Development of an Intelligent Framework for Early Diagnosis of Cardiovascular Diseases and Diabetes Mellitus Using Handheld Spectroscopy Techniques and Fingernails

Wilson, M (2026) Development of an Intelligent Framework for Early Diagnosis of Cardiovascular Diseases and Diabetes Mellitus Using Handheld Spectroscopy Techniques and Fingernails. Doctoral thesis, Liverpool John Moores University.

Preview	Text 2026Wilsonphd.pdf - Published Version Available under License Creative Commons Attribution Non-commercial. Download (8MB) | Preview
	Text 2026wilsoninternalphd.pdf - Published Version Access Restricted Available under License Creative Commons Attribution Non-commercial. Download (8MB)

Abstract

Aim: Cardiovascular diseases and diabetes mellitus represent leading contributors to morbidity and mortality, imposing substantial strain on healthcare systems across the globe. These burdens are particularly detrimental in low- and middle-income countries, where access to advanced diagnostic technologies and training is limited. Timely detection of CVDs and DM is imperative for effective clinical intervention and disease prevention. However, conventional diagnostic techniques are often invasive, resource intensive and dependent on user knowledge. Therefore, this research aimed to develop and evaluate an intelligent, non-invasive framework using fingernails as an alternative biological matrix, combined with vibrational spectroscopic techniques and machine learning for the detection of cardiovascular diseases and/or diabetes mellitus.
Methods: Two quantitative studies, including a systematic review and structured questionnaire were conducted. The first examined current trends in diagnostic practices, such as diagnostic technique and biomarkers. The second created an understanding of disease-related factors and contextualised spectral data. The detection of disease was assessed through the physical and biochemical analysis of fingernails and was conducted using a variety of vibrational spectroscopic techniques including infrared, near-infrared and Raman spectroscopy. Moreover, the impact of disease in terms of elemental composition and tissue damage to the fingernail plate were evaluated through scanning transmission electron microscopy and energy dispersive X-ray spectrometry, respectively.
Data analysis: Quantitative data obtained from the systemic review of literature and participants’ data from questionnaires were analysed using Microsoft Excel version 2019 and SPSS version 29.0.2.0, where descriptive statistics and logistic regression were applied. Machine learning algorithms including correlation in wavenumber space, principal component analysis and self-organising maps, as well as classification models, were applied to explore patterns in the spectroscopic data using MATLAB version R2024a.
Findings: A systematic review of the existing literature revealed frequent use of advanced analytical techniques and biomarkers for the detection of cardiovascular diseases or diabetes mellitus. Evidence from the systematic review, supported the notion of exploring alternative techniques and biological matrices for rapid and cost-effective analysis in low- and middle-income countries. The results of the questionnaires, paired with logistic regression, demonstrated the relationship between the manifestation of cardiovascular diseases and diabetes mellitus with patient- (age, biological sex and smoking habits) and clinical-related (comorbidities) risk factors. The chosen vibrational spectroscopic techniques highlighted the shared biochemical fingerprint of fingernails in the absence or presence of disease. However, the spectral activity of endogenous constituents (amino acids, lipids and proteins) varied dependent on factors such as age, biological sex, ethnicity and disease presence. Moreover, the spectroscopic techniques allowed for identification of biomarkers including cytochrome C and glucose. Disease manifestation was also visualised in the form of a rough topographic fingernail plate and indicated the presence of tissue damage attributed to poor circulation and/or hyperglycaemia.
Conclusions: This work established a proof-of-concept intelligent framework capable of identifying disease-related biochemical and structural patterns. By utilising a non-invasive sampling method, which requires minimal sample preparation and storage, this approach has the potential to significantly reduce per-test diagnostic costs and decrease reliance on conventional laboratory-based blood analysis. The integration of machine learning further enables automated data processing and classification, thus reducing clinician workload and shortening turnaround times. Factors that are beneficial for large scale and over-subscribed institutes such as the National Health Service. Moreover, traditional diagnostic pathways are often dependent on costly laboratory infrastructure and highly trained personnel, both of which are limited or inaccessible in low-resource settings. In contrast, vibrational spectroscopic techniques offer rapid and reagent-free analysis, which can be implemented using handheld devices. However, due to an imbalance of data, with limited disease representation, misclassification of machine learning and classification models were observed. Therefore, future work will focus on expanding and diversifying the datasets with larger-scale studies to encompass a greater number of cases, as well as participants with diverse backgrounds and characteristics. In addition, further research will include the application of diffuse reflectance infrared Fourier transform spectroscopy for deeper penetration of fingernails and application of additional Raman lasers for optimised instrumental parameters and standard operating procedures.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Spectroscopy; Machine Learning; Cardiovascular Diseases; Diabetes Mellitus; Fingernails
Subjects:	Q Science > QP Physiology R Medicine > RS Pharmacy and materia medica
Divisions:	Pharmacy and Biomolecular Sciences
Date of acceptance:	23 April 2026
Date of first compliant Open Access:	18 May 2026
Date Deposited:	18 May 2026 08:50
Last Modified:	18 May 2026 08:50
DOI or ID number:	10.24377/LJMU.t.00028450
Supervisors:	Assi, S, Birkett, J, Al-Jumeily OBE, D and Khan, I
URI:	https://researchonline.ljmu.ac.uk/id/eprint/28450

View Item