The implications of nanodiagnostics and artificial intelligence for detecting oral potentially malignant disorders and oral squamous cell carcinoma

Nanodiagnostics and artificial intelligence (AI) technology as emerging detection techniques provides advantages over early diagnosis of oral cancer compared with conventional techniques. We read with great interest two correspondence papers entitled “Conspectus on nanodiagnostics as an incipient platform for detection of oral potentially malignant disorders and oral squamous cell carcinoma” and “The use of multilayer perceptron and radial basis function: an artificial intelligence model to predict progression of oral cancer” recently published in International Journal of Surgery^1,2. Early detection of oral squamous cell carcinoma (OSCC) and the precursors known as oral potentially malignant disorders (OPMD) is very critical in order to improve the health benefits of the patient and the prognosis of this malignancy. As a large number of OPMD/OSCC patients are diagnosed in our hospital every year, we hope that the new diagnostic strategies on these diseases can emerge in our day-to-day practice.

We congratulate Shanmugam and colleagues for an interesting paper which introduces the conspectus on nanodiagnostics for detecting OPMD and OSCC. Although the authors attempted to provide a brief summary of nanodiagnostic strategies in the detection of oral cancer, this issue was actually well summarized by earlier review articles^3–5, which were published before the time of the correspondence submitted but not cited by Shanmugam et al. ¹. A well-written review article by Chakraborty et al.³ comprehensively summarized nanotechnology-based techniques used for oral cancer diagnosis, and described the details of these relevant nano-based oral cancer screening methods including Raman scattering, colorimetric-nano, fluorescent-nano, electrochemical-nano biosensors, and nanotechnology combined with optical coherence tomography (OCT), magnetic resonance imaging, diffusion reflection imaging, modified immunoassay, microfluidics (lab-on-a-chip). Shanmugam et al. ¹ briefly listed the types of nanoparticles (NPs), such as liposomes, gold NPs, and magnetic NPs, currently used in oral cancer diagnosis and nanomaterial promising structures for oral cancer management.

Although the term ‘oral potentially malignant disorders’ was entitled in the title, the existing evidence on nanodiagnostics of OPMD was not mentioned by Shanmugam et al. ¹. This allows us to focus attention on the nanodiagnostics of OPMD. Within above context, the current correspondence gives us an opportunity to provide a summary of nanodiagnostic strategies for OPMD for the consideration of the authors and readership. As presented in Table 1, there were three relevant nano-based diagnostic methods including diffusion reflection imaging nanocarrier-loaded gold nanorods, biosensors loaded nano-bio-chip, and OCT loaded Au-NPs^6–11. These nanocarriers can conjugate with Anti-epidermal growth factor receptor monoclonal antibodies to enhance the contrast and penetration depth of lesions in-vivo imaging. Tissue and brush samples from OPMD patients and chemical carcinogens-induced experimental murine oral carcinogenesis models were used to investigate the potential of OPMD nanodiagnostics. As Chakraborty et al. ³ discussed the challenges and prospects, nanodiagnostics for detecting and monitoring of OPMD/OSCC remains in its infancy. Significant research is required to improve them and expand their application from the bench-to-bedside^3,12.

Table 1.

Summary of included studies on nanotechnology-based methods used for OPMD diagnosis.

Technique area	Detection method	Nanocarrier	Conjugated with substances	Cell line/sample/model	Main results	References
Optical	Diffusion reflection imaging	Gold nanorods	Anti-EGFR monoclonal antibodies	A rat tongue model of 4NQO-induced carcinoma	Absorption of light increased significantly in cases of moderate-severe dysplasia/OSCC compared to low-risk lesions (86% sensitivity and 89% specificity, AUC=0.79)	Sudri et al.6
Optical	Diffusion reflection imaging	Gold nanorods	Anti-EGFR monoclonal antibodies	Tissue samples from 15 various dysplasia, 10 OSCC, and 5 HC	Reflectance intensity increased with the progression of the disease, lowest in the control group and increasing as the dysplastic changes increase (P<0.001 for linear trend of grade)	Hirshberg et al.7
Exfoliative cytology	Nano-bio-chip sensor	Nano-bio-chip	\	FA-OSCC cell lines and brush samples of a FA patient with OPMD	The programmable bio-nanochip test recognized dysplastic oral epithelial cells in a brush biopsy sample of a FA patient	Floriano et al.8
Exfoliative cytology	Nano-bio-chip sensor	Nano-bio-chip	Anti-EGFR monoclonal antibodies	Brush samples from 41 OPMD/OSCC patients and 11 HC	Enhance discrimination of oral cancer and precancerous lesions (AUC=0.94) with high sensitivity and specificity	Weigum et al.9
Optical	Optical coherence tomography	Plasmonic Au-NPs (15 nm)	SPDP	A hamster cheek pouch model of DMBA-induced carcinoma	The dysplastic tissue was visualized by optical coherence tomography imaging, which showed reduced scattering intensity and increased Doppler variance	Kim et al.10
Optical	Optical coherence tomography	Spherical Au-NPs (70 nm)	Anti-EGFR monoclonal antibodies	A hamster cheek pouch model of DMBA-induced carcinoma	Enhance the contrast and penetration depth in-vivo optical coherence tomography images of oral dysplasia	Kim et al.11

4NQO, 4-nitroquinoline-N-oxide; AUC, area under the curve; DMBA, 7,12-dimethlybenz(a)anthracene; EGFR, epidermal growth factor receptor; FA, Fanconi anaemia; HC, healthy controls; NPs, nanoparticles; OPMD, oral potentially malignant disorder; OSCC, oral squamous cell carcinoma; SPDP, N-succinimidyl 3-(2-pyridyldithio)-propionate.

On the other side, Jayaram et al.² introduced the use of multilayer perceptron and radial basis function as an AI model to predict progression of oral cancer. Multilayer perceptron and radial basis function are deep learning algorithms used in artificial neural networks. As mentioned in two correspondence papers^1,2, convolutional neural network is one such method of automated machine earning to diagnose oral cancer^13,14. In fact, the issue on AI-assisted diagnosis of OPMD/OSCC was also well reviewed by earlier systematic reviews and meta-analyses^15–17. AI-assisted diagnostic modalities in detecting OPMD/OSCC mainly contain photographic images, fluorescence spectroscopy, Raman spectroscopy, and OCT. Ferro et al. ¹⁵ estimated the overall AUC across all the 35 studies was 0.935 for automated classification of oral cavity lesions. Elmakaty et al. ¹⁶ pooled the sensitivity and specificity with 95% CI being 92.0% (86.7–95.4%) and 91.9% (86.5–95.3%), respectively, across 16 studies on AI-assisted technologies in detecting OSCC. Kim et al. ¹⁷ reported that OCT was more diagnostically accurate and more negatively predictive than photographic images and autofluorescence on the screening for all OPMD from normal mucosa.

Herein, we put forward a hypothesis whether ChatGPT can assist in the screening and diagnosis of oral cancer. ChatGPT is a recently released AI-based large language model chatbot that has gathered significant interest in the medical community. Of note, the appropriateness and accuracy of the ChatGPT responses to common questions about breast and liver cancer prevention and screening/diagnosis were newly assessed^18,19. Since ChatGPT provided appropriate responses for 22/25 (88%) questions posed about breast cancer prevention and screening, it has great potential to automate provision of accurate health care information related to breast cancer¹⁸. Conversely, ChatGPT did not reliably provide accurate information about liver cancer surveillance and diagnosis, owing to 15/20 (75%) questions being considered to have not reliable answers¹⁹. Although whether ChatGPT can assist in the screening and diagnosis of oral cancer is unknown, the potential application of ChatGPT in this field as follows.

The combination of medical imaging and AI tools such as ChatGPT is considered the most promising field in the diagnosis and intelligent classification of OPMD/OSCC. Further, ChatGPT could be used to comprehensively analyze clinical and histological data, imaging data, and even genetic data of the patients, and further optimize the lesion recognition and classification methods. The second is case management and data analysis. After ChatGPT access to case management systems, it could help clinicians record and extraction information such as patient history, examination results, diagnosis and treatment procedures²⁰. Additionally, ChatGPT could also play an important role in risk prediction model of OPMD/OSCC. Collectively, the development of AI has become an inexorable trend, with the potential to improve clinical workflow and responsible use of diagnosis, treatment and care services.

Ethical approval

This work does not include any human/animal subjects to acquire such approval, and then there is not the relevant Judgement’s reference number.

Source of funding

The authors are supported by Science and Technology Commission of Shanghai Municipality (21015800800, 20Y11903700), Two hundred talent project of Shanghai Jiao Tong University School of Medicine, Innovative Research Team of High-level Local Universities in Shanghai (SHSMU-ZLCX20212401, SSMU-ZLCX20212300), Crossdisciplinary Research Fund (JYJC202113) of Shanghai Ninth People’s Hospital, and Research Discipline Fund no. KQYJXK2020 of the Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine, College of Stomatology, Shanghai Jiao Tong University.

Author contribution

W.L., Y.W.: investigation, data collection, writing—original draft preparation. X.S., X.Y., L.S.: conceptualization, writing—review and editing, and supervision.

Conflicts of interest disclosure

There are no conflicts of interest.

Research registration unique identifying number (UIN)

The World Medical Association's Declaration of Helsinki 2013 states in article 35: 'Every research study involving human subjects must be registered in a publicly accessible database before recruitment of the first subject'. Editors of IJS require that all types of research studies involving human participants should be registered. There are many places to register your research and you can choose which is the most suitable for your needs:

* Clinicaltrials.gov - for all human studies - free

* Chinese Clinical Trial Registry chictr.org.cn - for all human studies - free

* Researchregistry.com - for all human studies - charge

* ISRCTN.com - for all human studies - charge

* Prospero - for systematic reviews - free (Review status must be Review Completed not Published)

* There are many national registries approved by the UN that can be found at http://www.who.int/ictrp/network/primary/en/

Elsevier does not support or endorse any registry.

Even retrospective studies should be registered prior to submission. Once registered, please add the following to your Author Disclosure Form:

1. Registry used.

2. Unique Identifying number or registration ID.

3. Hyperlink to your specific registration (must be publicly accessible and will be checked)

Not applicable.

Guarantor

Xuemin Shen.

Data statement

The data in this correspondence article are not sensitive in nature and is accessible in the public domain. The data are therefore available and not of a confidential nature.