ISMPO Insights

Artificial intelligence and head and neck cancers

Rakesh Katna1, Mahadeepa Kar2

1Consultant Head and Neck Onco-Surgery

Jaslok Hospital, Mumbai

Email ID: katna.rakesh@gmail.com

2Fellow Head and Neck Onco-Surgery, MDS

Vedant Hospital,Thane

INTRODUCTION

Artificial intelligence (AI) refers to the capability of machines to emulate human intelligence without explicit programming.[1] AI can complete tasks that require complicated decision-making. Recent developmentsincomputing power and big data handling havepromoted the use of AI to supporter substitute conventional approaches.[2]AI-based medical platforms have changed the way we approach healthcare. The use of AI in diagnosis, treatment planning, patient monitoring and personalized care is getting well established now.[1]

There are two main types of AI– Machine Learning and Deep Learning.Machine Learning (ML) is not programmed to accomplish a particular task but can rather learn interactively to make predictions or take decisions. The more data a ML model is fed, the better it performs over time.Deep Learning (DL) is a subcategory ofML that uses artificial neural networks modelled after how the human brain processes information to learn from vast amounts of data. A well-made andtrained DL model is capableofexecuting classification tasks and making accuratepredictions, sometimes surpassing human expert–level performance.[3]

ROLE OF “AI” IN HEAD AND NECK CANCER CARE.

Use of AI in identification of head and neck precancerous and cancerous lesions

Early diagnosis of potentially malignant head and neck lesions can inhibit cancer development in up to 88% - 94 % of cases, still most patients are diagnosed at a late stage of disease specially in the Indian subcontinent (62% detected at stage III or IV).[4] In the last decade, AI has grown in popularity in cancer research where it has demonstrated improved diagnostic accuracy and efficiency in predicting cancer behaviour and prognosis.[4]

“Segmentation” comprisesisolating high-resolution digital whole slide images (WSI) of human tissue into regions of clinical significance followed by deconstruction of the WSI into smaller images by a methodcalled ‘patch extraction’.[5]

Predictions from Imaging

AI-based modelshave become a budding area of interest in prognostic medicine and have the potential tohelp physician decision-making to improve patient outcomes. Radiomics and imaging play a critical role in thesemodels. Exploratory studies are encouraging;however, validation studies that exhibit consistency, reproducibility, and predictiveimpact remain rare. Potentialclinical trials with standardized procedures are necessary for clinical translation. [6]

Radiomic featuresmayshine newlight onthe fundamental pathophysiology of the tumors and may be used asimaging ‘‘biomarkers’’ to aid in the forecast of outcomes.[7]

Prediction of Oncologic Outcomes

Customized risk-based treatment selectionwill strongly dependon models that can deliverprecise predictions of oncologicoutcomes. In head and neck squamous cell carcinoma (HNSCC) treatment, compromises are madebetween disease control and adequate treatment toxicity. Thecapability to forecast treatment outcomes permits patient-specificselections around preferred treatment intensity.[6]

Patients with HPV-associated oropharyngeal cancer haveexceptional goodprognosis.[8]Nevertheless, roughly 10-15% ofpatients continue to perform badly despite treatment and might be pooroptionsfor reducedtreatment.Prediction models forsurvival may be able to recognize these patients before starting treatment and help in improving outcomes.[9]

Pathological findings based on Imaging

Pathological findings, such as lymph node number, location, and size of tumor, depth of invasion, and existence ofextranodal extension (ENE), lymphovascular invasion, or perineuralinvasion, have predictive implications and are frequentlyappliedto assistclinical decision-making. However, imagingoutcomes can be subtle,generic and difficultfor theradiologist.[10]

Predictive models for pathological discoveries are essential as the identification of ENE is a signal for adjuvant treatment intensification. Models were built which successfully determined nodal metastasis and ENE, which are crucial decision factors for cancer management in head and neck. [11]Potential work may include a look into specific prognostic factors more subjectively like Perineural invasion (PNI), lymph node ratio patterns of invasion etc. In addition to estimating patient oncologic outcomes, models able to deliver decision support around minimizing toxicity will play an essential role in clinical decision making for these patients.[12]

AI in Radiation Oncology

Patient treatment workflow in radiotherapy comprisesseveral steps like patient positioning and immobilization, acquisition of planning CT, segmentation of tumor and organs at risk (OARs), radiation planning and calculating the preferred dose, time and fractionation schedule. It also involvesoptimizing the beam positions for ideal dose coverage and normal tissue sparing followed by actual delivery of radiation, and lastly, post-treatment follow-up.[13]AI systems are mainly suited to aid and advance the efficiency of this workflow. Machine learning has been suggested for automatic organ segmentation, error avoidance, or treatment planning.[14]

CONCLUSION

Integration of AI technology in cancer care canincrease the accuracy and pace of diagnosis, assistwith clinical decision-making, and lead to improved health outcomes. AI-guided clinical care has the potential to play an essential role in decreasing health disparities, predominantly in low-resource settings & developing economics. The potential of AI in cancer care needs to be tapped with more focused research in this domain.

References
  • 1. Pomerol JC. Artificial intelligence and human decision making. European Journal of Operational Research. 1997 May 16;99(1):3-25.
  • 2. Tama BA, Do Hyun Kim GK, Kim SW, Lee S. Recent Advances in the Application of Artificial Intelligence in Otorhinolaryngology-Head and Neck Surgery. Clinical and Experimental Otorhinolaryngology. 2020 Nov;13(4):326.
  • 3. El Naqa I, Murphy MJ. What is machine learning? Inmachine learning in radiation oncology 2015 (pp. 3-11). Springer, Cham.
  • 4. Mahmood H, Shaban M, Indave BI, Santos-Silva AR, Rajpoot N, Khurram SA. Use of artificial intelligence in diagnosis of head and neck precancerous and cancerous lesions: a systematic review. Oral Oncology. 2020 Nov 1;110:104885.
  • 5. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521(7553):436–44.
  • 6. Chinnery T, Arifin A, Tay KY, Leung A, Nichols AC, Palma DA et al. Utilizing Artificial Intelligence for Head and Neck Cancer Outcomes Prediction From Imaging. Canadian Association of Radiologists Journal. 2021 Feb;72(1):73-85.
  • 7. Gillies RJ, Kinahan PE, Hricak H. Radiomics: images are more than pictures, they are data. Radiology. 2016 Feb;278(2):563-77.
  • 8. O’Sullivan B, Huang SH, Su J, et al. Development and validation of a staging system for HPV-related oropharyngeal cancer by the International Collaboration on Oropharyngeal cancer Network for Staging (ICON-S): a multicentre cohort study. Lancet Oncol. 2016;17(4):440-451.
  • 9. William T et al. Predictive Models for Newly Diagnosed Prostate Cancer Patients.Reviews in Urology 11(3):117-26
  • 10. Hoang JK, Vanka J, Ludwig BJ, Glastonbury CM. Evaluation of cervical lymph nodes in head and neck cancer with CT and MRI: tips, traps, and a systematic approach. Am J Roentgenol. 2013; 200(1):W17-W25.
  • 11. Kann BH, Hicks DF, Payabvash S, et al. Multi-institutional validation of deep learning for pre-treatment identification of extra nodal extension in head and neck squamous cell carcinoma. J ClinOncol. 2020;38(12):1304-1311.
  • 12. Ariji Y, Fukuda M, Kise Y, et al. Contrast-enhanced computed tomography image assessment of cervical lymph node metastasis in patients with oral cancer by using a deep learning system of artificial intelligence. Oral Surg Oral Med Oral Pathol Oral Radiol. 2019;127(5):458-463.
  • 13. Rattan R, Kataria T, Banerjee S, Goyal S, Gupta D, Pandita A, et al. Artificial intelligence in oncology, its scope and future prospects with specific reference to radiation oncology 2019; 1: 20180031.
  • 14. Feng M, Valdes G, Dixit N, Solberg TD. Machine learning in radiation oncology: opportunities, requirements, and needs. Frontiers in oncology. 2018 Apr 17;8:110.

Courtesy - Indian Journal of Medical and Paediatric Oncology (IJMPO)
Editor-in-Chief - Dr. Padmaj Kulkarni
Section Editor - Dr. Sneha Bothra
Editorial Assistant - Devika Joshi