An overview of the use of cutting-edge artificial intelligence (AI) modeling to produce synthetic medical data (SMD) in decentralized clinical machine learning (ML) for ovarian cancer(OC) and ovarian lymphoma(OL).

An overview of the use of cutting-edge artificial intelligence (AI) modeling to produce synthetic medical data (SMD) in decentralized clinical machine learning (ML) for ovarian cancer(OC) and ovarian lymphoma(OL) - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice J Ultrasound . 2025 Jan 22;28(2):483–492. doi: 10.1007/s40477-025-00983-3 An overview of the use of cutting-edge artificial intelligence (AI) modeling to produce synthetic medical data (SMD) in decentralized clinical machine learning (ML) for ovarian cancer(OC) and ovarian lymphoma(OL) Diana Donatello 1, ✉ Author information Article notes Copyright and License information 1 Costa Contina street n. 19, 66054 Vasto, Chieti Italy ✉ Corresponding author. Received 2024 Dec 30; Accepted 2025 Jan 7; Collection date 2025 Jun. © Società Italiana di Ultrasonologia in Medicina e Biologia (SIUMB) 2025. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law. PMC Copyright notice PMCID: PMC12145400  PMID: 39841353 Abstract Aim o point out how novel analysis tools of AI can make sense of the data acquired during OL and OC diagnosis and treatment in an effort to help improve and standardize the patient pathway for these disease. Material and methods ultilizing programmed detection of heterogeneus OL and OC habitats through radiomics and correlate to imaging based tumor grading plus a literature review. Results new analysis pipelines have been generated for integrating imaging and patient demographic data and identify new multi-omic biomarkers of response prediction and tumour grading using cutting-edge artificial intelligence (AI) in OL and OC. Description deline the main AI methods used in OL and OC that we can try to standardize in the clinical radiological and medical practice to ameliorate the patients diagnosis and theraphy. Conclusion through new AI methods it’s possible to combine research into a SwarmDeepSurv, generate new data flow channels, create medical imaging data channels of OL and OC using AI and identify new biomarkers of OL and OC . . Keywords: Ovarian lymphoma, Ovarian cancer, Artificial intelligence, Radiomics, Swarm intelligence Introduction Lymphomas, which account for around 8% of all neoplasms, are primary neoplasms of the immune system that originate inside lymphoid tissue. They are separated into two groups: the more diverse group of non-Hodgkin’s lymphomas and the more homogeneous group of Hodgkin’s lymphomas, which are named after the English physician Thomas Hodgkin, who made the discovery of them in 1832. In addition to differing from non-Hodgkin lymphomas in their histology, Hodgkin’s lymphomas typically manifest as a nodal disease that affects clusters of nearby lymph nodes while retaining contiguity. Only 2–5% of Hodgkin lymphomas have extranodal localizations. The branches of the thoracic duct—the supraclavicular, lateralcervical, and mediastinal—are the most severely afflicted. In contrast to non-Hodgkin lymphomas, they do not exhibit continuity in their nodal presentation. Instead, they frequently result in extranodal localizations, accounting for 25–50% of all non-Hodgkin lymphomas, and they frequently disseminate. Ovarian involvement by malignant lymphoma occurs at a frequency of 7–26%, with up to 25% of women dying with lymphomas exhibiting ovarian infltration at autopsy. There is no lymphoid tissue in a normal ovary. It has been proposed that chronic inflammation (such as endometriosis or pelvic inflammation disease) may make women more likely to develop primary ovarian lympholysis (OL). This condition is thought to be caused by lymphocytes that supply blood vessels, are distributed throughout the ovarian stroma, reside in ovarian follicles, and are connected to the corpus luteum. There are two ways in which the ovary might become involved in the lymphomatous process: primary or secondary. Frequently, the ovarian mass is a manifestation of a widely disseminated lymphomatous disease, whereas the initial clinical manifestation of an occult lymphoma as an ovarian mass is uncommon. Primary OL is rare, accounting for only 0.5% of all non-Hodgkin’s lymphomas (NHL), and 1.5% of all ovarian neoplasms [ 1 , 2 ]. On the other hand, although involvement of the genital tract by non-Hodgkin lymphoma is rare, the ovary is the most frequently affected genital organ and secondary lymphoma at this site is the most common presentation. At the same time, information on the treatment and clinical images of primary and secondary ovarian lymphoma is lacking. As with all non-Hodgkin lymphomas, extranodal lymphoma with ovarian involvement can also be staged with the Ann Arbor classification. (Table 1 ). Clinical presentation is non-specifc (abdominal pain and distension), and physical examination may reveal a pelvic mass. Imaging techniques such as Transvaginal Sonography (TVUS), Computer Tomography, Whole-Body Diffusion-Weighted MRI and F18FDG PETR/CT, are usually performed in the work-up and follow-up of women with OL and OC [ 1 ].Conversely, the most deadly gynecological cancer is ovarian cancer (OC) (Tables 2 and 3 ) [ 4 , 5 ]. Ovarian cancer ranks as the sixth most common cause of cancer-related deaths among women in the United States. According to estimates from the American Cancer Society (ACS), 22,240 women were newly diagnosed with ovarian cancer in 2018, and 14,075 of them women lost their lives to the disease. It is challenging to diagnose OC in a timely manner due to the absence of reliable biomarkers, not just in OL cases. The current “one size fits all” method to treating OC involves severe debulking surgery to remove any visible tumors, which is followed by chemotherapy with a mix of taxanes and platinum. The likelihood of survival is increased with gynecological oncologist-performed debulking surgery; yet, many patients are not treated by gynecological oncologists. Chemotherapy before surgery is therefore sometimes just as successful as primary debulking. Chemotherapy treatment is initially successful in 70–80% of individuals. Nonetheless, 80–90% of patients will experience a recurrence of the cancer within 5 years, and tumors frequently develop resistance to chemotherapy [ 2 ]. It is necessary to carefully assess the location and grade of OL and OC before developing appropriate therapy routes for specific patients. Traditionally, interdisciplinary teams at specific specialists like imaging, biopsies, and histopathology oversee diagnosis and treatment. A proper differential diagnosis between OL and primary OC is essential for treatment, because urgent chemotherapy is the optimal treatment for the former, while surgical resection with debulking is the treatment of choice for the latter [ 1 ]. Table 1. Ann Arbour classification Open in a new tab Table 2. Ovarian cancer classification Open in a new tab Table 3. FIGO staging for ovarian cancer Open in a new tab Material and methods To carry out our study we decided to considered mainly two previously studies: The study of Crispin-Ortuzar et al [ 23 ]. The study of Al-Tashi et al [ 14 ]. For the remainder of the project, we have defined a basic radiomics approach for correlating imaging features with tumor grading (using histopathology data as a gold standard) and for automatic detection of heterogeneous OL and OC environments using imaging, as shown in Fig. 1 . Practically, we’ve blindfolded radiomics with tumor grade recognition and tumor habitat identification. This process, after the collection of all of the patients’ data, can be fused through the novel AI methods (for example, swarm intelligence) with demographic patient’s data and multi-omics biomarkers to develop a big data set, the so-called SwarmDeep Surv Model for the composition of a multiframework that is capable of predicting survival and delineating renspose in pre and post treatment. We reviewed retrospectively a total of 74 MR imagies of ovarian cancer from 110 patients analyzed, and we have blindfolded radiomics with tumor grade recogniction and tumor habitat identification (Fig. 1 ). Fig. 1. Open in a new tab Classification scheme One of the limitation of our study is the small number of patients, so that we cant’ build a real “Swarm Deep Model” due to the slight number of patients. In fact in the study of Al-Tashi et al that they used four different type of cancer to obtain an adeguate number of cases. Results All of the aforementioned innovative AI models may be employed in current and future studies to determine the prognosis and survival of the OL and OC. Further research should be conducted to enhance the algorithms in order to achieve a suitable balance between exploration and exploitation, which is crucial for improving the discovery of cancer prognostic biomarkers. OL and OC are types of cancer for which surgery is the primary therapy. Following surgery, the majority of patients undergo relapses, which significantly reduce patient survival by causing both distant metastatic disease and local tumor recurrence. To reduce the likelihood of a relapse, systemic treatment is further confounded by the fact that these cancers (less OL) are very heterogeneous; various parts within the same tumour might respond differently to the same treatment. The grade of OC tumors must be thoroughly examined before appropriate treatment routes can be developed for individual patients. Traditionally, diagnosis and therapy are evaluated by multidisciplinary teams at chosen specialists using a variety of clinical investigations such as imaging, multi-core biopsies, histopathology, and genomic profiling. The outcomes of this study are to demonstrate that today it’s feasible to rely on innovative analytical tools in artificial intelligence (AI), that may be used alone or combination to comprehend and standardize these malignancies therapy in an effort to help improve and regulate the patient. Discussion This article describes the main new methodologies of artificial intelligence (AI) capable of implementing diagnosis and therapies and consequently survival in OL and OC. We’ve focused the discussion on the development of new AI techniques for response assessment of OL nad OC using multi-centre imaging data. We know infact that imaging provides perhaps the only non-invasive approach for monitoring how heterogeneous regions within OL and OC tumours change during the course of treatment, but analysis methods for accurately tracking these changes are still in their infancy. Furthermore, given the paucity of robust imaging guidelines for evaluating OL and OC tumours during treatment, include data for integrating imaging material with the other patient-related information (including age, family history, blood markers), combining disparate imaging datasets from multiple institutions, comprising of different imaging modalities and protocols. The developed techniques can be integrated into a software framework with an aim to provide a robust prototype for further integration into clinical practise and improvement in patient outcome. In this study, we demostrate how we can leverage imaging and other multiomic data to develop data-driven AI models of image-based OL and OC tumour grading, heterogeneous tumour response, and prediction of local recurrence, treatment toxicity and patient survival. Then these developments are integrate into a robust research software framework, with the aim for clinical dissemination of these tools into the medical area. With encouraging findings thus far shown in the literature, artificial intelligence does, in fact, have a great deal of promise to enhance the pre-operative diagnosis of OL and OC, for example, on ultrasound tests. The doctor still has the last say in all decisions, but working together with AI in a synergistic way may help identify problems sooner and continue to improve patient outcomes. AI in radiomics features extration based on tumour grading and in the detection of heterogeneous subcompartments from imaging in response to neoadjuvant radiotherapy The major AI innovation in the area of radiology is radiomics that sostatially is a rapidly evolving field of research concerned with the extraction of quantitative metrics the so-called radiomic features within medical images [ 6 , 19 ]. In OL and OC TVUS, CT, MRI and serum level measurements of protein biomarker CA125 are typically performed prior to surgery for OL and OC diagnosis, along with biopsy for tumour grading [ 8 , 10 , 12 , 20 ]. Radiomics is used to identify image-based models of tumor grade and extract high-order characteristics (such as tumor form and texture) from US, MRI, and CT images. Imaging allows for the imaging of the entire disease’s extent and can account for tumor heterogeneity. (Fig. 1 ) In this methodology, recorded OL and OC grades determined by histopathology serve as the “gold-standard” [ 3 , 4 , 8 – 10 ]. Because of the inter-tumoural heterogeneity seen in both OL and OC, traditional radiological measures of tumour size using CT or MRI fail as response indicators. Using multi-parametric MRI or CT data, artificial intelligence (AI) can identify heterogeneous subcompartments that may arise in response to chemotherapy. For instance, multi-parametric MRI images can be integrated and classified into one of four tissue groups: Cellular vascularized tumour Poorly vascularized tissue Dead tissue Fatty tissue This categorization system is also used to check for changes in each compartment’s volume after chemotherapy. Prior to therapy, a heterogeneous pattern may be seen, with highly vascular sections making up the bulk of the tumor. The percentage of tissue with poor vascularization may grow after chemotherapy, which might mean that the tumor’s blood supply has been disrupted. (Fig. 1 ). Response/prognostic models that integrate imaging phenotypes with regional histopathology and tissue molecular profiling will be developed using the archived data that investigate the biological foundation of these areas. This categorization system may also be used after radiation therapy to evaluate alterations in the volume of every compartment. Prior to therapy, a heterogeneous pattern is found, with highly vascular areas making up the bulk of the tumor. After radiation therapy, the percentage of poorly vascularized tissue increases even while the overall tumor volume does not significantly change.This might mean that the tumor’s blood supply has been disrupted. Thus, by demonstrating the biological underpinnings of these locations by data archiving, predictive models combining imaging phenotypes with tissue molecular profiling and regional histology will be developed. By comparing results with histology, additional studies of this methodology in the context of chemotherapy response verify the procedure. Al in potential multi-omic, predictive biomarkers of OL ana OC tumor response to neoadjuvant chemotherapy Artificial intellect (AI) is a branch of computer science that aims to replicate human intellect and carry out tasks like problem-solving and decision-making that typically need human cognition [ 9 ]. The two primary domains of AI, machine learning (ML) and deep learning (DL), have demonstrated superior performance than traditional techniques in the molecular characterisation of cancer, prognostication, diagnosis, patient categorization, and prediction-making in numerous cancer types, including ovarian cancer. The very diverse condition known as high grade serous ovarian cancer (HGSOC) usually manifests as advanced, metastatic illness [ 23 ]. One of the main challenges in predicting and comprehending the essential response factors of neoadjuvant chemotherapy (NACT) is the multi-scale complexity of HGSOC. IRON, an integrated radiogenomic framework to predict the volumetric response of heterogeneous, multi-site ovarian cancer to NACT, is presented by Crispin-Ortuzar et al. in their paper. Clinical, chemotherapeutic treatment, CA-125, ctDNA, and radiomics variables taken from every primary and metastatic illness at diagnosis were included in two separate, thoroughly annotated data sets that they employed [ 23 ]. An ensemble machine learning model trained to anticipate volume loss during NACT utilized the attributes as input.They showed that radiomics characteristics are necessary to achieve considerable prediction power and verified the method using an external training data set. Their strategy is simply generalized and is relevant to other malignancies with heterogeneous, multi-site illness. Testing on tumor tissue samples has been the norm for analyzing a tumor’s molecular characteristics. Without surgical excision, it is impossible to get samples of OC tissues since the ovaries are entirely intraperitoneal organs. In particular, because cancer cells readily spread into the peritoneal cavity and penetration would encourage peritoneal metastasis, needle biopsies for early-stage ovarian tumors should be avoided. In the past, ovarian surface epithelium was frequently used by researchers as a typical control tissue in comparative studies. In order to find molecular biomarkers for OC, particularly HGSOCs, recent research typically employ normal fallopian tube controls in comparison with tumor samples, as there is growing awareness that the fallopian tube is the main tissue of origin for many OCs. There is an urgent need for informative and less intrusive biomarkers that might be employed for early identification of OC, as tumor tissue samples for OC are difficult to collect serially. Improvements in technology, sample preparation, and data processing over the last 20 years have made high-quality, repeatable, and comprehensive data from a variety of clinical samples available. A potential substitute for traditional methods of diagnosing cancer is liquid biopsies. Compared to traditional tumor biopsies, liquid biopsies provide many distinct benefits. Firstly, it is safe and minimally intrusive, circumventing the possible risks associated with tissue biopsies. Secondly, it offers a chance to find diverse tumor-specific changes that tissue biopsies could overlook.More significantly, repeated sampling over time is made possible by liquid biopsies, and this information is crucial for informing clinical judgments. Body fluids such as blood, urine, uterine lavage, cervical smear, and ascites are characterized by a notable enrichment of nucleic acids, proteins, or metabolites, which may serve as a possible source of diagnostic biomarkers for occult OC patients. There are still issues about how to enhance and utilize the data from different research to the greatest of abilities as massive “omics” data continue to grow. Because the vast volumes of multi-omics data used in cancer research are frequently noisy, lack statistical power, and are physiologically and computationally diverse, it is very challenging to extract biological insights from these high-dimensional datasets using conventional data analysis techniques. The development of computational techniques ranging from data integration, statistical methodologies, and artificial intelligence (AI) systems is necessary for the analysis of datasets produced by multi-omics sequencing. Because of recent advancements in computer science, the use of AI in preclinical and translational cancer research has grown significantly. A subfield of artificial intelligence called machine learning (ML) allows for the rigorous analysis of numerous datasets to find patterns and correlations that had not been found before.26 In earlier investigations, researchers used numerous markers in peripheral blood and clinical parameters to establish an AI-based OC-specific prediction strategy for pretreatment assessment of clinical stages, histotypes, surgical outcomes, and prognosis of patients with epithelial ovarian cancer.When compared to earlier findings, they discovered that the ML technique could predict malignant tumors with a noticeably high degree of accuracy. The incorporation and evaluation of high-throughput molecular tests grounded on machine learning methodologies facilitate the identification of distinct variants associated with the illness and the pursuit of additional biomarkers. Regardless of changes at each individual omics level, biomarker candidates derived from integrated analysis of big data would remain physiologically relevant. Therefore, it is anticipated that the quickly developing AI-based analysis would greatly support precision medicine in OC. The last several years have seen a significant increase in the identification of prospective biomarkers thanks to multi-omics technologies, however there are still gaps in the knowledge between discovery research and clinical application.Study design flaws that lead to inadequate statistical power in many studies are a primary cause of the low rate at which discovery research gets translated into clinical practice. Experiments must be properly developed in order to improve the prediction ability of prospective diagnostic biomarkers. For instance, the sample type, omics technical parameters, and statistical analysis techniques should all be taken into consideration when determining the sample size. Additionally, standardizing sample handling and preservation might aid in lowering biological variability throughout research projects. Ultimately, before applying the findings to standard clinical care, more thorough validation has to be completed. Large-scale research must be carried out using proper statistical methodologies and strict quality control measures. The necessity of using OC mortality as the primary endpoint in screening studies was brought to light by the recent UKCTOCS research. Subsequent big studies are expected to track survival outcomes and determine whether particular screening techniques might lower death over an extended period of time (perhaps ten years). Studies have showed efficacy for predicting models of patient prognosis based on patient demographic information and ovarian cancer parameters such as size and location. Unlike imaging, these descriptors are unable to take spatial tumor heterogeneity into account. As far as we are aware, no such models have been established for treatment toxicities or illness recurrence [ 1 , 3 , 4 , 9 , 11 – 13 ]. The enhanced precision of the merged imaging and demographic characteristics will be contrasted with current models or models created with just demographic data using a cross-validation method, such the Swarm Deep or the IRON model.That’s why we focused on the research of Al-Thasi et al. on swarm learning, which integrates many modalities and machine learning techniques while permitting collaborative learning on a sizable data pool without jeopardizing patient data. Utilizing all silos yields a dataset that aids in overcoming research obstacles. By utilizing all silos, a dataset is produced that may be used to mitigate research biases and variations in collection techniques. Moreover, more robust detection may be achieved for uncommon illnesses (e.g., Primary OL) with limited local training data [ 3 ]. Swarm learning is a type of artificial intelligence that uses a network of endpoint devices that can generate and interpret data right at the source. Individual agents may process and respond to input from their peers without relying on a centralized data lake or decision matrix by instantly sharing pertinent information throughout the network that meets predefined criteria [ 4 – 7 ]. Systems like SL should be favored since they provide highly controlled, equitable, and transparent shared data analytics while protecting user privacy. In order to diagnose COVID-19 image-based using patterns found in CT or X-ray images, structured medical records, or data from wearables for illness tracking, SL should be investigated. When combined, SL and transcriptomics (or other medical data) offer a very promising way to improve data privacy, security, and confidentiality while also reducing data traffic and democratizing the use of SL among the many stakeholders in the medical field [ 2 , 4 ]. Furthermore, one of the many creative ways that companies may share datasets with a wider user base is by employing synthetic data. Unfortunately, there isn’t much literature that examines its possibilities and uses in the medical field. [ 8 , 12 ] AI in the SwarmDeepSurv (SDS) model By drawing inspiration from the Al-Tashi et al [ 14 ] model, a unique SDS toolbox that can integrate imaging data with patient demographics to identify new predictive models of local recurrence and patient survival of OL and OC may be created [ 14 ]. They provide SwarmDeepSurv, a survival model that creatively blends DeepSurv with SI-based feature selection. All of the aforementioned strengths of the DeepSurv model and SI algorithms are anticipated to be carried over to SwarmDeepSurv; in particular, the DeepSurv model incorporated two highly acclaimed SI algorithms, PSO and GWO, as well as one evolutionary algorithm, GA. Additionally, they developed four objective functions with the goal of striking a balance between the quantity of chosen characteristics and prediction performance. In their research, using seven multicenter cohorts with 1058 patients and four different forms of cancer, they quantitatively compared SwarmDeepSurv with four traditional models. In order to do a survival analysis, they additionally looked at the effects of each objective function and further examined how comparable the chosen radiomics properties were across all models.The institutional review board at MD Anderson authorized this study. The main objective was to create a survival analysis framework that works well and can detect imaging patterns associated with cancer prognosis. In order to do this, they developed a novel SI-based framework (dubbed SwarmDeepSurv) and tested it against cutting-edge algorithms. The diagram below shows the general architecture of the suggested SwarmDeepSurv, which is divided into four main stages. First, as described in supplemental experimental procedures, they gathered demographic information and pan-cancer imaging from patients with four different types of cancer: non-small cell lung cancer ( n = 616), breast cancer ( n = 186), brain cancer (glioblastoma; n = 128), and head and neck squamous cell carcinoma ( n = 128). The typical radiomics characteristics, such as first order histogram features, second order textural features, and shape-based features, were retrieved during the second phase of the study, which also included picture preparation (see additional experimental techniques). In the third stage, the training cohort was used to fit the model using various feature selection strategies and hyperparameter tweaking. Specifically, they layered the SI with the CoxPH (Cox proportional hazards) deep neural network. As a testing set for the ultimate assessment, model validation was carried out on unseen data in the last step. Using a sizable multicenter radiomics dataset, they created a novel survival prediction framework called SwarmDeepSurv in their study. This framework can predict survival in many cancer types. The deep survival network has been combined with three well acclaimed SI methods. To maximize prediction accuracy and regularize the amount of chosen characteristics, they developed four distinct goal functions. improved stratification in clinically important subgroups, such as high-risk and low-risk patients with improved mortality and substantial difference, is demonstrated by the suggested SwarmDeepSurv models. Together, their research demonstrates that deep learning and the stochastic nature of SI algorithms may generate reliable prognostic radiomics models that can be used as a gauge to assess cancer patients’ prognoses. Promising predictive value beyond traditional algorithms was proven by SwarmDeepSurv.Furthermore, SwarmDeepSurv demonstrated the ability to enhance patient risk categorization through improved CI and mortality hazard ratio. Medical researchers will be able to utilize the suggested SwarmDeepSurv models as a general tool to build, understand, and forecast the consequences of patients’ imaging characteristics on their survival endpoints because of their predictive and modeling capabilities. Due to the curse of dimensionality, existing deep learning survival models based on high-dimensional features have been shown to exhibit significant performance decreases. In order to bridge this gap, SwarmDeepSurv integrates swarm intelligence algorithms with deep survival models. [ 7 , 9 , 14 – 16 ]. For instance, medical researchers will be able to employ the prediction and modeling capabilities of the SwarmDeepSurv models that Qasem Al-Tashi et al. [ 14 ] have presented as a universal tool to construct, understand, and forecast the consequences of patients’ imaging characteristics on their survival endpoints. (Table 4 and Figs. 2 and 3 ) [ 8 , 15 , 16 ]. The prognostic model was the subject of the proof-of-concept research, which shows how it may be expanded to create predictive biomarkers outside of radiomics data in order to analyze clinico-genomic data and support clinical decision-making. Table 4. Steps of proposed SwarmDeepSurv [ 14 ] Open in a new tab Fig. 2. Open in a new tab SwarmDeepSurv: simple graph Fig. 3. Open in a new tab The general structure of the SwarmDeepSurv model that Qasem Al-Tashi et al. suggested consists of gathering large-scale clinical and imaging data related to pan-cancer, preprocessing and extracting features from images, fitting the model via feature selection and hyperparameter tweaking, and validating the model [ 14 ] Conclusion This study demonstrates how it is possible to combine research into AI models for use in future clinical trials and further clinical applications for OL and OC; additionally, it is possible to create new data flow channels for integrating imaging and patient demographic data in OL and OC, as well as generate medical imaging data of OL and OC using AI with radiomic feature extraction channels, and incorporate it in a “big data set” known as the SwarmDeepSurv Model. Finally, we can use AI to uncover novel biomarkers for predicting OL and OC response and grading tumours [ 17 – 24 ]. Abbreviations ML Machine learning DP Deep learning AI Artificial intelligence OL Ovarian lymphoma OC Ovarian cancer SI Swarm intelligence SL Swarm learning HGSOC High grade serous ovarian carcinoma NACT Neoadjuvant chemotherapy IRON Integrated radiogenomics for ovarian neoadjuvant therapy PSO Particle swarm optimization GWO Grey wolf optimizer GA Genetic algorithm UKCTOCS UK collaborative trial of ovarian cancer screening CT Computer tomography MRI Magnetic resonace imaging TVUS Transvaginal ultrasonography CA-125 Cancer antigen 125 ctDNA Circulating tumor DNA Author contributions Not applicable. Funding Not applicable. Data availability All data supporting the findings of this study are available within the paper. Declarations Conflict of interest The author declares that she has no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Ethical approval Not applicable. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References 1. Roy L, Cowden Dahl KD (2018) Can stemness and chemoresistance be therapeutically targeted via signaling pathways in ovarian cancer? Cancers 10:241. 10.3390/cancers10080241 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Donatello D, Battista G, Sassi C (2023) Imaging of ovarian lymphoma. J Ultrasound 26:799–807. 10.1007/s40477-023-00779-3 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Becker M (2022) Swarm learning for decentralized healthcare. Hautarzt 73(4):323–325. 10.1007/s00105-021-04940-z [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Bhardwaj T, Mittal R, Upadhyay H, Lagos L (2022) Applications of swarm intelligent and deep learning algorithms for image-based cancer recognition. In: Garg L, Basterrech S, Banerjee C, Sharma TK (eds) Artificial intelligence in healthcare advanced technologies and societal change. Springer, Singapore [ Google Scholar ] 5. de PinhoPinheiro CA, Nedjah N, de MacedoMourelle L (2020) Detection and classification of pulmonary nodules using deep learning and swarm intelligence. Multimed Tools Appl 79:15437–15465. 10.1007/s11042-019-7473-z [ Google Scholar ] 6. Macedo M, Santana M, dos Santos WP, Menezes R, Bastos-Filho C (2021) Breast cancer diagnosis using thermal image analysis: A data-driven approach based on swarm intelligence and supervised learning for optimized feature selection. Appl Soft Comput. 10.1016/j.asoc.2021.107533 [ Google Scholar ] 7. Rosenberg LB et al. (2018) Artificial swarm intelligence employed to amplify diagnostic accuracy in radiology. In: 2018 IEEE 9th annual information technology, electronics and mobile communication conference (IEMCON). 1186–1191. 10.1109/IEMCON.2018.8614883 8. Al-Rifaie MM, Aber A, Hemanth DJ (2015) Deploying swarm intelligence in medical imaging identifying metastasis, micro-calcifications and brain image segmentation. IET Syst Biol 9(6):234–244. 10.1049/iet-syb.2015.0036 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Cai Y, Sharma A (2021) Swarm intelligence optimization: an exploration and application of machine learning technology. J Intell Syst 30(1):460–469. 10.1515/jisys-2020-0084 [ Google Scholar ] 10. Jan YT, Tsai PS, Huang WH et al (2023) Machine learning combined with radiomics and deep learning features extracted from CT images: a novel AI model to distinguish benign from malignant ovarian tumors. Insights Imaging 14:68. 10.1186/s13244-023-01412 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Cai G, Huang F, Gao Y, Li X, Chi J, Xie J, Zhou L, Feng Y, Huang H, Deng T, Zhou Y, Zhang C, Luo X, Xie X, Gao Q, Zhen X, Liu J (2024) Artificial intelligence-based models enabling accurate diagnosis of ovarian cancer using laboratory tests in China: a multicentre, retrospective cohort study. Lancet Digit Health 6(3):e176–e186. 10.1016/S2589-7500(23)00245-5 [ DOI ] [ PubMed ] [ Google Scholar ] 12. Mitchell S, Nikolopoulos M, El-Zarka A, Al-Karawi D, Al-Zaidi S, Ghai A, Gaughran JE, Sayasneh A (2024) Artificial intelligence in ultrasound diagnoses of ovarian cancer: a systematic review and meta-analysis. Cancers 16:422. 10.3390/cancers16020422 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Saldanha OL, Quirke P, West NP, James JA et al (2022) Swarm learning for decentralized artificial intelligence in cancer histopathology. Nat Med 28(6):1232–1239. 10.1038/s41591-022-01768-5 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Al-Tashi Q, Saad MB, Sheshadri A, Wu CC, Chang JY, Al-Lazikani B, Gibbons C, Vokes NI, Zhang J, Lee JJ, Heymach JV, Jaffray D, Mirjalili S, Wu J (2023) SwarmDeepSurv: swarm intelligence advances deep survival network for prognostic radiomics signatures in four solid cancers. Patterns (N Y) 4(8):F100777. 10.1016/j.patter.2023.100777 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Warnat-Herresthal S, Schultze H, Shastry KL et al (2021) Swarm Learning for decentralized and confidential clinical machine learning. Nature 594:265–270. 10.1038/s41586021-03583-3 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Gonzales A, Guruswamy G, Smith SR (2023) Synthetic data in health care: a narrative review. PLoS Digit Health 2(1):e0000082. 10.1371/journal.pdig.0000082 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Shah R, AstutoAroucheNunes B, Gleason T et al (2023) Utilizing a digital swarm intelligence platform to improve consensus among radiologists and exploring its applications. J Digit Imaging 36:401–413. 10.1007/s10278-022-00662-3 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Majumder S, Katz S, Kontos D, Roshkovan L (2024) State of the art: radiomics and radiomics-related artificial intelligence on the road to clinical translation. BJR Open 6(1):tzad004. 10.1093/bjro/tzad004 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Mayerhoefer ME, Materka A, Langs G, Häggström I, Szczypiński P, Gibbs P, Cook G (2020) Introduction to radiomics. J Nucl Med 61(4):488–495. 10.2967/jnumed.118.222893 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Liberto JM, Chen SY, Shih IM, Wang TH, Wang TL, Pisanic TR 2nd (2022) Current and emerging methods for ovarian cancer screening and diagnostics: a comprehensive review. Cancers (Basel) 14(12):2885. 10.3390/cancers14122885 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Takeyama N, Sasaki Y, Ueda Y, Tashiro Y, Tanaka E, Nagai K, Morioka M, Ogawa T, Tate G, Hashimoto T, Ohgiya Y (2024) Magnetic resonance imaging-based radiomics analysis of the differential diagnosis of ovarian clear cell carcinoma and endometrioid carcinoma: a retrospective study. Jpn J Radiol 42(7):731–743. 10.1007/s11604-024-01545-z [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Sala E, Mema E, Himoto Y, Veeraraghavan H, Brenton JD, Snyder A, Weigelt B, Vargas HA (2017) Unravelling tumour heterogeneity using next-generation imaging: radiomics, radiogenomics, and habitat imaging. Clin Radiol 72(1):3–10. 10.1016/j.crad.2016.09.013 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Crispin-Ortuzar M, Woitek R, Reinius MAV, Moore E, Beer L, Bura V, Rundo L, McCague C, Ursprung S, Escudero Sanchez L, Martin-Gonzalez P, Mouliere F, Chandrananda D, Morris J, Goranova T, Piskorz AM, Singh N, Sahdev A, Pintican R, Zerunian M, Rosenfeld N, Addley H, Jimenez-Linan M, Markowetz F, Sala E, Brenton JD (2023) Integrated radiogenomics models predict response to neoadjuvant chemotherapy in high grade serous ovarian cancer. Nat Commun 14(1):6756. 10.1038/s41467-023-41820-7 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Xiao Y, Bi M, Guo H, Li M (2022) Multi-omics approaches for biomarker discovery in early ovarian cancer diagnosis. EBioMedicine 79:104001. 10.1016/j.ebiom.2022.104001 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Availability Statement All data supporting the findings of this study are available within the paper. Articles from Journal of Ultrasound are provided here courtesy of Springer ACTIONS View on publisher site PDF (2.1 MB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Applications of artificial intelligence and machine learning models in the prognosis and diagnosis of ovarian cancer.#269583
• Data-driven modeling in healthcare: From generative AI in healthcare education to epidemiological and ovarian cancer biomarker modeling#348773
• Improving ovarian cancer risk assessment using a machine learning model developed on data from the Prostate Lung Colorectal Ovarian Cancer Screening Trial.#421253
• [A Transformer-based model using laboratory indicators efficiently differentiates ovarian cancer].#143170
• Prediction Model for Overall Survival in Patients with Epithelial Ovarian Cancer Undergoing Surgery and Systemic Therapy Based on the 2010–2021 SEER Database.#151447