The development of artificial intelligence (AI) technology in the field of medicine marks a new stage in the detection, diagnosis and treatment of cancer. With oncology still considered a global health concern, the application of AI technologies throughout the oncology care continuum has aided in achieving earlier detection, more accurate diagnoses and improved treatment outcomes. This chapter looks into the ways AI is transforming oncology, including advanced imaging techniques, predictive modeling, personalization of treatment protocols and automated rehabilitation systems. AI imaging analysis with deep learning algorithms is already achieving the diagnostic accuracy standards set by radiologists and pathologists, as well as detecting previously missed abnormalities in images. In addition, machine learning is now routinely used to analyze genomic information, making targeted therapies possible. AI is also involved in prediction modeling: these models can detect patients highly predisposed to developing cancer and predict how they will respond to treatment with unparalleled precision. In therapeutics, AI supports the clinical decision-making process by offering individualized treatment strategies for each patient and streamlining the processes involved in drug development and repurposing. AI systems also play a vital role in cancer rehabilitation: intelligent assistive devices and virtual support systems help cancer survivors with customized psychosocial and physical rehabilitation. This chapter looks into the state of incorporation of AI in oncology, describes major AI clinical innovations and applications, and outlines ethical as well as technical issues such as data confidentiality, bias in algorithms and subsystem integration. This analysis based on clinical studies, technology evaluations and expert opinions, describes the holistic impact of automation through AI in transforming cancer care, optimizing healthcare services and reducing work overload on physicians. Despite remarkable progress in the field of medicine, cancer continues to be one of the primary causes of global mortality, contributing millions to the death toll each year. In simple terms, cancer is defined as the uncontrollable growth and dissemination of abnormal cells that have the potential to invade and affect nearly all organ systems within a human body. The differences in the type of cancer, or what organ system gets affected, contribute to the very complex nature of cancer, making it one of the hardest diseases to diagnose, treat and manage (Bray et al. 2018). Additionally, with the integration of modern lifestyles and age, the prevalence of cancer is expected to rise, emphasizing the need for efficient solutions to relieve future strains on healthcare systems. With the inclusion of oncologists, pathologists, radiologists and surgeons, the cancer care multidisciplinary team is relatively specialized and integrates different aspects of cancer care. This approach has advanced cancer therapy significantly over the years; however, there are still numerous gaps. Patients may experience poor outcomes due to variations in the clinical decision-making, nondiagnostic procedures, treatment delays and limited access to specialists who provide high-quality care (Hanahan and Weinberg 2011). In this regard, AI technologies in oncology have recently emerged in the form of a powerful innovation capable of solving multi-faceted problems. The term AI is used to describe a large variety of options for processing data on a computer; this enables devices to complete tasks that are conventionally associated with human cognition, including learning, reasoning and problem solving. Within healthcare applications, technologies such as machine learning (ML), deep learning, natural language processing (NLP) and computer vision are subfields of AI that particularly stand out. These branches of study are able to process and analyze complex and vast troves of data including but not limited to medical images, pathology slides, genomic sequences and electronic health records (EHR) with exceptional speed and precision, often outpacing humans in particular tasks (World Health Organization 2023). AI is actively being woven into the entire cancer care spectrum, including early detection, diagnosis, treatment planning, prognosis and even post-treatment monitoring. Medical imaging is one of the fields where cancer care has benefitted through the incorporation of AI technology. AI, especially deep learning-based algorithms, has shown remarkable performance in detecting cancerous lesions from radiologic images such as mammograms, CT scans and MRIs (Siegel et al. 2020). These systems can act as aids to radiologists and avoid diagnostic oversights, improve consistency in interpretations and illuminate minute details that the naked eye may overlook. AI is also breaking new ground in pathology. AI-powered platforms can scan whole slide images to automatically classify tumor types, grade tumor malignancy and even classify tumors into certain molecular subtypes based on the histological architecture of the tissue. This not only helps in completing a diagnosis much faster, but also improves the quality of treatment through better tailoring of the individual treatment regimens. For example, AI tools can incorporate histology and genomic data to evaluate the likelihood of patients responding to targeted therapies or immunotherapy, and treating them with more precise cancer care (Duffy et al. 2013). AI is transforming many sectors, including treatment planning. Depending on relevant data points – such as the patient’s tumor features, existing illnesses and previous treatments – AI algorithms help clinicians devise tailored treatment strategies that achieve high efficacy while reducing adverse effects (Topol 2019). In radiation oncology, AI can streamline treatment planning to improve efficiency, consistency and quality in dose delivery. Additionally, AI-powered clinical decision support tools can analyze complex clinical databases, including clinical trials, guidelines and real-world data, to provide patient-specific evidence-based recommendations that integrate the patient’s medical history and unique characteristics (Esteva et al. 2017). Another crucial domain in cancer care where AI is used is prognosis – estimating the possible course and outcome of a cancer disease. Prognosis greatly benefits from modeling biological systems with large datasets, where ML algorithms provide accurate estimates for advancement, risk of recurrence and survival possibilities. These forecasts, alongside other critical insights, allow for greater preparedness within healthcare systems and facilitate timely intervention for patients (LeCun et al. 2015). In addition, AI aids the discovery of biomarkers and signatures for treatment resistance, leading to better therapeutic approaches and preemptive measures. AI technologies are also changing the post-treatment care and rehabilitation processes. AI-powered wearable sensors and mobile health applications can track and monitor a patient’s vital signs, physical activity and symptoms during recovery, ensuring that complications are recognized and clinically managed as early as possible. Patients can be engaged during the follow-up treatment through chatbots, virtual health assistants and NLP tools which can help provide personalized health information, emotional support and medication reminders (Shen et al. 2017). This is particularly useful in rural or disadvantaged regions where there is little access to tailored oncology care. Even with the new possibilities brought by AI, there are some issues that remain, including data privacy, algorithm transparency, trust in regulations and ethical issues surrounding AI’s role in cancer care. The implementation of AI in healthcare needs to be done carefully to ensure equitable safety for all users. There are also gaps in the quality of AI models; AI depends heavily on fresh, diverse data. Work needs to be done to ensure adequate representation of diverse, multi-ethnic populations in order to counter biases that may worsen health inequities (Sengupta et al. 2022). Educating and involving healthcare professionals is critical to ensuring AI tools are integrated properly and optimally aligned with clinical workflows. AI is one of the most powerful and versatile tools that has the potential to transform cancer care by improving the accuracy of diagnostics, personalizing treatment plans, forecasting outcomes and assisting in postoperative recovery (Rajkomar et al. 2018). With continued technological developments and the healthcare work force’s growing familiarity with AI tools, these systems are anticipated to be used routinely in oncology practices. Still, harnessing the benefits of AI to their maximum comes from the multidisciplinary collaboration between clinicians, data experts, ethicists and policy shapers to resolve the technological, ethical and pragmatic difficulties. Under proper policies and protective measures, AI can significantly enhance the efficacy and quality of cancer care, ultimately lessening the impact of this disease on the population worldwide (Reinkensmeyer et al. 2016). AI is transforming oncology in a very pronounced way, particularly concerning the use of medical imaging. In the past decade, deep learning, especially using convolutional neural networks (CNN), has changed the processes of capturing and interpreting medical images to improve cancer detection. Such AI models undergo supervised learning on large datasets of radiographic images through the identification of identifiable patterns and details which are often beyond the grasp of seasoned radiologists. Accordingly, diagnostic processes are being aided greatly by AI systems capable of image analyses with improved sensitivity and specificity (Esteva et al. 2017; Litjens et al. 2017). All radiographic techniques – X-ray, MRI, CT scan, PET scan and mammography – stand to gain immensely from the application of AI. For instance, AI algorithms in the field of breast cancer detection have been shown to detect tumors at early stages with an accuracy level comparable to or surpassing that of human experts. In the detection of lung cancer, AI algorithms have also been used for pulmonary nodule detection from low-dose CT scans with greater accuracy which reduced false negative rates to improve early detection of malignant tumors (Ardila et al. 2019; McKinney et al. 2020). AI not only helps with pattern detection, but also improves the standardization of evaluation by removing the variability brought about by human judgment. Standardization can be important in the context of low-resource settings where trained radiologists may not be readily available. Moreover, AI-based systems have the ability to process imaging data at greater speeds, enabling real-time diagnosis and drastically cutting the time needed between testing and the commencement of treatment (Hosny et al. 2018). In addition, these models are designed to improve themselves with feedback loops and more sophisticated mechanisms of continuous learning. Some systems today are able to identify images requiring further review, assign them a ranking for the likelihood of malignancy or even recommend additional imaging studies and biopsies based on available clinical information (Wang et al. 2016). With all advances, however, some challenges persist. A critical area that requires comprehensive validation is the generalizability of the models across different populations, types of scanners used and protocols for imaging. Alongside this, there is a pressing need for increased transparency and a way to explain the abilities of AI models to instill trust for clinical usage and adoption. Still, the application of deep learning techniques for medical imaging is a striking development for early cancer detection and highlights the tremendous influence of AI technology on current oncology practices (Topol 2019). Pathology has experienced astonishing changes with the development of AI tools that fully automate and enhance the analysis of histological tissue samples. Histopathology has, until recently, depended on the physical examination of tissues under a microscope. It is now augmented by powerful ML paradigms, especially deep learning methods with their CNNs that can recognize and measure intricate patterns in scanned pathology slides (Komura and Ishikawa 2018; Campanella et al. 2019). Such artificial intelligence (AI) algorithms can, with great accuracy, determine the type of tumors, grade their degree of malignancy, count mitotic figures and detect necrosis, lymphocytic infiltration and vascular invasion. One of the most important benefits AI provides in histopathology is ensuring improved accuracy and reliability for diagnostics. A human interpretation is prone to variation from factors such as fatigue, level of experience or personal bias, while AI algorithms provide a uniform standard of analysis at a high speed for different datasets, and can reliably and reproducibly process large amounts of data. For instance, several studies have demonstrated that AI is able to match the diagnostic performance of an expert pathologist when differentiating between ductal carcinoma in situ (DCIS) and breast cancer or grading prostate cancer (Steiner et al. 2018; Bulten et al. 2022). Additionally, AI is being integrated into predicting genetic mutations and molecular subtypes directly from histopathological images, which is known as computational pathology. This allows for genomic alterations such as TP53, EGFR and KRAS mutations to be hypothesized without expensive molecular testing, thus accelerating the diagnostic process. AI also aids in tumor microenvironment characterization by offering details on immune cell infiltration and spatial relations that are important for predicting tumor response to immunotherapy (Fu et al. 2020; Lu et al. 2021). In histopathology, AI technology is making a significant advancement in the diagnosis of cancer using tissue samples, altering the entire process by improving workflow efficiency, cutting down time and empowering the pathologists. With the expansion of digital pathology and the subsequent increase in available dataset diversity, the efficacy and utility of AI models will only improve, cementing their application in contemporary oncologic practice (Echle et al. 2021). AI systems have the ability to radically change the interpretation of highly complicated genomic data which remains foundational in precision oncology. Interpreting the data in a single human genome is nearly impossible at scale due to the vast amount of information; however, algorithms based on ML and deep learning are capable of analyzing such data in order to extract clinically valuable insights. These AI tools are able to scrutinize large volumes of WGS, WES and RNA-seq data to detect mutations, structural variations and gene expressions associated with cancer (Kather et al. 2019; Zou et al. 2019). AI models excel in identifying both common and rare genetic variants contributing to oncogenesis. For instance, mutation detection in oncogenes and tumor suppressor genes such as BRCA1, TP53 or EGFR and the evaluation of their functional impact are all within the capability of AI. This assists in the genetic risk profile classification of patients enabling tailored screening and early intervention approaches (Yuan and Bar-Joseph 2019). Moreover, ML algorithms can forecast the possible success outcomes of focused therapies and immunotherapies by examining the tumor mutational burden, microsatellite instability and other relevant biomarkers from the corresponding genomic data (Libbrecht and Noble 2015). AI has proved rather crucial for actionable mutations in the field of genomics and for patient–clinical trial match making. Through an AI approach, genomic data with relational drug–gene interaction databases are interlinked, providing innovative treatment strategies or suggesting participation in precision medicine trials. NLP algorithms significantly contribute to the genomics knowledge base by mining literature, clinical trial registries and updating relevant genomic information (Lee et al. 2018
9
Automating Cancer Detection and Rehabilitation: The AI Revolution
9.1. Introduction
9.2. AI in cancer detection
9.2.1. Medical imaging and deep learning
9.2.2. Histopathological analysis
9.2.3. Interpreting genomic data
Related posts:
The Role of Artificial Intelligence in Personalized Physiotherapy and Cancer Treatment
Model Validation Techniques for AI in Cancer Research Based on Physiotherapy and Oncology
Machine Learning in Oncology and Physiotherapy: A New Era of Precision Medicine
AI and Wearable Technology in Physiotherapy for Oncology Patients
Smart Robotics in Physiotherapy and Oncology: Redefining Patient Outcomes
Recurrent Neural Networks for Predictive Modeling in Cancer Time Series Data
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
