top of page

Technological advancements in head and neck cancer detection – looking into the future

Hanya mahmood- acf in oral surgery, sheffield, uk.


The global incidence of head and neck cancers (HNC) is on the rise, despite the introduction of public health awareness campaigns and advancements in medical and surgical techniques. HNC is now the sixth leading group of cancers worldwide accounting for more than 650,000 new cases and 33,000 deaths each year. In the UK alone, figures have increased by 22% over the last decade, with almost 12,000 new diagnoses each year, or 33 every day. An early diagnosis can prevent up to 88% of cases, however most patients are diagnosed at a late stage of disease (62% diagnosed at stage III or IV) resulting in poor survival. Histopathological interpretation remains the diagnostic gold standard and the only determinant of prognosis, playing a key role in informing clinical treatment decisions. However, there are recognised limitations of a manual diagnosis using light microscopy, including variationS in observation and differences in interpretation even amongst the most experienced pathologists. This subjectivity CAN POTENTIALLY INFLUENCE cancer diagnosis AND have a negative impact on patient survival, emphasising the need for improved methods and technologies to increase diagnostic CONSISTENCY AND EFFICIENCY.

Over the last decade, Artificial Intelligence (AI) has gained popularity with application to cancer research to increase diagnostic accuracy and efficiency. AI is a branch of computer science concerned with building smart machines that can perform tasks which typically require human intelligence. AI applications are increasingly being used in everyday life (e.g. Siri, Alexa, Google maps) and its popularity in cancer research is rapidly evolving with improvements in image analysis algorithms. Machine learning (ML) is the most common type of AI method which uses computational methods to “learn” information directly from data. ML algorithms adaptively improve their performance as the number of samples available for learning increases, enabling the computer to essentially “learn from experience”. ML has been shown to increase diagnostic accuracy and efficiency by removing subjective variability of cancer classification, in addition to providing quantifiable outputs for cancer prediction and prognosis, which can play a key role in guiding treatment decisions and standardising patient care.

The increase in computational power coupled with the recent advent and cost–effectiveness of digital slide scanners enables tissue histopathology slides to be digitised, providing the ability for whole slide image (WSI) analysis. Research has demonstrated the success of ML methods to identify prognostic markers and to predict disease outcome and survival in a range of cancers including colorectal, lung, skin and breast malignancies. Furthermore, there is evidence to support the use of AI-based methods for prediction of HNC behaviour, prognosis, survival and treatment success from radiological, cytological, clinical and genomic data. However, there remains a paucity of research exploring the role of AI/ML for detection, grading and classification of HNC from histology slides.

Early research has suggested the potential for classical ML methods to detect specific histological features in oral precancerous lesions. In one study, ML was used to produce an automated nuclear phenotypic score to allow binary classification of oral epithelial dysplasia lesions into low and high-risk groups (78% sensitivity and 71% specificity). In another study, ML methods were used to quantify sub-epithelial connective tissue cells based on eccentricity and compactness, to enable differentiation of normal tissue from oral submucous fibrosis tissue with/without dysplasia or atrophy (classification accuracy of 88.69%). Another study used a modern ML approach involving a 12-layer deep convolutional neural network to segment constituent layers of the oral mucosa (into epithelial, subepithelial and keratin layers) and extract keratin pearls from segmented keratin regions in oral cancer slides. The detection accuracy was reported as 96.88%, although results were based on a small dataset consisting of small visual fields of tissue sections. Currently, there is still extremely limited evidence studying the performance of AI/ML methods for detection of other types of HNC, highlighting scope for future research in this field, particularly given the increasing incidence.

We are living in an exciting time with emerging evidence and increasing potential for computerised imaging methods to assist pathologists in making an accurate diagnosis of disease and identifying morphological features correlated with prognosis. However, more high-quality research using state-of-the-art AI methods is needed to develop novel digital biomarkers which will be crucial to help understand the underlying mechanism of HNC development and progression in addition to determining which histological features are most important in early detection. Ultimately, this will help to develop targeted, patient-specific therapeutics to reduce HNC associated morbidity.


relevant reading

- Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA: a cancer journal for clinicians. 2011 Mar;61(2):69-90.

- Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin 2020; 70:7.

- Bray F, Ferlay J, Soerjomataram I, et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2018; 68:394.

- Cancer Research UK (2011) Cancer Incidence in the UK in 2011.

- Bejnordi, B. E. et al. Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer. Jama 318, 2199–2210 (2017).

- Kourou K, Exarchos TP, Exarchos KP, Karamouzis MV, Fotiadis DI. Machine learning applications in cancer prognosis and prediction. Computational and structural biotechnology journal. 2015 Dec 31;13:8-17

- Kourou K, Exarchos TP, Exarchos KP, Karamouzis MV, Fotiadis DI. Machine learning applications in cancer prognosis and prediction. Computational and structural biotechnology journal. 2015 Dec 31;13:8-17

- LeCun Y, Bengio Y, Hinton G. Deep learning. nature. 2015 May;521(7553):436-44.

- Bera K, Schalper KA, Rimm DL, Velcheti V, Madabhushi A. Artificial intelligence in digital pathology—new tools for diagnosis and precision oncology. Nature Reviews Clinical Oncology. 2019 Nov;16(11):703-15.

- Zormpas-Petridis K, Failmezger H, Raza SE, Roxanis I, Jamin Y, Yuan Y. Superpixel-based Conditional Random Fields (SuperCRF): Incorporating global and local context for enhanced deep learning in melanoma histopathology. Frontiers in oncology. 2019;9:1045.

- Wang S, Yang DM, Rong R, Zhan X, Fujimoto J, Liu H, Minna J, Wistuba II, Xie Y, Xiao G. Artificial Intelligence in Lung Cancer Pathology Image Analysis. Cancers. 2019 Nov;11(11):1673.

- Veta M, Van Diest PJ, Willems SM, Wang H, Madabhushi A, Cruz-Roa A, Gonzalez F, Larsen AB, Vestergaard JS, Dahl AB, Cireşan DC. Assessment of algorithms for mitosis detection in breast cancer histopathology images. Medical image analysis. 2015 Feb 1;20(1):237-48.

- Song B, Sunny S, Uthoff RD, Patrick S, Suresh A, Kolur T, Keerthi G, Anbarani A, Wilder-Smith P, Kuriakose MA, Birur P. Automatic classification of dual-modalilty, smartphone-based oral dysplasia and malignancy images using deep learning. Biomedical optics express. 2018 Nov 1;9(11):5318-29.

- Baik J, Ye Q, Zhang L, Poh C, Rosin M, MacAulay C, Guillaud M. Automated classification of oral premalignant lesions using image cytometry and random forests-based algorithms. Cellular Oncology. 2014 Jun 1;37(3):193-202.

- Krishnan MM, Pal M, Bomminayuni SK, Chakraborty C, Paul RR, Chatterjee J, Ray AK. Automated classification of cells in sub-epithelial connective tissue of oral sub-mucous fibrosis—An SVM based approach. Computers in biology and medicine. 2009 Dec 1;39(12):1096-104.

- Das DK, Bose S, Maiti AK, Mitra B, Mukherjee G, Dutta PK. Automatic identification of clinically relevant regions from oral tissue histological images for oral squamous cell carcinoma diagnosis. Tissue and Cell. 2018 Aug 1;53:111-9.


bottom of page