Page 127 - EJMO-9-1

Eurasian Journal of Medicine and
Oncology
Oncology care with AI chatbots and assistants

reviewing existing research on AI chatbots in healthcare, predictive models for chemotherapy and radiotherapy, and
with a focus on oncology, highlighting their applications, genomic data analysis, emphasizing their clinical impact,
benefits, and challenges. Section 3 introduces the proposed merits, limitations, and future enhancements.
chatbot algorithm, detailing the key components, such as The survey highlights significant advancements in using
NLP, ML models, and decision-support systems. In Section AI technologies across cancer types. For example, Esteva
4, a comparison with existing methods is made, evaluating et al. demonstrated how CNNs achieved dermatologist-
6
the proposed chatbots against current oncology chatbots, level accuracy in detecting skin cancer, enabling earlier
such as RITA, based on accuracy, patient engagement, and diagnosis and reduced diagnostic time. Similarly, successes
treatment outcomes. Finally, Section 5 offers a conclusion were reported in diabetic retinopathy detection, breast
7
and outlines future work, focused on enhancing chatbots cancer diagnosis, and lung cancer detection, where AI
10
8
capabilities, integrating them into healthcare systems, and models significantly outperformed traditional diagnostic
ensuring equitable access for diverse patient populations. methods. 10,12,16 In treatment planning, AI has been
9
2. Literature survey employed in radiotherapy and chemotherapy response
prediction, where its ability to optimize treatment plans
17
The literature survey aims to evaluate the current applications and personalized therapies has shown promise. AI
9,17
of AI in oncology, focusing on its transformative impact also plays a critical role in genomic data analysis and
23
on cancer detection, diagnosis, treatment planning, and tumor classification, aiding in individualized treatment
16
prognosis. The review included studies published between strategies. 20,24
2015 and 2023, ensuring relevance by adhering to specific Despite these advancements, there remain several
inclusion and exclusion criteria.
challenges. A recurring limitation is the need for large,
The inclusion criteria consisted of peer-reviewed high-quality datasets, as noted by Tran et al. in prostate
14
articles published in journals or conference proceedings cancer detection. Concerns about biases in training data
that explicitly focused on the use of AI, including chatbots and generalization to diverse populations further limit
and virtual assistants, in oncology care. Studies addressing the scalability of AI solutions. In addition, computational
key themes such as diagnostic accuracy, patient care demands and the lack of transparency in AI decision-
improvement, and treatment adherence were prioritized. making processes remain significant barriers to clinical
All included studies were published between 2015 and adoption. 11,13
2023 to capture the most recent advancements in the field. Future enhancements need to focus on integrating
The exclusion criteria were non-peer-reviewed articles, multimodal data – combining imaging, genomic, and
editorials, or opinion pieces, as well as studies unrelated clinical datasets – to improve diagnostic precision and
to oncology or AI-driven healthcare tools and non-English enable real-time treatment adaptations. The development
publications. The time frame of the review was limited to of adaptive AI systems for continuous patient monitoring
the past 8 years (2015 – 2023) to ensure the relevance and and personalized care pathways is a particularly promising
timeliness of the research.
direction. These observations align with the authors’
The review analyzed several key attributes: (i) Diagnostic expertise in oncology AI applications and are supported
accuracy, assessing AI models’ ability to detect or classify by emerging evidence. Quantitatively, the survey reveals
cancer, (ii) patient care impact, focusing on improvements that over 75% of studies have focused on image-based
in treatment adherence, emotional support, and patient diagnostics using deep learning algorithms, particularly
empowerment, and (iii) technological components, CNNs, with approximately 50% demonstrating improved
examining AI methodologies, such as NLP and ML diagnostic accuracy compared to traditional methods. 10,12,16
models. In addition, the review addressed challenges, This highlights the potential of AI to transform oncology
such as data privacy, system accuracy, and bias in AI care while underscoring the need to overcome limitations
applications, to provide a comprehensive understanding related to data quality, model generalization, and
of the current landscape in AI-driven oncology care. This computational requirements. Overall, while AI has
structured approach ensured that the survey captured shown substantial promise in advancing cancer care,
a comprehensive and up-to-date understanding of AI’s its widespread clinical adoption depends on addressing
role in transforming oncology care while identifying gaps these challenges and refining models to ensure robustness
and opportunities for further research. Table 1 provides across diverse patient populations. The potential for
a comprehensive overview of the 20 identified studies, personalized medicine and the seamless integration of AI
showcasing various innovations in AI for oncology. These with healthcare technologies offers exciting prospects for
include applications, such as AI-driven imaging analysis, improving oncology practices in the future.

Volume 9 Issue 1 (2025) 119 doi: 10.36922/ejmo.6251