Artificial Intelligence in Oncology: Current Landscape, Challenges, and Future Directions

In the paper

“AI is increasingly being applied to all aspects of oncology, where several applications are maturing beyond research and development to direct clinical integration.”

In the paper

“FDA-cleared AI algorithms have also been developed to aid in the interpretation of magnetic resonance imaging (MRI; DEN1700, RRID:SCR_012945) and breast ultrasound exams (K190442, K210670, P150043 RRID:SCR_012945).”

What the sources say

• Not checked yetcites DEN1700
• Not checked yetcites K190442
• Not checked yetcites K210670

In the paper

“the use of AI for breast cancer detection in X-ray mammography is one of the most active areas of development and clinical translation.”

In the paper

“A subsequent study presented an AI approach that encompasses both 2D and 3D mammography, with evidence of detecting breast cancer 1 to 2 years earlier than standard radiographic review for some cases (10).”

What the sources say

• Not checked yetcites [10] Lotter, 2021, Robust breast cancer detection in mammography and digital breast tomosynthesis using an annotation-efficient deep learning approach, Nat Med source

In the paper

“Several studies have also examined the use of AI to predict future breast cancer risk.”

In the paper

“A later study led by Google in 2020 garnered much attention in reporting evidence of expertlevel breast cancer detection by an AI system (9).”

What the sources say

• Not checked yetcites [9] McKinney, 2020, International evaluation of an AI system for breast cancer screening, Nature source

In the paper

“AI is being heavily applied all along the cancer care continuum using different types of data (Fig. 2).”

In the paper

“An early catalyst in this space was the Digital Mammography DREAM Challenge, an open AI competition where groups were able to benchmark and compare different AI approaches (8).”

What the sources say

• Not checked yetcites [8] Schaffter, 2020, Evaluation of combined artificial intelligence and radiologist assessment to interpret screening mammograms, JAMA Netw Open source

In the paper

“Raciti and colleagues (60) reported on Paige Prostate Alpha, a commercial decision support system trained to detect prostate cancer from core-needle biopsies.”

What the sources say

• Not checked yetcites [60] Raciti, 2020, Novel artificial intelligence system increases the detection of prostate cancer in whole slide images of core needle biopsies, Mod Pathol source

In the paper

“In the context of colorectal cancer screening, deep learning has been applied extensively to colonoscopy images and video.”

In the paper

“multiple randomized controlled trials have also reported promising results (16)(17)(18)(19)(20)(21)(22).”

What the sources say

• Not checked yetcites [16] Wang, 2020, Effect of a deep-learning computer-aided detection system on adenoma detection during colonoscopy (CADe-DB trial): a double-blind randomised study, Lancet Gastroenterol Hepatol source
• Not checked yetcites [17] Repici, 2020, Efficacy of real-time computer-aided detection of colorectal neoplasia in a randomized trial, Gastroenterology source
• Not checked yetcites [18] Karsenti, 2023, Effect of real-time computer-aided detection of colorectal adenoma in routine colonoscopy (COLO-GENIUS): a single-centre randomised controlled trial, Lancet Gastroenterol Hepatol source

In the paper

“Wang and colleagues (58) have developed DeepGrade, a deep learning model that aims to improve on the Nottingham grading system.”

What the sources say

• Not checked yetcites [58] Wang, 2022, Improved breast cancer histological grading using deep learning, Ann Oncol source

In the paper

“multiple groups have reported on Gleason scoring predictions from prostate cancer H&E images (62)(63)(64).”

What the sources say

• Not checked yetcites [62] Nagpal, 2019, Development and validation of a deep learning algorithm for improving Gleason scoring of prostate cancer, NPJ Digit Med source
• Not checked yetcites [63] Bulten, 2020, Automated deep-learning system for Gleason grading of prostate cancer using biopsies: a diagnostic study, Lancet Oncol source
• Not checked yetcites [64] Ström, 2020, Artificial intelligence for diagnosis and grading of prostate cancer in biopsies: a population-based, diagnostic study, Lancet Oncol source

In the paper

“This evaluation includes comparison with a reference panel of pathologists working on the same data set with predefined time constraints designed to mimic routine pathology workflows, where many of the models in the challenge exhibited higher performance than the panel of pathologists.”

What the sources say

• Not checked yetcites [45] Ehteshami Bejnordi, 2017, Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer, JAMA source

In the paper

“AI is being used to guide biopsies themselves through prostate segmentation and quantitative feature extraction on MRI images (K193283, RRID:SCR_012945).”

What the sources say

• Not checked yetcites K193283

In the paper

“Hamm and colleagues (38) presented an explainable AI model that goes beyond classification to output AI predictions in terms of PI-RADS criteria (39)”

What the sources say

• Not checked yetcites [38] Hamm, 2023, Interactive explainable deep learning model informs prostate cancer diagnosis at MRI, Radiology source
• Not checked yetcites [39] Turkbey, 2019, Prostate imaging reporting and data system Version 2.1: 2019 update of prostate imaging reporting and data system version 2, Eur Urol source

In the paper

“Challa and colleagues (47) evaluated the Visiopharm Integrator System metastasis AI algorithm in pathology workflows in a retrospective study at Ohio State University with promising results.”

What the sources say

• Not checked yetcites [47] Challa, 2023, Artificial Intelligence-aided diagnosis of breast cancer lymph node metastasis on histologic slides in a digital workflow, Mod Pathol source

In the paper

“Zhou and colleagues (15) have reported on CRCNet, a deep learning model for detecting the presence of colorectal cancer within endoscopy images, with reported high performance across three independent data sets.”

What the sources say

• Not checked yetcites [15] Zhou, 2020, Diagnostic evaluation of a deep learning model for optical diagnosis of colorectal cancer, Nat Commun source

In the paper

“CEmarked products exist for quantifying estrogen receptor, progesterone receptor, human epidermal growth factor receptor 2 (HER2), and Ki-67 status from digital pathology images (55-57).”

What the sources say

• Not checked yetcites [55] Aiforia Clinical Suite.
• Not checked yetcites [56] Mindpeak Solutions
• Not checked yetcites [57] Panakeia Products

In the paper

“several studies have demonstrated the potential of using AI to facilitate intraoperative diagnosis (e.g., during surgical resection) using label-free (40,41) and cryosectioned (42,43) tissue samples.”

What the sources say

• Not checked yetcites [40] You, 2019, Real-time intraoperative diagnosis by deep neural network driven multiphoton virtual histology, NPJ Precis Oncol source
• Not checked yetcites [41] Hollon, 2020, Near real-time intraoperative brain tumor diagnosis using stimulated Raman histology and deep neural networks, Nat Med source
• Not checked yetcites [42] Nasrallah, 2023, Machine learning for cryosection pathology predicts the 2021 WHO classification of glioma, Med source
• Not checked yetcites [43] Ozyoruk, 2022, A deep-learning model for transforming the style of tissue images from cryosectioned to formalin-fixed and paraffin-embedded, Nat Biomed Eng source

In the paper

“including the National Lung Cancer Screening Trial (NLST), which has become a key public data resource for machine learning research.”

What the sources say

• Not checked yetcites [27] National Lung Screening Trial Research Team, 2011, Reduced lung-cancer mortality with low-dose computed tomographic screening, N Engl J Med source
• Not checked yetcites [28] Field, 2021, Lung cancer mortality reduction by LDCT screening: UKLS randomised trial results and international meta-analysis, Lancet Reg Health Eur source

In the paper

“Several of the clinical studies have reported an increased adenoma detection rate with AI, a central criterion in evaluating the effectiveness of colonoscopy procedures.”

In the paper

“Multiple academic groups and companies have developed AI models for prostate cancer diagnosis and grading from H&E images.”

In the paper

“These AI risk prediction algorithms have been shown to outperform traditional risk models (13,14)”

What the sources say

• Not checked yetcites [13] Arasu, 2023, Comparison of mammography AI algorithms with a clinical risk model for 5-year breast cancer risk prediction: an observational study, Radiology source
• Not checked yetcites [14] Eriksson, 2023, Long-term performance of an image-based short-term risk model for breast cancer, J Clin Oncol source

In the paper

“Multiple companies have received FDA clearance or EU certification for computer-aided detection (CADe) systems in identifying polyps from colonoscopy images (K211951, K223473 RRID:SCR_012945)”

What the sources say

• Not checked yetcites K211951
• Not checked yetcites K223473

In the paper

“Current U.S. and European guidelines recommend prostate cancer detection via MRI prior to biopsy (33,34).”

What the sources say

• Not checked yetcites [33] European Association of Urology. Guidelines on prostate cancer—diagnostic evaluation
• Not checked yetcites [34] American College of Radiology. Prostate MRI model policy

In the paper

“a commercial prototype from Siemens was shown to increase the diagnostic accuracy of clinically significant prostate cancer in a retrospective reader study involving seven radiologists (37).”

What the sources say

• Not checked yetcites [37] Winkel, 2021, A novel deep learning based computer-aided diagnosis system improves the accuracy and efficiency of radiologists in reading biparametric magnetic resonance images of the prostate: results of a multireader, multicase study, Invest Radiol source

In the paper

“Multiple groups have now leveraged NLST imaging data to develop lung cancer detection models.”

In the paper

“Multiple commercial platforms for breast cancer lymph node metastasis are also clinically available or in development.”

In the paper

“However, most of these same studies also found an increased detection rate of diminutive adenomas (defined as polyps ≤5 mm in size), which are very rarely malignant (23,24).”

What the sources say

• Not checked yetcites [23] Ahmad, 2020, Deep learning for colorectal polyp detection: time for clinical implementation?, Lancet Gastroenterol Hepatol source
• Not checked yetcites [24] Misawa, 2023, Computer-aided detection in real-world colonoscopy: enhancing detection or offering false hope?, Lancet Gastroenterol Hepatol source

In the paper

“Nam and colleagues (31) have published promising results from a singlecenter, open-label randomized control trial that utilized an AI-based system for chest X-rays from the company Lunit, and reported that the AI arm had an improved detection rate of actionable nodules compared with the non-AI arm.”

What the sources say

• Not checked yetcites [31] Nam, 2023, AI improves nodule detection on chest radiographs in a health screening population: a randomized controlled trial, Radiology source

In the paper

“You and colleagues (40) have presented an approach that combines intraoperative microscopy with a deep learning model to classify malignant versus benign breast tissue.”

What the sources say

• Not checked yetcites [40] You, 2019, Real-time intraoperative diagnosis by deep neural network driven multiphoton virtual histology, NPJ Precis Oncol source

In the paper

“there are increasing efforts of using digital pathology images to guide treatment selection by predicting patient prognosis (73)(74)(75) and/or existing treatment biomarkers (76-79) using AI.”

What the sources say

• Not checked yetcites [73] Mobadersany, 2018, Predicting cancer outcomes from histology and genomics using convolutional networks, Proc Natl Acad Sci U S A source
• Not checked yetcites [74] Chen, 2021, Multimodal co-attention transformer for survival prediction in gigapixel whole slide images
• Not checked yetcites [75] Chen, 2022, Pan-cancer integrative histology-genomic analysis via multimodal deep learning, Cancer Cell source

In the paper

“DoMore Diagnostics has a CE-marked product that predicts colorectal cancer prognosis from H&E slides (94).”

What the sources say

• Not checked yetcites [94] DoMore Diagnostics: Histotype Px Colorectal

In the paper

“The relatively poor prognosis of NSCLC and the variable success of immunotherapies for NSCLC have motivated many efforts in using AI to predict prognosis, treatment response, and existing biomarkers for these patients.”

In the paper

“Lu and colleagues (69) used a weakly supervised deep learning method-dubbed clustering-constrained-attention multiple-instance learning (CLAM)-to predict subtypes in renal cell carcinoma and NSCLC.”

What the sources say

• Not checked yetcites [69] Lu, 2021, Data-efficient and weakly supervised computational pathology on whole-slide images, Nat Biomed Eng source

In the paper

“Vanguri and colleagues (108) demonstrated that a multimodal model that included CT images, digitized PD-L1 immunohistochemistry (IHC) slides, and molecular alteration data performed significantly better than unimodal models in predicting response to immunotherapy in NSCLC.”

What the sources say

• Not checked yetcites [108] Vanguri, 2022, Multimodal integration of radiology, pathology and genomics for prediction of response to PD-(L)1 blockade in patients with non-small cell lung cancer, Nat Cancer source

In the paper

“Along with directly predicting prognosis, there have been several studies using AI to predict previously identified prognostic/predictive biomarkers.”

In the paper

“A recent study by Ogier du Terrail and colleagues (83) presented a federated learning approach for predicting response to neoadjuvant chemotherapy in triple-negative breast cancer.”

What the sources say

• Not checked yetcites [83] Ogier du Terrail, 2023, Federated learning for predicting histological response to neoadjuvant chemotherapy in triple-negative breast cancer, Nat Med source

In the paper

“There are now several CE-marked products for Gleason grading (55,65,66).”

What the sources say

• Not checked yetcites [55] Aiforia Clinical Suite.
• Not checked yetcites [65] Ibex: Galen Prostate
• Not checked yetcites [66] Indica Labs: HALO Prostate AI.

In the paper

“Prognostic AI models for colorectal cancer have been developed for various data modalities, including MRI (88), histopathology (89-91), and multiplex imaging (92,93).”

What the sources say

• Not checked yetcites [88] Jiang, 2023, An MRI deep learning model predicts outcome in rectal cancer, Radiology source
• Not checked yetcites [89] Wulczyn, 2021, Interpretable survival prediction for colorectal cancer using deep learning, NPJ Digit Med source
• Not checked yetcites [90] Kather, 2019, Predicting survival from colorectal cancer histology slides using deep learning: a retrospective multicenter study, PLOS Med source
• Not checked yetcites [91] Bychkov, 2018, Deep learning based tissue analysis predicts outcome in colorectal cancer, Sci Rep source
• Not checked yetcites [92] Wu, 2022, Graph deep learning for the characterization of tumour microenvironments from spatial protein profiles in tissue specimens, Nat Biomed Eng source
• Not checked yetcites [93] Foersch, 2023, Multistain deep learning for prediction of prognosis and therapy response in colorectal cancer, Nat Med source

In the paper

“Numerous studies have demonstrated the potential for using AI to predict immunotherapy biomarkers from H&E images of patients with NSCLC, including tumor mutation burden (TMB; ref. 76), PD-L1 expression (105), and TILs (106, 107).”

What the sources say

• Not checked yetcites [76] Jain, 2020, Predicting tumour mutational burden from histopathological images using multiscale deep learning, Nat Mach Intell source
• Not checked yetcites [105] Sha, 2019, Multi-field-of-view deep learning model predicts non–small cell lung cancer programmed death-ligand 1 status from whole-slide hematoxylin and eosin images, J Pathol Inform source
• Not checked yetcites [106] Park, 2022, Artificial intelligence-powered spatial analysis of tumor-infiltrating lymphocytes as complementary biomarker for immune checkpoint inhibition in non-small-cell lung cancer, J Clin Oncol source
• Not checked yetcites [107] Rakaee, 2023, Association of machine learning-based assessment of tumor-infiltrating lymphocytes on standard histologic images with outcomes of immunotherapy in patients with NSCLC, JAMA Oncol source

In the paper

“This technology has been commercialized as the ArteraAI Prostate Test, which is now clinically available through a Clinical Laboratory Improvement Amendments (CLIA)-certified laboratory (115).”

What the sources say

• Not checked yetcites [115]

In the paper

“a line of work (97-99) has focused on developing and validating AI for predicting microsatellite instability (MSI), a feature associated with clinical outcomes and a biomarker for immunotherapy (100).”

What the sources say

• Not checked yetcites [97] Echle, 2022, Artificial intelligence for detection of microsatellite instability in colorectal cancer—a multicentric analysis of a pre-screening tool for clinical application, ESMO Open source
• Not checked yetcites [98] Echle, 2020, Clinical-grade detection of microsatellite instability in colorectal tumors by deep learning, Gastroenterology source
• Not checked yetcites [99] Kather, 2019, Deep learning can predict microsatellite instability directly from histology in gastrointestinal cancer, Nat Med source
• Not checked yetcites [100] Li, 2020, Microsatellite instability: a review of what the oncologist should know, Cancer Cell Int source

In the paper

“Other efforts have used AI on H&E images to directly predict TIL presence (85,86), a notable prognostic and predictive biomarker in breast cancer, and programmed cell death ligand 1 (PD-L1) expression status, a predictive biomarker for immunotherapies (87).”

What the sources say

• Not checked yetcites [85] Saltz, 2018, Spatial organization and molecular correlation of tumor-infiltrating lymphocytes using deep learning on pathology images, Cell Rep source
• Not checked yetcites [86] Binder, 2021, Morphological and molecular breast cancer profiling through explainable machine learning, Nat Mach Intell source
• Not checked yetcites [87] Shamai, 2022, Deep learning-based image analysis predicts PD-L1 status from H&E-stained histopathology images in breast cancer, Nat Commun source

In the paper

“Korbar and colleagues (72) developed a deep learning model to classify colorectal polyps into one of five common types based on H&E whole slide images.”

What the sources say

• Not checked yetcites [72] Korbar, 2017, Deep learning for classification of colorectal polyps on whole-slide images, J Pathol Inform source

In the paper

“Kartasalo and colleagues (67) have developed an AI model using H&E images to detect perineural invasion-a key prognostic marker associated with poor outcomes in prostate cancer.”

What the sources say

• Not checked yetcites [67] Kartasalo, 2022, Detection of perineural invasion in prostate needle biopsies with deep neural networks, Virchows Arch source

In the paper

“This application is beginning to be commercialized, such as the CE-marked product (101) developed by Owkin to predict MSI from H&E images.”

What the sources say

• Not checked yetcites [101] Owkin: MSIntuit CRC | AI Cancer Testing

In the paper

“Ozyoruk and colleagues (43) also demonstrated the potential of using AI to facilitate intraoperative subtyping of NSCLC by transforming cryosectioned tissue slides into the style of standard formalin-fixed and paraffinembedded slides.”

What the sources say

• Not checked yetcites [43] Ozyoruk, 2022, A deep-learning model for transforming the style of tissue images from cryosectioned to formalin-fixed and paraffin-embedded, Nat Biomed Eng source

In the paper

“AI company ArteraAI has developed and tested a multimodal AI approach (113, 114) that uses histology images and clinical data to predict prostate cancer prognosis.”

What the sources say

• Not checked yetcites [113] Esteva, 2022, Prostate cancer therapy personalization via multi-modal deep learning on randomized phase III clinical trials, NPJ Digit Med source
• Not checked yetcites [114] Spratt, 2023, Artificial intelligence predictive model for hormone therapy use in prostate cancer, NEJM Evid source

In the paper

“Histopathologic classification studies have been reported in colorectal cancer (70,71).”

What the sources say

• Not checked yetcites [70] Shaban, 2020, Context-aware convolutional neural network for grading of colorectal cancer histology images, IEEE Trans Med Imaging source
• Not checked yetcites [71] Yu, 2021, Accurate recognition of colorectal cancer with semi-supervised deep learning on pathological images, Nat Commun source

In the paper

“Efforts are also under way to use AI to improve prostate cancer prognostication.”

In the paper

“Deep learning methods are also having an impact on emerging novel therapeutic approaches involving molecular oncology.”

In the paper

“Robust, generalizable AI models require vast amounts of training data, ideally sourced from multiple, independent sources (160).”

What the sources say

• Not checked yetcites [160] Sheller, 2020, Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data, Sci Rep source

In the paper

“Of these, a particularly clinically advanced area is the use of deep learning to automatically contour Organs at Risk within clinical images such as CT, positron emission tomography and computed tomography (PET-CT), and MRI.”

In the paper

“This includes the use of AI in nearly every step of next-generation sequencing (NGS) processing and interpretation, including mutation identification (122, 123), artifact filtering (124), and driver gene identification (125).”

What the sources say

• Not checked yetcites [122] Khazeeva, 2022, DeNovoCNN: a deep learning approach to de novo variant calling in next-generation sequencing data, Nucleic Acids Res source
• Not checked yetcites [123] Sahraeian, 2022, Achieving robust somatic mutation detection with deep learning models derived from reference data sets of a cancer sample, Genome Biol source
• Not checked yetcites [124] Jongbloed, 2023, Machine learning-based somatic variant calling in cell-free DNA of metastatic breast cancer patients using large NGS panels, Sci Rep source
• Not checked yetcites [125] Luo, 2019, Predicting cancer driver genes based on somatic mutations using deep convolutional neural networks, Front Genet source

In the paper

“Several companies have developed AI-assisted scribing technologies to generate documentation based on transcriptions of a clinical encounter (155, 156).”

What the sources say

• Not checked yetcites [155] DeepScribe: AI-powered medical scribe
• Not checked yetcites [156] Nuance Communications: ambient clinical intelligence | automatically document care

In the paper

“It is even more acutely important for rare cancers, where individual sites may only treat a small number of such patients, and multisite data integration is necessary to establish meaningful scientific or clinical insights (161).”

What the sources say

• Not checked yetcites [161] Pati, 2022, Federated learning enables big data for rare cancer boundary detection, Nat Commun source

In the paper

“A study by Lu and colleagues (138) has also demonstrated the potential of using AI to predict tumor origin from routine histology slides.”

What the sources say

• Not checked yetcites [138] Lu, 2021, AI-based pathology predicts origins for cancers of unknown primary, Nature source

In the paper

“Multiple non-deep learning clinical trial matching applications now exist.”

In the paper

“Documentation and coding requirements in modern EMRs are extensive, imposing a significant time burden on clinicians (153)”

What the sources say

• Not checked yetcites [153] Arndt, 2017, Tethered to the EHR: primary care physician workload assessment using EHR event log data and time-motion observations, Ann Fam Med source

In the paper

“Multiple AI methods have been developed to perform these predictions.”

In the paper

“ChatGPT was shown to exhibit good performance at answering questions about cancer myths and misconceptions (159).”

What the sources say

• Not checked yetcites [159] Johnson, 2023, Using ChatGPT to evaluate cancer myths and misconceptions: artificial intelligence and cancer information, JNCI Cancer Spectr source

In the paper

“Multiple companies have now received FDA clearances and/or EU certification for these approaches (K201232, K223774, K211881 RRID:SCR_012945).”

What the sources say

• Not checked yetcites K201232
• Not checked yetcites K223774
• Not checked yetcites K211881

In the paper

“This is the standard, default approach used in most commercial and academic AI efforts to date.”

What the sources say

• Not checked yetcites [161] Pati, 2022, Federated learning enables big data for rare cancer boundary detection, Nat Commun source

In the paper

“clinical trial matching is usually framed as a natural language processing (NLP) challenge, where patient medical records-usually only available as unstructured clinical notes-are matched against inclusion and exclusion clinical trial criteria (148).”

What the sources say

• Not checked yetcites [148] von Itzstein, 2021, Application of information technology to clinical trial evaluation and enrollment: a review, JAMA Oncol source

In the paper

“Many patients with advanced cancer, for example, do not fully grasp the palliative intent of their treatment (158).”

What the sources say

• Not checked yetcites [158] Weeks, 2012, Patients’ expectations about effects of chemotherapy for advanced cancer, N Engl J Med source

In the paper

“and contributing to burnout (154)”

What the sources say

• Not checked yetcites [154] Robertson, 2017, Electronic health record effects on work-life balance and burnout within the i3 population collaborative, J Grad Med Educ source

In the paper

“The advent of LLMs now offers a renewed opportunity to improve NLP trial matching.”

In the paper

“Overall, these recent endeavors suggest that LLMs offer promising applications for clinical decision support in oncology”

In the paper

“An overarching challenge in AI in medicine is ensuring fairness in performance and use across populations (173).”

What the sources say

• Not checked yetcites [173] Chen, 2023, Algorithmic fairness in artificial intelligence for medicine and healthcare, Nat Biomed Eng source

In the paper

“The United States currently follows a centralized model of medical device regulation, overseen by the FDA.”

In the paper

“A study by Pati and colleagues (161) also used FL across 71 sites to predict tumor boundaries within multiparametric MRI scans of glioblastoma patients.”

What the sources say

• Not checked yetcites [161] Pati, 2022, Federated learning enables big data for rare cancer boundary detection, Nat Commun source

In the paper

“Multiple clinical trials have now also published results in the past two years, such as those described in the breast cancer and colorectal screening sections above.”

In the paper

“Conversely, AI offers opportunities to mitigate healthcare disparities that are already known to exist (174, 175).”

What the sources say

• Not checked yetcites [174] Chen, 2020, Treating health disparities with artificial intelligence, Nat Med source
• Not checked yetcites [175] Badal, 2023, Guiding principles for the responsible development of artificial intelligence tools for healthcare, Commun Med source

In the paper

“Beyond using general-purpose foundation models like GPT-4, there are increasing efforts in building domain-specific foundation models for healthcare (171), including pathology (ref.172; arXiv:2309.07778 and arXiv:2307.12914).”

What the sources say

• Not checked yetcites [171] Moor, 2023, Foundation models for generalist medical artificial intelligence, Nature source
• Not checked yetcites [172] Huang, 2023, A visual–language foundation model for pathology image analysis using medical Twitter, Nat Med source
• Not checked yetcites arXiv:2309.07778

In the paper

“One challenge in monitoring the progress of regulatory clearances is the lack of up-to-date, frequently refreshing databases focused on AI devices and their specific characteristics.”

In the paper

“The FDA has released a list of AI-enabled devices (193) on a yearly basis for the last few years, which is very valuable for the field, yet the rapid progress and value of aggregating more detailed characteristics creates a strong need for complementary solutions.”

What the sources say

• Not checked yetcites [193] FDA. Artificial intelligence and machine learning (AI/ML)-enabled medical devices

In the paper

“These include trials involving AI-based chatbots, such as use in genetic counseling (NCT05562778, NCT04354675 RRID:SCR_002309).”

What the sources say

• Not checked yetcites NCT05562778
• Not checked yetcites NCT04354675

In the paper

“To this end, the U.S. FDA is developing a pathway for a”

What the sources say

• Not checked yetcites [189] FDA. marketing submission recommendations for a predetermined change control plan for artificial intelligence/machine learning (AI/ML): enabled device software functions: draft guidance for industry and Food and Drug Administration staff

In the paper

“Similarly, the EU AI Act will require device manufactures to document their strategy for modifying and subsequently testing and validating AI medical devices (190).”

What the sources say

• Not checked yetcites [190] Vokinger, 2021, Regulating AI in medicine in the United States and Europe, Nat Mach Intell source

In the paper

“The majority of AI approaches described above have been unimodal, such as making predictions from a medical image alone.”

In the paper

“For instance, Chen and colleagues (75) developed an approach to fuse genomic data and histopathology images to predict survival in a pan-cancer cohort.”

What the sources say

• Not checked yetcites [75] Chen, 2022, Pan-cancer integrative histology-genomic analysis via multimodal deep learning, Cancer Cell source

In the paper

“The diversity of public data sets is a challenge across other oncology applications, including popular genomics data sets, such as TCGA (181) and GENIE (182).”

What the sources say

• Not checked yetcites [181] Spratt, 2016, Racial/ethnic disparities in genomic sequencing, JAMA Oncol source
• Not checked yetcites [182] Cheung, 2023, Racial and ethnic disparities in a real-world precision oncology data registry, NPJ Precis Oncol source

In the paper

“However, oncologists rely on the totality of information regarding a patient-including radiology images, pathology images, lab values, genomic data, family history, and prior clinical notes-to provide optimal care.”

In the paper

“Boehm and colleagues (170) used TCGA and an internal cohort to fuse genomic data, histopathology imaging, and radiology imaging to predict survival in high-grade serous ovarian cancer.”

What the sources say

• Not checked yetcites [170] Boehm, 2022, Multimodal data integration using machine learning improves risk stratification of high-grade serous ovarian cancer, Nat Cancer source

In the paper

“Muehlematter and colleagues (195) surveyed all approved AI medical devices in both the United States and the EU for the period 2015 to 2020 and also found that a majority of approved devices in both the United States and the EU were radiology focused.”

What the sources say

• Not checked yetcites [195] Muehlematter, 2021, Approval of artificial intelligence and machine learning-based medical devices in the USA and Europe (2015–20): a comparative analysis, Lancet Digit Health source

In the paper

“The FDA regulates AI applications under the umbrella of Software as a Medical Device (SaMD) with similar regulatory processes as other SaMD devices and non-AI algorithms (187).”

What the sources say

• Not checked yetcites [187] Harvey, 2020, How the FDA regulates AI, Acad Radiol source

In the paper

“Additionally, ArteraAI has developed a clinically available laboratory-developed test to predict prostate cancer prognosis by integrating digital pathology images with clinical metadata (115).”

What the sources say

• Not checked yetcites [115]

In the paper

“This lack of reimbursement also makes it challenging to assess the true extent of clinical integration, as traditional methods for measuring medical device use such as insurance claims data cannot be used except for a small number of applications (211).”

What the sources say

• Not checked yetcites [211] Wu, 2023, Characterizing the clinical adoption of medical AI devices through U.S. insurance claims, NEJM AI source

In the paper

“In particular, there is a growing number of AI”

In the paper

“There are ongoing efforts to provide more structured explainability, such as representing AI predictions in terms of human-interpretable concepts (205-207) and language-based explanations (medRxiv 2023.06.07.23291119; refs.208, 209).”

What the sources say

• Not checked yetcites [205] Kim, 2018, Interpretability beyond feature attribution: quantitative testing with concept activation vectors (TCAV)
• Not checked yetcites [206] Koh, 2020, Concept bottleneck models.
• Not checked yetcites [207] Yuksekgonul, 2023, Post-hoc concept bottleneck models. In:
• Not checked yetcites medRxiv 2023.06.07.23291119
• Not checked yetcites [208] Agarwal, 2023, Representing visual classification as a linear combination of words

In the paper

“The EQUATOR (Enhancing the QUAlity and Transparency Of health Research) Network currently provides a growing set of reporting guidelines specific to AI in healthcare and oncology (197).”

What the sources say

• Not checked yetcites [197] Simera, 2010, Transparent and accurate reporting increases reliability, utility, and impact of your research: reporting guidelines and the EQUATOR Network, BMC Med source

In the paper

“To this end, explainability is an often-cited challenge for”

What the sources say

• Not checked yetcites [203] Rudin, 2019, Stop explaining black box machine learning models for high-stakes decisions and use interpretable models instead, Nat Mach Intell source
• Not checked yetcites [204] van der Velden, 2022, Explainable artificial intelligence (XAI) in deep learning-based medical image analysis, Med Image Anal source

In the paper

“Rather than having each AI application interface separately with clinical systems, these platforms, such as those developed by Nvidia (212) and CARPL (213), serve as a single interface and orchestrate the execution of multiple AI algorithms based on rule-based logic.”

What the sources say

• Not checked yetcites [212] NVIDIA Clara supercharges next gen medical devices
• Not checked yetcites [213] CARPL Platform

In the paper

“with a recent report indicating that the scope of explainability offered in FDA-cleared devices for medical imaging is often limited (medRxiv 2023.11.28.23299132).”

What the sources say

• Not checked yetcites medRxiv 2023.11.28.23299132