ROLE OF ARTIFICIAL INTELLIGENCE IN POST- TREATMENT CHALLENGES IN CANCER PATIENTS

171 CHAPTER-16 ROLE OF ARTIFICIAL INTELLIGENCE IN POSTTREATMENT CHALLENGES IN CANCER PATIENTS Swati Singh Uttaranchal School of Computing Sciences Uttaranchal University, Dehradun, India Sonal Sharma Uttaranchal School of Computing Sciences Uttaranchal University, Dehradun, India Deepak Bhatt Uttaranchal School of Computing Sciences Uttaranchal University, Dehradun, India Abstract Cancer is a biological and lifestyle disease growing significantly in all the regions of the world. Global cancer statistics 2022 demographicbased predictions indicates that the number of new cases of cancer will reach 35 million by 2050. Recent advancements in Artificial Intelligence (AI) offers promising solutions to address early detection, classification of cancer, prediction, advanced treatment regimens and planning, prevention from reoccurrence, and analysis of genes variations responsible for causing cancer. These advancements are highly helpful throughout the treatment journey for the oncologists as well as the patients. But in this journey patients often face numerous physical and mental challenges right after completing their primary treatment. The list of these challenges includes most prominent side effects like fatigue, stress, sleep disturbances, sleep latency, insomnia, joints pain, lymphedema, mood swings, anger issues, depression, various allergies, body pain, and many more. These side effects persist even after many years of completion of the treatment and cause deterioration in their quality of life. AI driven technologies can provide significant solutions to address these challenges effectively by providing the predictive models for early detection of these issues, while personalized rehabilitation plans based on machine learning and deep learning algorithms can assist in managing longterm side effects. This chapter explores the key post-treatment challenges faced by cancer patients and current innovative AI-driven solutions to mitigate these issues.Additionally, it provides the recommendations to improve patient’s health by providing home, hospital, and societydriven solutions. Keywordsartificial intelligence, cancer, posttreatment issues, physical issues, mental issues 1. Introduction The journey for individuals diagnosed with cancer extends significantly beyond the initial diagnosis and the completion of active treatment. A rapidly expanding 172 global population of cancer survivors faces a complex array of post-treatment challenges, encompassing the management of long-term physical and psychological side effects, vigilant monitoring for disease recurrence, addressing emotional distress, and striving for an enhanced overall quality of life. Conventional follow-up care, while foundational, frequently encounters limitations in its capacity for scalability, personalization, and the efficient processing of the vast and intricate datasets inherent in oncology. Artificial Intelligence (AI) and Machine Learning (ML) are emerging as profoundly transformative technologies, poised to redefine cancer care delivery by offering more precise, efficient, and patient-centred methodologies across the entire continuum of care, with a particular emphasis on the critical post-treatment phase(Khorsand, 2024).The increasing number of cancer survivors underscores an urgent and growing demand for innovative solutions to support their ongoing health and well-being. These post-treatment challenges are inherently multifaceted, spanning physical, psychological, and social dimensions that significantly impact a survivor's daily life and long-term prognosis. The integration of AI with medical science represents a groundbreaking shift towards healthcare models that are not only more precise but also deeply patient-centric. This technological convergence holds substantial promise for enhancing patient management by leveraging advanced computational tools and the accelerating digitization of healthcare data(Swarnkar et al., 2025). The expansion of AI into post-treatment care marks a fundamental change in how survivorship is approached. Historically, cancer care has often been reactive and generalized, responding to symptoms as they arise and applying broad treatment protocols. The current trajectory, however, points towards proactive, highly personalized survivorship strategies. This profound transformation is fundamentally enabled by two key shifts: the advent of sophisticated computational tools, particularly deep learning, and the comprehensive digitization of healthcare data. Electronic Health Records (EHRs), advanced radiology and pathology images, and standardized genomic profiling are generating an unprecedented volume of high-quality, longitudinal data to best support the objectives of cancer care and costeffectiveness(Yu, 2011). AI can effectively process and analyze this immense data reservoir allows for the development of nuanced, individualized patient management plans. This capacity is particularly vital in the post-treatment phase, where individual variability in response to therapy, the manifestation of long-term side effects, and the precise risk of recurrence become paramount considerations for optimizing patient outcomes. Figure 1 represents the transformation of cancer care by utilizing artificial intelligence. 173 Figure 1. Funnel approach representing fourstage process of transforming cancer care using AI. 2. Literature Review: AI for Enhanced Symptom Monitoring and Management Effective monitoring and proactive management of symptoms are paramount for improving the quality of life and long-term outcomes for cancer survivors. AIdriven tools are proving to be indispensable in predicting, tracking, and alleviating the diverse spectrum of physical and psychological symptoms that patients often experience in the post-treatment period. A comprehensive systematic review(Vakili et al., 2024) encompassing 41 studies predominantly published between 2021 and 2023, has extensively explored the application of AI in symptom monitoring for adult cancer survivors. This review highlighted that machine learning algorithms were the most frequently employed AI methods, appearing in 43.9% of the studies. Following this, natural language processing (NLP) was utilized in 29.3% of studies, AI-driven chatbots in 17.1%, and decision support tools in 9.8%. The data inputs commonly fed into these AI algorithms included textual data, patient-reported symptoms, and physiological measurements. Among the myriad symptoms, pain was the most extensively examined, addressed in 34.2% of the studies, while fatigue and nausea were each investigated in 17.1% of the studies. The overarching trend observed was a clear increase in the adoption of AI technology for both the prediction and ongoing monitoring of cancer-related symptoms. A review was conducted (Zeinali et al., 2024) that stated, machine learning algorithms are specifically designed to predict the onset and progression of various cancer symptoms by identifying significant contributing factors. Among the most prevalent ML algorithms identified for symptom prediction are Logistic 174 Regression (used in 17% of studies), Random Forest (13%), Artificial Neural Networks (9%), and Decision Trees (9%). These models leverage a wide array of demographic features, such as age, gender, body mass index (BMI), income, education level, marital status, and zip code-level poverty. They also incorporate extensive clinical characteristics, including smoking and alcohol use, initial diagnosis, cancer stage, course, and site, specific treatment types and numbers, chemotherapy and radiotherapy details, the presence of chronic diseases and comorbidities, physical and psychological factors, care fragmentation, polypharmacy, hormone levels, physical activity, diet, heart rate, and social support factors. For instance, specific predictors for xerostomia include age, gender, treatment type, and hypertension. For pain, factors such as age, BMI, cancer site and stage, and psychological factors are highly significant. In the context of depression, socioeconomic factors and comorbidities play a crucial role, while for fatigue, existing fatigue levels, cancer site, and sleep disturbances are key indicators. AI models are also being specifically developed to predict debilitating symptoms such as cancer cachexia, a condition characterized by significant muscle wasting. An AI model known as SMAART-AI, for example, integrates imaging scans, specifically computed tomography (CT) scans for quantifying skeletal musclewith routine clinical data, including tumour stage, patient demographics, weight, height, laboratory results, and clinical notes, to predict the likelihood of cachexia. The early detection of cachexia through such AI-driven approaches is critical, as it enables the timely initiation of lifestyle and pharmacological interventions that can help slow muscle wasting, improve metabolic function, and significantly enhance the patient's quality of life(Pancholi, 2025). The effectiveness of AI in symptom monitoring stems from its unparalleled ability to process and identify intricate patterns within heterogeneous and high-volume data. This includes textual data from clinical notes, patient-reported information, and physiological measurements. Such a comprehensive analysis would be exceedingly time-consuming or even impossible for human clinicians to perform manually. This capability allows for earlier, more personalized, and proactive interventions for a wide range of symptoms, including pain, fatigue, nausea, and cachexia, thereby directly improving patient outcomes and overall quality of life(Vakili et al., 2024). The ongoing evolution towards advanced NLP models, such as transformer-based architectures, further enhances this capability by extracting nuanced understandings from unstructured clinical notes and patient narratives. This allows AI to capture the "emotional content" and "daily challenges" that are often central to a patient's experience but might be overlooked in structured data collection, leading to more targeted and holistic care(Voigt et al., 2025). Natural Language Processing (NLP) plays a pivotal role in analyzing PatientReported Outcomes (PROs) embedded within unstructured clinical notes in Electronic Health Records (EHRs). This capability offers invaluable insights that 175 extend beyond the limitations of structured data(Bilal, Hamza and Malik, 2024). NLP tasks involve a multi-step process. This typically begins with preprocessing, which includes annotation, removal of stop-words, Part-of-Speech (POS) tagging, entity linking, normalization, and lemmatization or stemming. Following this, feature extraction and representation techniques are applied, such as detecting affirmation/negation, employing rule-based NLP, Named Entity Recognition (NER), N-gram analysis, Word Embedding (e.g., Word2vec, GloVe), Topic Modelling (e.g., Latent Dirichlet Allocation), and Sentiment Analysis. Finally, these features are used for model development, utilizing both traditional machine learning (e.g., Logistic Regression, SVM, Random Forest) and neural machine learning approaches (e.g., Artificial Neural Networks, transformer-based language models like BERT)(Sim et al., 2024). The advantages of NLP are substantial as it converts large volumes of unstructured PRO data into practical clinical utilities, significantly reducing the labour-intensive task of manual review and enhancing efficiency. This approach holds the potential to complement or even replace traditional PRO surveys in busy clinical settings, thereby improving the feasibility of PRO collection for both research and clinical applications. NLP enables a comprehensive assessment of PROs, facilitates risk prediction and stratification, and allows for the investigation of associations between PROs and clinical outcomes(Sim et al., 2024). A notable trend in cancer research is the increasing adoption of NLP applications, with a clear shift from older rule-based methods to more advanced machine learning techniques, particularly transformer-based models, reflecting the growing sophistication in extracting complex information from clinical text(Bilal, Hamza and Malik, 2024). The successful integration of AI for symptom monitoring requires substantial investment in robust data infrastructure and the standardization of data capture methodologies. Without these foundational elements, the full potential of AI to deliver personalized and timely care, and to alleviate the burden on clinicians, cannot be fully realized(Vakili et al., 2024). This underscores the necessity of a systemic approach to AI adoption, one that extends beyond the development of isolated tools to encompass comprehensive infrastructure development and seamless integration into existing clinical workflows. Figure 2 represents AI driven symptom monitoring process and key AI methods and applications in cancer care symptom monitoring in Table 1. 176 Figure 2. AI driven symptom monitoring process. Table 1: Key AI Methods and Applications in Cancer Symptom Monitoring AI Method Primary Applicat ion Data Inputs Common Symptoms Addressed Key Contribu tions to Care Releva nt Source s Machine Learning Algorithms (e.g., Logistic Regression , Random Forest, ANN, Decision Trees, XGBoost) Predicti on and monitori ng of physical and psychol ogical sympto ms Textual data, patientreported symptom s, physiolo gic measure ments, demogra phic features, clinical character istics (treatmen t type, cancer site/stage , Pain, Fatigue, Nausea, Xerostomia , Depression, Cancer Cachexia Identifie s complex , nonlinear relations hips between diverse patient factors and sympto m develop ment; enables proactiv e intervent ions and (Vakili et al., 2024) 177 comorbid ities, psycholo gical factors) personal ized care. Natural Language Processing (NLP) Analysis of unstruct ured patientreported outcome s (PROs) from EHRs and patient forums Free-text clinical notes, patient forum posts Emotional content, daily challenges, concerns about test results/recu rrence, mindsetrelated attitudes Converts qualitati ve data into measura ble insights; captures nuances missed by structure d data; reduces manual review burden, enhancin g understa nding of patient experien ce. (Voigt et al., 2025) AIdrivenC hatbots Providin g psychol ogical support and informat ion Patient interactio ns (textual data) Anxiety, Stress, Unshared personal concerns Offers 24/7 accessibl e, nonjudgmen tal support; enhance s treatmen t (Khors and, 2024) 178 motivati on; comple ments tradition al mental health services by filling access gaps. Decision Support Tools Guiding clinical decision s related to sympto m manage ment Clinical data, patient data General symptom manageme nt, risk stratificatio n Provides tailored informat ion and assistanc e in managin g treatmen t plans; aids timely clinical decision s for improve d patient outcome s. (Khors and, 2024) 3. AI in Recurrence Prediction and Risk Stratification One of the most profound anxieties for cancer survivors is the potential for disease recurrence. Artificial intelligence, particularly through advanced machine learning and deep learning methodologies, is demonstrating remarkable capabilities in accurately predicting recurrence and precisely stratifying patient risk, thereby moving significantly beyond the limitations of traditional prognostic methods.In the context of breast cancer, AI applications are actively analyzing extensive datasets to predict recurrence, paving the way for highly personalized medical treatment strategies. Methods such as Support Vector Machines (SVM) 179 and Neural Networks, especially when applied to combined clinical and imaging datasets, exhibit substantial potential for enhancing prediction accuracy. SVMs prove particularly effective with high-dimensional clinical data, while Neural Networks excel in the analysis of genetic and molecular information. Despite these advancements, challenges persist, including issues related to dataset diversity, sample size limitations, and the need for standardized evaluation protocols, all of which underscore the necessity for continued rigorous validation before widespread clinical adoption(Silveira, Silva and Lima, 2025).A study (Pourakbar et al., 2025) for lung cancer stated that AI and ML models are integrating genomic biomarkers, clinical data, and imaging features to predict recurrence with superior accuracy compared to conventional methods. A variety of AI models are employed, including SVM, Gradient Boosting, XGBoost, Regression Models (such as LASSO and Cox), and sophisticated Deep Learning models like Genotype-Guided Radiomics (GGR), BPN-ALO, and the IBPGNET framework. Key genomic biomarkers that enhance these predictions include PDIA3, MYH11, SMARCA4, TP53, Fraction of Genome Altered (FGA), CpG methylation markers, specific 12-gene signatures, immune-related markers (e.g., FOXP3, PD-L1 on tumour-infiltrating lymphocytes (TILs), STK11, KEAP1), and multi-omics data such as copy number variations (CNVs), single nucleotide variants (SNVs), and long non-coding RNAs (lncRNAs). These AI models have achieved Area Under the Curve (AUC) values ranging from 0.73 to 0.92, significantly outperforming traditional TNM staging, which typically yields an AUC of 0.61. Multi-modal approaches, which integrate gene expression, radiomics, and clinical data, have consistently demonstrated improved accuracy. However, significant challenges remain, including concerns about data quality and availability (often relying on retrospective datasets and expensive genetic data), the risk of overfitting, the inherent "black box" nature of many AI models which limits interpretability, ethical considerations (such as privacy, bias, and fairness), limitations due to small sample sizes and data heterogeneity, insufficient external validation, complexities in multi-omics integration, high costs, and a general lack of standardization across studies. In cervical cancer, deep learning models are being developed to predict postoperative recurrence risk by leveraging multiparametric MRI images. The novel ConvXGB model exemplifies this, combining a Convolutional Neural Network (CNN), such as ResNet18, with eXtreme Gradient Boost (XGBoost). The image processing for this model is meticulous, involving precise image acquisition (T2-weighted imaging with fat suppression, diffusion-weighted imaging, and apparent diffusion coefficient maps), accurate segmentation of regions of interest (ROIs), and thorough data preprocessing (including cropping, gray-scaling, resampling, data augmentation, and normalization).ConvXGB has demonstrated excellent predictive performance, achieving high AUCs for 1-year (0.872) and 3-year (0.882) recurrence-free survival (RFS) in test cohorts. This performance significantly surpasses that of traditional clinical models and 186 misinformation or directs critical care decisions(Hantel et al., 2022). The imperative for human accountability in AI-associated tasks in oncology is therefore paramount(Shah and Karlovits, 2025). Many of these challenges are deeply interconnected. For instance, if the underlying data used to train an AI model is biased or heterogeneous (an implementation barrier), the AI's predictions will inherently be biased and potentially inaccurate (an ethical concern). Similarly, the "black box" nature of some AI models (an ethical concern) directly contributes to a lack of trust among clinicians, which in turn hinders clinical validation and seamless integration into existing workflows (implementation barriers)(Hantel et al., 2022). This complex interplay signifies that a holistic, multi-stakeholder approach is required to address these issues effectively, rather than attempting to solve each problem in isolation. (ii)Regulatory Landscape The regulatory environment for AI in oncology is still evolving and presents its own set of challenges. There is a notable lack of standardized guidelines, such as those from the National Comprehensive Cancer Network (NCCN), specifically for AI tools. This absence can inadvertently permit the use of models that may have comparatively low accuracy or contain racial biases, potentially leading to suboptimal patient care.An inconsistent regulatory approach is also evident. A discrepancy exists between the demonstrated efficacy of emergent AI tools and their exclusion from established guidelines, which can result in the continued reliance on outdated methodologies. Furthermore, many non-AI tools currently utilized in healthcare are not subject to FDA regulation, while AI tools are increasingly expected to be, creating an uneven and potentially confusing regulatory playing field. There is a clear and pressing need to apply the same rigorous regulatory scrutiny to AI tools as to all other technologies developed in oncology to ensure patient safety and the delivery of optimal care. (iii)Implementation Barriers Beyond ethical and regulatory concerns, several practical barriers impede the widespread implementation of AI in post-treatment cancer care. Data quality and heterogeneity are significant issues. Healthcare data are frequently recorded and stored in disparate, idiosyncratic, and unstructured formats across different systems. This inherent heterogeneity means that AI algorithms developed using data from one system may perform poorly when applied to data from another, necessitating comprehensive standardization of terminology and data collection methods(Pourakbar et al., 2025).The burdens of data management and collection are also substantial. AI solutions often require data from multiple sources, including patient-level EHR data and medical knowledge databases. This exacerbates the administrative and financial costs associated with managing and maintaining diverse data types, which can be prohibitive for smaller clinical practices(Khorsand, 2024). Moreover, the increased demand for data collection could contribute to clinician burnout(Chua et al., 2021).Insufficient clinical 187 validation is another critical barrier. While many AI models demonstrate strong performance in internal validation, a significant proportion lack sufficient external validation across diverse patient populations and healthcare systems. This raises concerns about their generalizability and real-world reliability, hindering trust and adoption among clinicians(Pourakbar et al., 2025).Workflow and user-design challenges are also prevalent. Integrating AI tools seamlessly into existing clinical workflows presents considerable difficulties. This requires careful user-centred design considerations and comprehensive training for the oncology workforce to ensure that AI tools are intuitive, efficient, and genuinely enhance, rather than disrupt, clinical practice(Vakili et al., 2024).Finally, the cost and accessibility of advanced AI-driven treatments and genomic profiling technologies can limit their availability, particularly in developing regions or low-resource settings. This contributes to existing healthcare disparities, making it difficult for all patients to benefit from these technological advancements(Pourakbar et al., 2025). The rapid pace of scientific discovery and the dynamic nature of medical knowledge also mean that AI models must be continuously updated to remain relevant and effective(Chua et al., 2021). The current state of AI implementation in oncology highlights a critical gap between technological capability and practical, ethical, and equitable deployment. The emphasis on "human accountability" (Shah and Karlovits, 2025) and the consistent message that AI should "complement doctors, not replace them" suggests that the future of AI in cancer care is not about full automation, but about a symbiotic relationship where AI enhances human decision-making while preserving the essential human touch. This necessitates a proactive, collaborative effort involving researchers, clinicians, policymakers, and patients to co-create AI solutions that are not only effective but also trustworthy, transparent, and equitable. Table 2 represents the ethical and implementation challenges for AI in post treatment cancer care with mitigation strategies. Table 2: Ethical and Implementation Challenges of AI in Post-Treatment Cancer Care with Mitigation Strategies Challenge Category Specific Challenge Impact on PostTreatment Care Mitigation Strategies Releva nt Sources Ethical Implicatio ns Data Bias & Incomplet eness Inaccurate predictions of recurrence/side effects; exacerbation of health disparities; misinformed Rigorous data governance; diverse, representativ e datasets; bias detection & (Pourak bar et al., 2025) 188 patient management. mitigation algorithms; transparent data collection practices. Privacy & Data Security Misuse of sensitive patient information; erosion of patient trust; legal liabilities. Robust data anonymizati on/deidentification ; secure storage (e.g., cloud-based systems like Azure); explicit patient consent protocols; adherence to GDPR/HIPA A. (Hasei et al., 2025) Transpare ncy & Accountab ility ("Black Box") Lack of understanding of AI decisions; difficulty in validating predictions; unclear responsibility for errors. Explainable AI (XAI) techniques (e.g., GradCAM, SHAP); public disclosure of algorithms (where appropriate); clear human accountabilit y frameworks; regulatory oversight for transparency. (Pourak bar et al., 2025) 189 Implement ation Barriers Data Heterogen eity & Manageme nt Burden Inability of AI models to generalize across systems; high administrative/fi nancial costs for data integration; clinician burnout. Standardizati on of terminology (e.g., mCODE); interoperable EHR systems; automated data collection (e.g., PROMs); investment in robust data infrastructure . (Vakili et al., 2024) Insufficien t Clinical Validation AI models performing well in training but failing in realworld settings; lack of trust among clinicians. Multiinstitutional, prospective studies; diverse, large sample sizes; rigorous external validation; standardized reporting practices for AI research. (Silveir a, Silva and Lima, 2025) Workflow & UserDesign Challenges Difficulty integrating AI into existing clinical routines; resistance from healthcare professionals. Usercentered design principles; iterative development with clinician feedback; (Vakili et al., 2024) 190 comprehensi ve training and education for the oncology workforce; pilot programs. Cost & Accessibili ty High cost of AI tools and associated technologies limiting access, particularly in low-resource settings. Development of costeffective AI solutions; publicprivate partnerships; equitable distribution strategies; focus on open-source tools where feasible. (Pourak bar et al., 2025) 7. Future Directions The integration of Artificial Intelligence (AI) in post-treatment cancer care is still in its nascent stages, with enormous untapped potential to transform survivorship outcomes. Moving forward, a crucial direction lies in the development of more robust, ethically grounded, and explainable AI models that not only ensure high predictive accuracy but also build trust among clinicians and patients. There is a pressing need to invest in the creation of large-scale, diverse, and representative datasets that minimize algorithmic bias and facilitate generalizable predictions across varied populations. Additionally, the expansion of multi-modal AI systemsthose integrating clinical, genomic, imaging, and patient-reported datawill significantly enhance precision in recurrence prediction, side effect management, and holistic patient care. Another promising frontier involves enhancing AIenabled telemedicine platforms and empathetic generative AI chatbots that support emotional well-being, especially in underserved and remote areas. Future research should also focus on establishing standardized regulatory frameworks that ensure safety, transparency, and accountability of AI-driven interventions. Encouraging cross-disciplinary collaboration among oncologists, data scientists, psychologists, and policymakers will be critical to design usercentric AI tools that can be seamlessly integrated into existing clinical workflows 191 without adding burdens to healthcare professionals. Moreover, the shift toward Explainable AI (XAI) and human-AI collaborative modelswhere AI serves as a decision support tool rather than a replacement for clinicianswill likely shape the ethical and practical future of cancer care. Ensuring affordability and accessibility of these technologies, particularly in low-resource settings, must remain a central priority to avoid widening health disparities. 8. Conclusion AI offers a transformative opportunity to redefine the landscape of post-treatment cancer care by enabling predictive, personalized, and proactive interventions. From symptom monitoring and recurrence risk stratification to managing longterm physical and psychological side effects, AI has demonstrated significant potential to enhance patient outcomes and quality of life. However, the path to widespread adoption is laden with challengesranging from ethical dilemmas and regulatory uncertainty to implementation barriers and data limitations. To truly realize the promise of AI in oncology, a holistic and collaborative approach is imperativeone that places patients at the centre of innovation, upholds ethical standards, and builds a healthcare ecosystem where human empathy and artificial intelligence work in harmony. References 1. AI in Oncology: Managing & Monitoring Cancer Treatment, accessed on June 14, 2025, https://elementtechnologies.net/2023/07/26/ai-in-oncologyadvancements-in-managing-and-monitoring-cancer-treatment-side-effects/ 2. Bilal, M., Hamza, A. and Malik, N. (2024) ‘Natural Language Processing for Analyzing Electronic Health Records and Clinical Notes in Cancer Research: A Review’, arXiv, October. Available at: https://doi.org/https://arxiv.org/html/2410.22180v1. 3. Chua, I.S. et al. (2021) ‘Artificial intelligence in oncology: Path to implementation’, Cancer Medicine, 10(70). Available at: https://doi.org/10.1002/cam4.3935. 4. Considerations for Equity and Regulations for AI Tools in Oncology - CancerNetwork, accessed on June 14, 2025, https://www.cancernetwork.com/view/considerations-for-equity-andregulations-for-ai-tools-in-oncology 5. Contemplating AI Assistance as a Cancer Survivor - Curetoday, accessed on June 14, 2025, https://www.curetoday.com/view/contemplating-aiassistance-as-a-cancer-survivor 6. Ethical Challenges and Limitations of AI in Cancer Treatment, accessed on June 14, 2025, https://vcarecancercenter.com/ethical-challenges-andlimitations-of-ai-in-cancer-treatment/ 7. Hantel, A. et al. (2022) ‘A Process Framework for Ethically Deploying Artificial Intelligence in Oncology’, Journal of Clinical Oncology, 40(34), pp. 3907–3911. Available at: https://doi.org/10.1200/JCO.22.01113. 192 8. Hasei, J. et al. (2025) ‘Empowering pediatric, adolescent, and young adult patients with cancer utilizing generative AI chatbots to reduce psychological burden and enhance treatment engagement: a pilot study’, Frontiers in Digital Health, 7(2025). Available at: https://doi.org/https://doi.org/10.3389/fdgth.2025.1543543. 9. Hilbers, D. et al. (2025) ‘Patient Attitudes Toward Artificial Intelligence in Cancer Care: A Scoping Review’, medRxiv : the preprint server for health sciences [Preprint]. Available at: https://doi.org/https://doi.org/10.1101/2025.03.15.25324029. 10. Improving Cancer Care With Artificial Intelligence | Blog | AACR, accessed on June 14, 2025, https://www.aacr.org/blog/2025/04/28/improving-cancercare-with-artificial-intelligence/ 11. Khorsand, M. (2024) AI Chatbots and Telemedicine in Cancer Care : Supporting Patients and Enhancing Communications. PreferPub. 12. Liu, Z. et al. (2025) ‘Predicting Recurrence in Locally Advanced Rectal Cancer Using Multitask Deep Learning and Multimodal MRI’, Radiology, 7(3). Available at: https://doi.org/https://doi.org/10.1148/rycan.240359. 13. Pancholi, N.J. (2025) ‘Improving Cancer Care With Artificial Intelligence’. Available at: https://www.aacr.org/blog/2025/04/28/improving-cancer-carewith-artificial-intelligence/. 14. Pourakbar, N. et al. (2025) ‘Effectiveness of Artificial Intelligence Models in Predicting Lung Cancer Recurrence: A Gene Biomarker-Driven Review’, Cancers, 17(11), p. 1892. Available at: https://doi.org/https://doi.org/10.3390/cancers17111892. 15. Shah, C. and Karlovits, S.M. (2025) ‘Artificial Intelligence in Oncology: Fulfilling Its Promise While Avoiding Its Peril’, JCO Oncology Practice [Preprint]. Available at: https://doi.org/https://doi.org/10.1200/OP-25-0007. 16. Silveira, J.A., Silva, A.R. da and Lima, M.Z.T. de (2025) ‘Harnessing artificial intelligence for predicting breast cancer recurrence: a systematic review of clinical and imaging data’, Discover Oncology, 8(16), p. 135. Available at: https://doi.org/10.1007/s12672-025-01908-6. 17. Sim, J. et al. (2024) ‘Using Natural Language Processing to Analyze Unstructured Patient-Reported Outcomes Data Derived from Electronic Health Records for Cancer Populations: A Systematic Review’, National Library of Medicine, 24(4), pp. 467–475. Available at: https://doi.org/10.1080/14737167.2024.2322664. 18. Swarnkar, S.K. et al. (2025) Artificial Intelligence Revolutionizing Cancer Care Precision Diagnosis and Patient-Centric Healthcare. Routledge. 19. Vakili, S.T. et al. (2024) ‘Application of Artificial Intelligence in Symptom Monitoring in Adult Cancer Survivorship: A Systematic Review’, JCO clinical cancer informatics, 8(December). Available at: https://doi.org/https://doi.org/10.1200/CCI.24.00119. 193 20. Voigt, K. et al. (2025) ‘A Machine Learning Approach Using Topic Modeling to Identify and Assess Experiences of Patients With Colorectal Cancer: Explorative Study’, JMIR Publications, 11. Available at: https://doi.org/10.2196/58834. 21. Wu, J. et al. (2025) ‘ConvXGB: A novel deep learning model to predict recurrence risk of early-stage cervical cancer following surgery using multiparametric MRI images’, Translational Oncology, 11(52). Available at: https://doi.org/10.1016/j.tranon.2025.102281. 22. Yu, P.P. (2011) ‘The Evolution of Oncology Electronic Health Records’, The Cancer Journal, 17(4), pp. 197–202. Available at: https://doi.org/10.1097/PPO.0b013e3182269629. 23. Zeinali, N. et al. (2024) ‘Machine Learning Approaches to Predict Symptoms in People With Cancer: Systematic Review’, JMIR Publications, 10(March). Available at: https://doi.org/10.2196/52322.