Deep learning enables the differentiation between early and late stages of hip avascular necrosis.

OBJECTIVES: To develop a deep learning methodology that distinguishes early from late stages of avascular necrosis of the hip (AVN) to determine treatment decisions. METHODS: Three convolutional neural networks (CNNs) VGG-16, Inception ResnetV2, InceptionV3 were trained with transfer learning (ImageNet) and finetuned with a retrospectively collected cohort of (n = 104) MRI examinations of AVN patients, to differentiate between early (ARCO 1-2) and late (ARCO 3-4) stages. A consensus CNN ensemble decision was recorded as the agreement of at least two CNNs. CNN and ensemble performance was benchmarked on an independent cohort of 49 patients from another country and was compared to the performance of two MSK radiologists. CNN performance was expressed with areas under the curve (AUC), the respective 95% confidence intervals (CIs) and precision, and recall and f1-scores. AUCs were compared with DeLong's test. RESULTS: On internal testing, Inception-ResnetV2 achieved the highest individual performance with an AUC of 99.7% (95%CI 99-100%), followed by InceptionV3 and VGG-16 with AUCs of 99.3% (95%CI 98.4-100%) and 97.3% (95%CI 95.5-99.2%) respectively. The CNN ensemble the same AUCs Inception ResnetV2. On external validation, model performance dropped with VGG-16 achieving the highest individual AUC of 78.9% (95%CI 51.6-79.6%) The best external performance was achieved by the model ensemble with an AUC of 85.5% (95%CI 72.2-93.9%). No significant difference was found between the CNN ensemble and expert MSK radiologists (p = 0.22 and 0.092 respectively). CONCLUSION: An externally validated CNN ensemble accurately distinguishes between the early and late stages of AVN and has comparable performance to expert MSK radiologists. CLINICAL RELEVANCE STATEMENT: This paper introduces the use of deep learning for the differentiation between early and late avascular necrosis of the hip, assisting in a complex clinical decision that can determine the choice between conservative and surgical treatment. KEY POINTS: • A convolutional neural network ensemble achieved excellent performance in distinguishing between early and late avascular necrosis. • The performance of the deep learning method was similar to the performance of expert readers.

Radiotranscriptomics of non-small cell lung carcinoma for assessing high-level clinical outcomes using a machine learning-derived multi-modal signature.

BACKGROUND: Multi-omics research has the potential to holistically capture intra-tumor variability, thereby improving therapeutic decisions by incorporating the key principles of precision medicine. The purpose of this study is to identify a robust method of integrating features from different sources, such as imaging, transcriptomics, and clinical data, to predict the survival and therapy response of non-small cell lung cancer patients. METHODS: 2996 radiomics, 5268 transcriptomics, and 8 clinical features were extracted from the NSCLC Radiogenomics dataset. Radiomics and deep features were calculated based on the volume of interest in pre-treatment, routine CT examinations, and then combined with RNA-seq and clinical data. Several machine learning classifiers were used to perform survival analysis and assess the patient's response to adjuvant chemotherapy. The proposed analysis was evaluated on an unseen testing set in a k-fold cross-validation scheme. Score- and concatenation-based multi-omics were used as feature integration techniques. RESULTS: Six radiomics (elongation, cluster shade, entropy, variance, gray-level non-uniformity, and maximal correlation coefficient), six deep features (NasNet-based activations), and three transcriptomics (OTUD3, SUCGL2, and RQCD1) were found to be significant for therapy response. The examined score-based multi-omic improved the AUC up to 0.10 on the unseen testing set (0.74 ± 0.06) and the balance between sensitivity and specificity for predicting therapy response for 106 patients, resulting in less biased models and improving upon the either highly sensitive or highly specific single-source models. Six radiomics (kurtosis, GLRLM- and GLSZM-based non-uniformity from images with no filtering, biorthogonal, and daubechies wavelets), seven deep features (ResNet-based activations), and seven transcriptomics (ELP3, ZZZ3, PGRMC2, TRAK1, ATIC, USP7, and PNPLA2) were found to be significant for the survival analysis. Accordingly, the survival analysis for 115 patients was also enhanced up to 0.20 by the proposed score-based multi-omics in terms of the C-index (0.79 ± 0.03). CONCLUSIONS: Compared to single-source models, multi-omics integration has the potential to improve prediction performance, increase model stability, and reduce bias for both treatment response and survival analysis.

Prediction of abdominal aortic aneurysm growth by artificial intelligence taking into account clinical, biologic, morphologic, and biomechanical variables.

OBJECTIVES: To develop a prediction model that could risk stratify abdominal aortic aneurysms (AAAs) into high and low growth rate groups, using machine learning algorithms based on variables from different pathophysiological fields. METHODS: A cohort of 40 patients with small AAAs (maximum diameter 32-53 mm) who had at least an initial and a follow-up CT scan (median follow-up 12 months, range 3-36 months) were included. 29 input variables from clinical, biological, morphometric, and biomechanical pathophysiological aspects extracted for predictive modeling. Collected data were used to build two supervised machine learning models. A gradient boosting (XGboost) and a support vector machines (SVM) algorithm were trained with 60% and tested with 40% of the data to predict which AAA would achieve a growth rate higher than the median of our study cohort. Receiver operating characteristics (ROC) curves and areas under the curve (AUC) were used for the evaluation of the developed algorithms. RESULTS: XGboost achieved the highest AUC in predicting high compared to low AAA growth rate with an AUC of 81.2% (95% CI from 61.1 to 100%). SVM achieved the second highest performance with an AUC of 68.8% (95% CI from 46.5 to 91%). Based on the best performing algorithm, variable importance was estimated. Diameter-diameter ratio (maximum diameter/neck diameter), Tortuosity from Renal arteries to aortic bifurcation, and maximum thickness of the intraluminal thrombus were found to be the most important factors for model predictions. Other factors were also found to play a significant but less important role. CONCLUSIONS: A prediction model that can risk stratify AAAs into high and low growth rate groups could be developed by analyzing several factors implicated in the multifactorial pathophysiology of this disease, with the use of machine learning algorithms. Future studies including larger patient cohorts and implementing additional risk markers may aid in the establishment of such methodology during AAA rupture risk estimation.

Personalized Risk Analysis to Improve the Psychological Resilience of Women Undergoing Treatment for Breast Cancer: Development of a Machine Learning-Driven Clinical Decision Support Tool.

BACKGROUND: Health professionals are often faced with the need to identify women at risk of manifesting poor psychological resilience following the diagnosis and treatment of breast cancer. Machine learning algorithms are increasingly used to support clinical decision support (CDS) tools in helping health professionals identify women who are at risk of adverse well-being outcomes and plan customized psychological interventions for women at risk. Clinical flexibility, cross-validated performance accuracy, and model explainability permitting person-specific identification of risk factors are highly desirable features of such tools. OBJECTIVE: This study aimed to develop and cross-validate machine learning models designed to identify breast cancer survivors at risk of poor overall mental health and global quality of life and identify potential targets of personalized psychological interventions according to an extensive set of clinical recommendations. METHODS: A set of 12 alternative models was developed to improve the clinical flexibility of the CDS tool. All models were validated using longitudinal data from a prospective, multicenter clinical pilot at 5 major oncology centers in 4 countries (Italy, Finland, Israel, and Portugal; the Predicting Effective Adaptation to Breast Cancer to Help Women to BOUNCE Back [BOUNCE] project). A total of 706 patients with highly treatable breast cancer were enrolled shortly after diagnosis and before the onset of oncological treatments and were followed up for 18 months. An extensive set of demographic, lifestyle, clinical, psychological, and biological variables measured within 3 months after enrollment served as predictors. Rigorous feature selection isolated key psychological resilience outcomes that could be incorporated into future clinical practice. RESULTS: Balanced random forest classifiers were successful at predicting well-being outcomes, with accuracies ranging between 78% and 82% (for 12-month end points after diagnosis) and between 74% and 83% (for 18-month end points after diagnosis). Explainability and interpretability analyses built on the best-performing models were used to identify potentially modifiable psychological and lifestyle characteristics that, if addressed systematically in the context of personalized psychological interventions, would be most likely to promote resilience for a given patient. CONCLUSIONS: Our results highlight the clinical utility of the BOUNCE modeling approach by focusing on resilience predictors that can be readily available to practicing clinicians at major oncology centers. The BOUNCE CDS tool paves the way for personalized risk assessment methods to identify patients at high risk of adverse well-being outcomes and direct valuable resources toward those most in need of specialized psychological interventions.

Predicting long-term outcomes of ultrasound-guided percutaneous irrigation of calcific tendinopathy with the use of machine learning.

OBJECTIVE: To evaluate the performance of two machine learning models in predicting the long-term complete pain resolution in patients undergoing ultrasound-guided percutaneous irrigation of calcific tendinopathy (US-PICT). MATERIALS AND METHODS: Within a 3-year period, 100 consecutive patients who underwent US-PICT for rotator cuff disease were prospectively enrolled. The location, maximum diameter, and type of each calcification were recorded. The degree of calcium retrieval was graded as complete or incomplete. Shoulder pain was assessed with the visual analogue scale (VAS) at baseline, 1-week, 1-month, and 1-year post-treatment. Measurements related to procedural details, patient, and calcification characteristics were used to construct a machine learning model for the prediction of pain at 1-year post-US-PICT. Two distinct models were built, one including VAS data at 1 week and another additionally including pain data at 1-month post-treatment. Variable importance analysis was performed for the 1-week model. Model performance was evaluated by using receiver operating characteristics (ROC) curves and the respective areas under the curve (AUC). RESULTS: The model exhibited an AUC of 69.2% for the prediction of complete pain resolution at 1 year. The addition of VAS scores at 1 month did not significantly alter the performance of the algorithm. Age and baseline VAS scores were the most important variables for classification performance. CONCLUSION: The presented machine learning model exhibited an AUC of almost 70% in predicting complete pain resolution at 1 year. Pain data at 1 month do not appear to improve the performance of the algorithm.

Patient empowerment for cancer patients through a novel ICT infrastructure.

As a result of recent advances in cancer research and "precision medicine" approaches, i.e. the idea of treating each patient with the right drug at the right time, more and more cancer patients are being cured, or might have to cope with a life with cancer. For many people, cancer survival today means living with a complex and chronic condition. Surviving and living with or beyond cancer requires the long-term management of the disease, leading to a significant need for active rehabilitation of the patients. In this paper, we present a novel methodology employed in the iManageCancer project for cancer patient empowerment in which personal health systems, serious games, psychoemotional monitoring and other novel decision-support tools are combined into an integrated patient empowerment platform. We present in detail the ICT infrastructure developed and our evaluation with the involvement of cancer patients on two sites, a large-scale pilot for adults and a small-scale test for children. The evaluation showed mixed evidences on the improvement of patient empowerment, while ability to cope with cancer, including improvement in mood and resilience to cancer, increased for the participants of the adults' pilot.

Status and Recommendations of Technological and Data-Driven Innovations in Cancer Care: Focus Group Study.

BACKGROUND: The status of the data-driven management of cancer care as well as the challenges, opportunities, and recommendations aimed at accelerating the rate of progress in this field are topics of great interest. Two international workshops, one conducted in June 2019 in Cordoba, Spain, and one in October 2019 in Athens, Greece, were organized by four Horizon 2020 (H2020) European Union (EU)-funded projects: BOUNCE, CATCH ITN, DESIREE, and MyPal. The issues covered included patient engagement, knowledge and data-driven decision support systems, patient journey, rehabilitation, personalized diagnosis, trust, assessment of guidelines, and interoperability of information and communication technology (ICT) platforms. A series of recommendations was provided as the complex landscape of data-driven technical innovation in cancer care was portrayed. OBJECTIVE: This study aims to provide information on the current state of the art of technology and data-driven innovations for the management of cancer care through the work of four EU H2020-funded projects. METHODS: Two international workshops on ICT in the management of cancer care were held, and several topics were identified through discussion among the participants. A focus group was formulated after the second workshop, in which the status of technological and data-driven cancer management as well as the challenges, opportunities, and recommendations in this area were collected and analyzed. RESULTS: Technical and data-driven innovations provide promising tools for the management of cancer care. However, several challenges must be successfully addressed, such as patient engagement, interoperability of ICT-based systems, knowledge management, and trust. This paper analyzes these challenges, which can be opportunities for further research and practical implementation and can provide practical recommendations for future work. CONCLUSIONS: Technology and data-driven innovations are becoming an integral part of cancer care management. In this process, specific challenges need to be addressed, such as increasing trust and engaging the whole stakeholder ecosystem, to fully benefit from these innovations.

MinePath: Mining for Phenotype Differential Sub-paths in Molecular Pathways.

Pathway analysis methodologies couple traditional gene expression analysis with knowledge encoded in established molecular pathway networks, offering a promising approach towards the biological interpretation of phenotype differentiating genes. Early pathway analysis methodologies, named as gene set analysis (GSA), view pathways just as plain lists of genes without taking into account either the underlying pathway network topology or the involved gene regulatory relations. These approaches, even if they achieve computational efficiency and simplicity, consider pathways that involve the same genes as equivalent in terms of their gene enrichment characteristics. Most recent pathway analysis approaches take into account the underlying gene regulatory relations by examining their consistency with gene expression profiles and computing a score for each profile. Even with this approach, assessing and scoring single-relations limits the ability to reveal key gene regulation mechanisms hidden in longer pathway sub-paths. We introduce MinePath, a pathway analysis methodology that addresses and overcomes the aforementioned problems. MinePath facilitates the decomposition of pathways into their constituent sub-paths. Decomposition leads to the transformation of single-relations to complex regulation sub-paths. Regulation sub-paths are then matched with gene expression sample profiles in order to evaluate their functional status and to assess phenotype differential power. Assessment of differential power supports the identification of the most discriminant profiles. In addition, MinePath assess the significance of the pathways as a whole, ranking them by their p-values. Comparison results with state-of-the-art pathway analysis systems are indicative for the soundness and reliability of the MinePath approach. In contrast with many pathway analysis tools, MinePath is a web-based system (www.minepath.org) offering dynamic and rich pathway visualization functionality, with the unique characteristic to color regulatory relations between genes and reveal their phenotype inclination. This unique characteristic makes MinePath a valuable tool for in silico molecular biology experimentation as it serves the biomedical researchers' exploratory needs to reveal and interpret the regulatory mechanisms that underlie and putatively govern the expression of target phenotypes.

The INTEGRATE project: Delivering solutions for efficient multi-centric clinical research and trials.

The objective of the INTEGRATE project (http://www.fp7-integrate.eu/) that has recently concluded successfully was the development of innovative biomedical applications focused on streamlining the execution of clinical research, on enabling multidisciplinary collaboration, on management and large-scale sharing of multi-level heterogeneous datasets, and on the development of new methodologies and of predictive multi-scale models in cancer. In this paper, we present the way the INTEGRATE consortium has approached important challenges such as the integration of multi-scale biomedical data in the context of post-genomic clinical trials, the development of predictive models and the implementation of tools to facilitate the efficient execution of postgenomic multi-centric clinical trials in breast cancer. Furthermore, we provide a number of key "lessons learned" during the process and give directions for further future research and development.

Assessment of DCE-MRI parameters for brain tumors through implementation of physiologically-based pharmacokinetic model approaches for Gd-DOTA.

Dynamic-contrast enhanced magnetic resonance imaging (DCE-MRI) is used for detailed characterization of pathology of lesions sites, such as brain tumors, by quantitative analysis of tracer's data through the use of pharmacokinetic (PK) models. A key component for PK models in DCE-MRI is the estimation of the concentration-time profile of the tracer in a nearby vessel, referred as Arterial Input Function (AIF). The aim of this work was to assess through full body physiologically-based pharmacokinetic (PBPK) model approaches the PK profile of gadoteric acid (Gd-DOTA) and explore potential application for parameter estimation in DCE-MRI based on PBPK-derived AIFs. The PBPK simulations were generated through Simcyp(®) platform and the predicted PK parameters for Gd-DOTA were compared with available clinical data regarding healthy volunteers and renal impairment patients. The assessment of DCE-MRI parameters was implemented by utilizing similar virtual profiles based on gender, age and weight to clinical profiles of patients diagnosed with glioblastoma multiforme. The PBPK-derived AIFs were then used to compute DCE-MRI parameters through the Extended Tofts Model and compared with the corresponding ones derived from image-based AIF computation. The comparison involved: (i) image measured AIF of patients vs AIF of in silico profile, and, (ii) population average AIF vs in silico mean AIFs. The results indicate that PBPK-derived AIFs allowed the estimation of comparable imaging biomarkers with those calculated from typical DCE-MRI image analysis. The incorporation of PBPK models and potential utilization of in silico profiles to real patient data, can provide new perspectives in DCE-MRI parameter estimation and data analysis.

Technology-Based Innovations to Foster Personalized Healthy Lifestyles and Well-Being: A Targeted Review.

BACKGROUND: New community-based arrangements and novel technologies can empower individuals to be active participants in their health maintenance, enabling people to control and self-regulate their health and wellness and make better health- and lifestyle-related decisions. Mobile sensing technology and health systems responsive to individual profiles combined with cloud computing can expand innovation for new types of interoperable services that are consumer-oriented and community-based. This could fuel a paradigm shift in the way health care can be, or should be, provided and received, while lessening the burden on exhausted health and social care systems. OBJECTIVE: Our goal is to identify and discuss the main scientific and engineering challenges that need to be successfully addressed in delivering state-of-the-art, ubiquitous eHealth and mHealth services, including citizen-centered wellness management services, and reposition their role and potential within a broader context of diverse sociotechnical drivers, agents, and stakeholders. METHODS: We review the state-of-the-art relevant to the development and implementation of eHealth and mHealth services in critical domains. We identify and discuss scientific, engineering, and implementation-related challenges that need to be overcome to move research, development, and the market forward. RESULTS: Several important advances have been identified in the fields of systems for personalized health monitoring, such as smartphone platforms and intelligent ubiquitous services. Sensors embedded in smartphones and clothes are making the unobtrusive recognition of physical activity, behavior, and lifestyle possible, and thus the deployment of platforms for health assistance and citizen empowerment. Similarly, significant advances are observed in the domain of infrastructure supporting services. Still, many technical problems remain to be solved, combined with no less challenging issues related to security, privacy, trust, and organizational dynamics. CONCLUSIONS: Delivering innovative ubiquitous eHealth and mHealth services, including citizen-centered wellness and lifestyle management services, goes well beyond the development of technical solutions. For the large-scale information and communication technology-supported adoption of healthier lifestyles to take place, crucial innovations are needed in the process of making and deploying usable empowering end-user services that are trusted and user-acceptable. Such innovations require multidomain, multilevel, transdisciplinary work, grounded in theory but driven by citizens' and health care professionals' needs, expectations, and capabilities and matched by business ability to bring innovation to the market.

Semantic biomedical resource discovery: a Natural Language Processing framework.

BACKGROUND: A plethora of publicly available biomedical resources do currently exist and are constantly increasing at a fast rate. In parallel, specialized repositories are been developed, indexing numerous clinical and biomedical tools. The main drawback of such repositories is the difficulty in locating appropriate resources for a clinical or biomedical decision task, especially for non-Information Technology expert users. In parallel, although NLP research in the clinical domain has been active since the 1960s, progress in the development of NLP applications has been slow and lags behind progress in the general NLP domain. The aim of the present study is to investigate the use of semantics for biomedical resources annotation with domain specific ontologies and exploit Natural Language Processing methods in empowering the non-Information Technology expert users to efficiently search for biomedical resources using natural language. METHODS: A Natural Language Processing engine which can "translate" free text into targeted queries, automatically transforming a clinical research question into a request description that contains only terms of ontologies, has been implemented. The implementation is based on information extraction techniques for text in natural language, guided by integrated ontologies. Furthermore, knowledge from robust text mining methods has been incorporated to map descriptions into suitable domain ontologies in order to ensure that the biomedical resources descriptions are domain oriented and enhance the accuracy of services discovery. The framework is freely available as a web application at ( http://calchas.ics.forth.gr/ ). RESULTS: For our experiments, a range of clinical questions were established based on descriptions of clinical trials from the ClinicalTrials.gov registry as well as recommendations from clinicians. Domain experts manually identified the available tools in a tools repository which are suitable for addressing the clinical questions at hand, either individually or as a set of tools forming a computational pipeline. The results were compared with those obtained from an automated discovery of candidate biomedical tools. For the evaluation of the results, precision and recall measurements were used. Our results indicate that the proposed framework has a high precision and low recall, implying that the system returns essentially more relevant results than irrelevant. CONCLUSIONS: There are adequate biomedical ontologies already available, sufficiency of existing NLP tools and quality of biomedical annotation systems for the implementation of a biomedical resources discovery framework, based on the semantic annotation of resources and the use on NLP techniques. The results of the present study demonstrate the clinical utility of the application of the proposed framework which aims to bridge the gap between clinical question in natural language and efficient dynamic biomedical resources discovery.

The current status of noninvasive intracranial pressure monitoring: A literature review.

Elevated intracranial pressure (ICP) is a life-threatening condition that must be promptly diagnosed. However, the gold standard methods for ICP monitoring are invasive, time-consuming, and they involve certain risks. To address these risks, many noninvasive approaches have been proposed. This study undertakes a literature review of the existing noninvasive methods, which have reported promising results. The experimental base on which they are established, however, prevents their application in emergency conditions and thus none of them are capable of replacing the traditional invasive methods to date. On the other hand, contemporary methods leverage Machine Learning (ML) which has already shown unprecedented results in several medical research areas. That said, only a few publications exist on ML-based approaches for ICP estimation, which are not appropriate for emergency conditions due to their restricted capability of employing the medical imaging data available in intensive care units. The lack of such image-based ML models to estimate ICP is attributed to the scarcity of annotated datasets requiring directly measured ICP data. This ascertainment highlights an active and unexplored scientific frontier, calling for further research and development in the field of ICP estimation, particularly leveraging the untapped potential of ML techniques.

Image Quality Assessment Tool for Conventional and Dynamic Magnetic Resonance Imaging Acquisitions.

Image quality assessment of magnetic resonance imaging (MRI) data is an important factor not only for conventional diagnosis and protocol optimization but also for fairness, trustworthiness, and robustness of artificial intelligence (AI) applications, especially on large heterogeneous datasets. Information on image quality in multi-centric studies is important to complement the contribution profile from each data node along with quantity information, especially when large variability is expected, and certain acceptance criteria apply. The main goal of this work was to present a tool enabling users to assess image quality based on both subjective criteria as well as objective image quality metrics used to support the decision on image quality based on evidence. The evaluation can be performed on both conventional and dynamic MRI acquisition protocols, while the latter is also checked longitudinally across dynamic series. The assessment provides an overall image quality score and information on the types of artifacts and degrading factors as well as a number of objective metrics for automated evaluation across series (BRISQUE score, Total Variation, PSNR, SSIM, FSIM, MS-SSIM). Moreover, the user can define specific regions of interest (ROIs) to calculate the regional signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR), thus individualizing the quality output to specific use cases, such as tissue-specific contrast or regional noise quantification.

Public data homogenization for AI model development in breast cancer.

BACKGROUND: Developing trustworthy artificial intelligence (AI) models for clinical applications requires access to clinical and imaging data cohorts. Reusing of publicly available datasets has the potential to fill this gap. Specifically in the domain of breast cancer, a large archive of publicly accessible medical images along with the corresponding clinical data is available at The Cancer Imaging Archive (TCIA). However, existing datasets cannot be directly used as they are heterogeneous and cannot be effectively filtered for selecting specific image types required to develop AI models. This work focuses on the development of a homogenized dataset in the domain of breast cancer including clinical and imaging data. METHODS: Five datasets were acquired from the TCIA and were harmonized. For the clinical data harmonization, a common data model was developed and a repeatable, documented "extract-transform-load" process was defined and executed for their homogenization. Further, Digital Imaging and COmmunications in Medicine (DICOM) information was extracted from magnetic resonance imaging (MRI) data and made accessible and searchable. RESULTS: The resulting harmonized dataset includes information about 2,035 subjects with breast cancer. Further, a platform named RV-Cherry-Picker enables search over both the clinical and diagnostic imaging datasets, providing unified access, facilitating the downloading of all study imaging that correspond to specific series' characteristics (e.g., dynamic contrast-enhanced series), and reducing the burden of acquiring the appropriate set of images for the respective AI model scenario. CONCLUSIONS: RV-Cherry-Picker provides access to the largest, publicly available, homogenized, imaging/clinical dataset for breast cancer to develop AI models on top. RELEVANCE STATEMENT: We present a solution for creating merged public datasets supporting AI model development, using as an example the breast cancer domain and magnetic resonance imaging images. KEY POINTS: • The proposed platform allows unified access to the largest, homogenized public imaging dataset for breast cancer. • A methodology for the semantically enriched homogenization of public clinical data is presented. • The platform is able to make a detailed selection of breast MRI data for the development of AI models.

Unveiling the Power of Model-Agnostic Multiscale Analysis for Enhancing Artificial Intelligence Models in Breast Cancer Histopathology Images.

Developing AI models for digital pathology has traditionally relied on single-scale analysis of histopathology slides. However, a whole slide image is a rich digital representation of the tissue, captured at various magnification levels. Limiting our analysis to a single scale overlooks critical information, spanning from intricate high-resolution cellular details to broad low-resolution tissue structures. In this study, we propose a model-agnostic multiresolution feature aggregation framework tailored for the analysis of histopathology slides in the context of breast cancer, on a multicohort dataset of 2038 patient samples. We have adapted 9 state-of-the-art multiple instance learning models on our multi-scale methodology and evaluated their performance on grade prediction, TP53 mutation status prediction and survival prediction. The results prove the dominance of the multiresolution methodology, and specifically, concatenating or linearly transforming via a learnable layer the feature vectors of image patches from a high (20x) and low (10x) magnification factors achieve improved performance for all prediction tasks across domain-specific and imagenet-based features. On the contrary, the performance of uniresolution baseline models was not consistent across domain-specific and imagenet-based features. Moreover, we shed light on the inherent inconsistencies observed in models trained on whole-tissue-sections when validated against biopsy-based datasets. Despite these challenges, our findings underscore the superiority of multiresolution analysis over uniresolution methods. Finally, cross-scale analysis also benefits the explainability aspects of attention-based architectures, since one can extract attention maps at the tissue- and cell-levels, improving the interpretation of the model's decision. The code and results of this study can be found at github.com/tsikup/multiresolution_histopathology.

Analysis of domain shift in whole prostate gland, zonal and lesions segmentation and detection, using multicentric retrospective data.

Despite being one of the most prevalent forms of cancer, prostate cancer (PCa) shows a significantly high survival rate, provided there is timely detection and treatment. Computational methods can help make this detection process considerably faster and more robust. However, some modern machine-learning approaches require accurate segmentation of the prostate gland and the index lesion. Since performing manual segmentations is a very time-consuming task, and highly prone to inter-observer variability, there is a need to develop robust semi-automatic segmentation models. In this work, we leverage the large and highly diverse ProstateNet dataset, which includes 638 whole gland and 461 lesion segmentation masks, from 3 different scanner manufacturers provided by 14 institutions, in addition to other 3 independent public datasets, to train accurate and robust segmentation models for the whole prostate gland, zones and lesions. We show that models trained on large amounts of diverse data are better at generalizing to data from other institutions and obtained with other manufacturers, outperforming models trained on single-institution single-manufacturer datasets in all segmentation tasks. Furthermore, we show that lesion segmentation models trained on ProstateNet can be reliably used as lesion detection models.

LKB1 Loss Correlates with STING Loss and, in Cooperation with ß-Catenin Membranous Loss, Indicates Poor Prognosis in Patients with Operable Non-Small Cell Lung Cancer.

To investigate the incidence and prognostically significant correlations and cooperations of LKB1 loss of expression in non-small cell lung cancer (NSCLC), surgical specimens from 188 metastatic and 60 non-metastatic operable stage I-IIIA NSCLC patients were analyzed to evaluate their expression of LKB1 and pAMPK proteins in relation to various processes. The investigated factors included antitumor immunity response regulators STING and PD-L1; pro-angiogenic, EMT and cell cycle targets, as well as metastasis-related (VEGFC, PDGFRα, PDGFRß, p53, p16, Cyclin D1, ZEB1, CD24) targets; and cell adhesion (ß-catenin) molecules. The protein expression levels were evaluated via immunohistochemistry; the RNA levels of LKB1 and NEDD9 were evaluated via PCR, while KRAS exon 2 and BRAFV600E mutations were evaluated by Sanger sequencing. Overall, loss of LKB1 protein expression was observed in 21% (51/248) patients and correlated significantly with histotype (p < 0.001), KRAS mutations (p < 0.001), KC status (concomitant KRAS mutation and p16 downregulation) (p < 0.001), STING loss (p < 0.001), and high CD24 expression (p < 0.001). STING loss also correlated significantly with loss of LKB1 expression in the metastatic setting both overall (p = 0.014) and in lung adenocarcinomas (LUACs) (p = 0.005). Additionally, LKB1 loss correlated significantly with a lack of or low ß-catenin membranous expression exclusively in LUACs, both independently of the metastatic status (p = 0.019) and in the metastatic setting (p = 0.007). Patients with tumors yielding LKB1 loss and concomitant nonexistent or low ß-catenin membrane expression experienced significantly inferior median overall survival of 20.50 vs. 52.99 months; p < 0.001 as well as significantly greater risk of death (HR: 3.32, 95% c.i.: 1.71-6.43; p <0.001). Our findings underscore the impact of the synergy of LKB1 with STING and ß-catenin in NSCLC, in prognosis.

Documenting the de-identification process of clinical and imaging data for AI for health imaging projects.

Artificial intelligence (AI) is revolutionizing the field of medical imaging, holding the potential to shift medicine from a reactive "sick-care" approach to a proactive focus on healthcare and prevention. The successful development of AI in this domain relies on access to large, comprehensive, and standardized real-world datasets that accurately represent diverse populations and diseases. However, images and data are sensitive, and as such, before using them in any way the data needs to be modified to protect the privacy of the patients. This paper explores the approaches in the domain of five EU projects working on the creation of ethically compliant and GDPR-regulated European medical imaging platforms, focused on cancer-related data. It presents the individual approaches to the de-identification of imaging data, and describes the problems and the solutions adopted in each case. Further, lessons learned are provided, enabling future projects to optimally handle the problem of data de-identification. CRITICAL RELEVANCE STATEMENT: This paper presents key approaches from five flagship EU projects for the de-identification of imaging and clinical data offering valuable insights and guidelines in the domain. KEY POINTS: ΑΙ models for health imaging require access to large amounts of data. Access to large imaging datasets requires an appropriate de-identification process. This paper provides de-identification guidelines from the AI for health imaging (AI4HI) projects.

MRI Methods to Visualize and Quantify Adipose Tissue in Health and Disease.

MRI is the modality of choice for a vast range of pathologies but also a sensitive probe into human physiology and tissue function. For this reason, several methodologies have been developed and continuously evolve in order to non-invasively monitor underlying phenomena in human adipose tissue that were difficult to assess in the past through visual inspection of standard imaging modalities. To this end, this work describes the imaging methodologies used in medical practice and lists the most important quantitative markers related to adipose tissue physiology and pathology that are currently supporting diagnosis, longitudinal evaluation and patient management decisions. The underlying physical principles and the resulting markers are presented and associated with frequently encountered pathologies in radiology in order to set the frame of the ability of MRI to reveal the complex role of adipose tissue, not as an inert tissue but as an active endocrine organ.