targeted Molecular Imaging of CA125

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. The Radiomics 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.

Further information on the Targeted Molecular Imaging of CA125 project.

Ultrasound-guided targeted biopsies of CT-based radiomic tumour habitats: technical development and initial experience in metastatic ovarian cancer

Beer, L., Martin-Gonzalez, P., et al, European Radiology, 14 Dec 2020.

We developed a precision tissue sampling technique that uses radiomic habitats to guide in vivo biopsies using CT/US fusion and that can be seamlessly integrated in the clinical routine for patients with HGSOC.

Six patients with suspected HGSOC scheduled for US-guided biopsy before starting neoadjuvant chemotherapy were included in this prospective study from September 2019 to February 2020. The tumour segmentation was performed manually on the pre-biopsy contrast-enhanced CT scan. Spatial radiomic maps were used to identify tumour areas with similar or distinct radiomic patterns, and tumour habitats were identified using the Gaussian mixture modelling. CT images with superimposed habitat maps were co-registered with US images by means of a landmark-based rigid registration method for US-guided targeted biopsies. The dice similarity coefficient (DSC) was used to assess the tumour-specific CT/US fusion accuracy.

Further information on the Real Time Habitat Guided Multimodal Fusion Biopsies project.

Read Publication