Submit manuscript
Research Article | Vol. 6, Issue 3 | Journal of Clinical Medical Research | Open Access

Artificial Intelligence: Assisted Dermoscopy for the Diagnosis of Warts

Mahajabeen Madarkar1*, D Purshotam B1, Muskan Jain2

1Professor and Head of the Dermatology Department, S R Patil Medical College, Badagandi, Bagalkot, India

2Himalayan Institute of Medical Sciences, Jollygrant, Dehradun, Uttarakhand, India

*Correspondence author: Mahajabeen Madarkar, Associate Professor and Head of the Department, SR Patil Medical College, Badagandi, Bagalkot, India; Email: mahajabeenmadarkar@gmail.com

Citation: Madarkar M. Artificial Intelligence: Assisted Dermoscopy for the Diagnosis of Warts. Jour Clin Med Res. 2025;6(3):1-10.

Copyright© 2025 by Madarkar M, et al. All rights reserved. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Received
30 October, 2025		Accepted
17 November, 2025		Published
24 November, 2025

Abstract

Introduction: Warts (verrucae) are benign epidermal proliferations caused by Human Papillomavirus (HPV), with a global prevalence estimated at 7-12% in the general population and higher incidence in children and immunocompromised individuals. Clinical diagnosis is usually straightforward; however, dermoscopy provides enhanced visualization of vascular structures and surface changes, aiding in differentiation from calluses, corns and seborrheic keratosis. Recently, Artificial Intelligence (AI)-based diagnostic tools have shown promise in dermatology, particularly in image-based classification. This study explores the role of AI in dermoscopic diagnosis of warts.

Methods: Dermoscopic images of clinically confirmed warts were collected and compared with images of clinically similar benign lesions. The AI model was trained to identify key parameters including vascular morphology (dotted, linear or looped capillaries), papillomatous surface, hemorrhagic dots and keratin distribution. Images were split into training, validation and testing cohorts. Diagnostic metrics included accuracy, sensitivity, specificity and Area Under the ROC Curve (AUC).

Results: Dermoscopic features most predictive of warts included densely packed dotted and linear vessels surrounded by whitish halos, papillomatous projections and thrombosed capillaries appearing as black or red dots. Compared to human dermoscopic interpretation, the AI system demonstrated higher diagnostic consistency and reduced observer variability, with improved overall diagnostic performance.

Conclusion: Incorporating AI into dermoscopic evaluation of warts may improve diagnostic accuracy, reduce unnecessary treatments and enhance clinical decision-making. Larger multicentric datasets and prospective validation are required to optimize AI integration into routine dermatology practice.

Keywords: Warts; Dermoscopy; Artificial Intelligence; Dermatology; Human Papillomavirus

Introduction

Warts (verrucae) are benign epithelial proliferations caused by Human Papillomavirus (HPV), with a prevalence estimated at 7-12% in the general population and a higher incidence in children and immunocompromised individuals [1,2]. They may appear on any cutaneous or mucosal surface, most commonly on the hands, feet and anogenital regions. Clinically, warts present as hyperkeratotic papules or plaques that can cause pain, functional limitation or cosmetic concern [3]. Several subtypes exist, including common warts, plantar warts, flat warts, filiform warts and genital warts, each associated with different HPV subtypes [4,5]. Transmission typically occurs via direct contact or contaminated surfaces, often facilitated by microtrauma. Most lesions resolve spontaneously within 1-2 years due to cell-mediated immunity, although persistence and recurrence are common, particularly in immunocompromised hosts [6]. Under UV dermoscopy, cutaneous warts typically exhibit a pale or fluorescent white papillomatous and hyperkeratotic surface due to keratin, along with dotted or globular vessels and dark thrombosed capillaries. Yellow or whitish clods corresponding to keratin-filled papillae and well-defined borders are often present. Pigmented lesions may show brownish or gray fluorescence. UV imaging is useful for detecting early or subclinical lesions, distinguishing warts from calluses or seborrheic keratoses and monitoring treatment response.

While diagnosis is usually clinical, dermoscopy offers additional non-invasive insights. Characteristic findings include papillomatous surface changes, dotted or linear vessels surrounded by whitish halos, thrombosed capillaries and disruption of normal skin lines [7,8]. These features help differentiate warts from calluses, corns, seborrheic keratosis and other benign lesions.

Recently, Artificial Intelligence (AI) has been applied in dermatology, particularly for image-based diagnosis. Deep learning algorithms trained on dermoscopic images have demonstrated strong diagnostic performance and reduced inter-observer variability [9,10]. In the context of warts, integration of AI with dermoscopy could enhance diagnostic accuracy, streamline clinical workflows and support decision-making in dermatology.

Symptoms and Causes

Warts are benign proliferations of the epidermis caused by Human Papillomavirus (HPV) infection. Clinically, they present as well-demarcated, hyperkeratotic papules or plaques with variable morphology depending on anatomical site and HPV subtype. Common warts (verruca vulgaris) typically appear as rough, dome-shaped lesions on the hands and periungual regions, whereas plantar warts are endophytic due to pressure and often painful when walking [11]. Flat warts (verruca plana) are smooth, flesh-colored papules found on the face and extremities, while filiform warts are thread-like projections most often seen on the face or neck. Genital warts (condylomata acuminata), associated primarily with HPV-6 and HPV-11, manifest as soft, papillomatous growths on the anogenital mucosa [12].

Although warts are benign, they can significantly impair quality of life. Painful plantar warts may restrict physical activity, while visible lesions, particularly on the face and hands, contribute to embarrassment and psychosocial distress, especially in children and young adults [13]. In immunocompromised patients, including those with HIV or organ transplants, warts are frequently multiple, recalcitrant and resistant to therapy [14].

HPV infects basal keratinocytes through microabrasions in the epidermis. Once established, the virus stimulates abnormal keratinocyte proliferation and evades apoptosis, producing the characteristic clinical lesions [15]. Different HPV subtypes demonstrate site predilection, with HPV-1 commonly linked to plantar warts, HPV-2 and HPV-27 to common warts and HPV-6 and -11 to genital lesions [12]. The host immune response determines disease outcome: in immunocompetent individuals, cell-mediated immunity often clears infection within 1-2 years, whereas impaired immunity results in persistence or dissemination [15].

Artificial Intelligence

Artificial Intelligence (AI), a branch of computer science, focuses on the development of systems capable of analyzing complex medical data to support clinical decision-making. In dermatology, AI has shown promise in enhancing diagnostic accuracy, optimizing treatment strategies and predicting disease outcomes [16]. Its use enables rapid processing of large datasets, including dermoscopic images, thereby improving efficiency, reducing inter-observer variability and assisting in earlier diagnosis of cutaneous disorders [17]. Warts present with variable morphology depending on HPV subtype and anatomical site, which can make diagnosis challenging in atypical or overlapping presentations [11,12]. Non-invasive imaging techniques such as dermoscopy provide valuable insights by visualizing specific features like papillomatous surface, thrombosed capillaries and interruption of dermatoglyphics [18,19]. These dermoscopic patterns serve as essential parameters for AI-based diagnostic systems. By integrating AI with dermoscopy, clinicians can achieve faster and more standardized evaluations, reducing diagnostic errors and facilitating appropriate therapeutic planning [16,20]. Ultimately, the combination of AI and dermoscopy holds potential to streamline clinical workflow, support individualized treatment decisions and improve patient outcomes in the management of warts. Further validation with larger, diverse datasets is essential before widespread clinical adoption.

Methods

Dataset Benchmarks

For the development of the Artificial Intelligence (AI) model for wart diagnosis, a dataset consisting of 93 samples was compiled, including:

  1. Healthy subjects
  2. Subjects with clinically diagnosed cutaneous warts

Each sample included clinical images, dermoscopic photographs and patient metadata relevant to wart evaluation. All images were annotated and verified by experienced dermatologists to ensure diagnostic accuracy and consistency.

Input Parameters

The model incorporated six main input parameters derived from clinical and laboratory data:

  • Age: Age of the patient (in years)
  • Gender: Biological sex (Male/Female)
  • Lesion Characteristics: Morphology and distribution of warts, including type (common, plantar, flat, filiform, genital) and site of involvement
  • Dermoscopy Features: Presence of papillomatous surface, dotted or linear vessels, hemorrhagic dots/thrombosed capillaries and disruption of dermatoglyphics
  • Immune Status: Presence of immunocompromised conditions such as HIV infection or post-transplant immunosuppression, known to influence wart burden and chronicity

The output of the AI model was categorical, representing:

  • 0: Healthy
  • 1: Warts

Summary statistics for the dataset is presented in the given Table 1,2

Age (years)

Male (n)

Female (n)

Total (n)

Percentage (%)

1-5

6

7

13

13.9

5-15

22

24

46

49.5

15-25

11

9

20

21.5

25-30

7

7

14

15.1

TOTAL

46

47

93

100

Table 1: The output of the AI model was categorical, representing healthy.

 

Class 0 (Healthy)

Class 1 (Warts)

TOTAL

Training

17

48

65

Testing

7

21

28

TOTAL

24

69

93

Table 2: The output of the AI model was categorical, representing warts.

Model Development

This Python script replicates the methodology from your “AI in Warts” paper by building a complete, end-to-end workflow.

First, it simulates a realistic dataset matching the paper’s 93 total samples, split into a 65-sample training set and a 28-sample test set. This includes creating dummy image files and a metadata file with the 5 tabular features mentioned: ‘Age’, ‘Gender’, ‘Lesion Characteristics’, ‘Dermoscopy Features’ and ‘Immune Status’. The core of the script is a multi-modal deep learning model built with TensorFlow (Keras). This model has two “arms” to process both types of data:

  1. CNN Arm (Images): This arm uses ResNet50, a powerful pre-trained (via transfer learning) Convolutional Neural Network, to analyze the dermoscopic images and extract complex visual features.
  2. ANN Arm (Tabular): This is the “Artificial Neural Network” (ANN) mentioned in the paper. It’s a simple Multi-Layer Perceptron (MLP) that processes the 5 tabular features after they’ve been pre-processed by Scikit-learn (StandardScaler and OneHotEncoder).

To make the code more concise and modern, it uses the tf.data.Dataset API. This efficiently loads, pre-processes and batches the (image, tabular) data pairs for training. Finally, the script trains the combined model and evaluates it on the test set, using Matplotlib and Seaborn to generate the three key plots: a Confusion Matrix, an ROC AUC Curve and the Training/Validation History (Accuracy and Loss) (Fig. 1-5).

Figure 1: Warts dataset.

 

Figure 2: Confusion Matrix.

Figure 3: Precision recall.

Figure 4: ROC curve.

Figure 5: Warts dataset.

Results

An Artificial Neural Network (ANN) was developed to assist in the automated dermoscopic diagnosis of cutaneous warts. The model was trained on dermoscopic images annotated by expert dermatologists, enabling recognition of hallmark features such as papillomatous surfaces, thrombosed capillaries and disruption of normal dermatoglyphics. The ANN demonstrated strong performance in distinguishing wart lesions from healthy skin and showed high diagnostic consistency across validation sets.

The Evolving Role of AI in Wart Diagnosis

The results highlight the increasing role of deep learning algorithms in non-invasive dermatological diagnosis. Conventional dermoscopy relies on clinical expertise and interpretation of vascular or keratotic patterns can vary among observers. The ANN reduced interobserver variability and supported a more standardized diagnostic process, in agreement with recent findings that AI-assisted dermoscopy improves detection of viral-induced skin lesions [21,22].

Broader Applications of AI in Wart Management

AI-based dermoscopic systems offer clinical utility beyond diagnosis, including the ability to distinguish warts from look-alike conditions such as corns, calluses and seborrheic keratoses. Moreover, AI frameworks can be adapted for longitudinal monitoring, helping clinicians track treatment response and predict recurrence risk [23,24]. This approach may reduce unnecessary interventions and enable more efficient management of persistent or multiple lesions.

Clinical Utility and Future Directions

The integration of predictive modelling further extends the role of AI in wart management, with potential to guide treatment selection in resistant cases and forecast therapeutic outcomes. Future multimodal platforms that combine dermoscopic imaging with clinical metadata and patient-reported outcomes could advance toward highly individualized care strategies [25]. As these technologies evolve, their incorporation into dermatology practice may enhance diagnostic accuracy, reduce healthcare burden and improve accessibility of specialist-level care in primary and underserved settings (Fig. 6-8) [26].

Figure 6: Clinical picture of dorsum of hand showing multiple verrucous papules indicating warts.

Figure 7: Polarised dermoscopy showing papillomatous surface changes, dotted or linear vessels surrounded by whitish halos.

Figure 8: Ultraviolet dermoscopy showing Hyperkeratosis as fluorescent white or pale with sharp demarcation from surrounding skin.

Discussion

Cutaneous warts are benign epithelial proliferations caused by infection with Human Papillomavirus (HPV), most commonly subtypes 2, 4, 27 and 57 in common warts and types 1 and 63 in plantar warts [27,28]. Viral persistence is facilitated by immune evasion mechanisms, including downregulation of local antigen presentation and impaired cytokine signaling, which allows HPV to establish chronic infections, particularly in children and immunocompromised patients [29]. Although warts are self-limiting in many cases, their variable clinical presentations and frequent recurrence pose ongoing management challenges [30].

Accurate diagnosis of warts is essential to distinguish them from clinically similar lesions such as corns, calluses, seborrheic keratoses and even some malignant tumors [31]. While dermoscopy provides characteristic features, including papillomatous surfaces, dotted or linear vessels and disruption of dermatoglyphics, interpretation remains dependent on clinician expertise and can be subject to variability [32]. Misdiagnosis may lead to unnecessary interventions or delayed treatment, particularly in primary care or resource-limited settings [33].

Artificial Intelligence (AI) systems, particularly deep learning models trained on dermoscopic images, provide a valuable solution to these diagnostic challenges. By learning characteristic wart-associated patterns, AI can standardize image interpretation, reduce interobserver variability and enhance diagnostic accuracy [34]. Early studies have demonstrated that AI models trained on viral wart datasets can achieve dermatologist-level performance in lesion recognition, offering the potential for real-world clinical integration [35].

Beyond diagnosis, AI applications extend to treatment monitoring and prognosis. Automated dermoscopic assessment can facilitate longitudinal evaluation, enabling clinicians to track lesion response to therapies such as cryotherapy, salicylic acid or immunomodulators. Predictive models may further help estimate recurrence risk, assisting in the development of more personalized treatment strategies [36].

Future directions include the integration of multimodal data, combining dermoscopy with clinical metadata, virological subtyping and patient-reported outcomes, to build more robust and individualized predictive models [37]. Such systems could improve therapeutic decision-making and resource allocation, especially in settings with limited access to dermatology specialists.

In summary, warts are highly prevalent benign skin lesions with diverse clinical presentations that often complicate diagnosis and management. AI-assisted dermoscopy has shown promising potential to improve diagnostic reliability, guide therapeutic monitoring and enable personalized care strategies. As these technologies mature, their incorporation into routine practice may improve efficiency, reduce diagnostic errors and enhance patient outcomes across varied healthcare contexts [34-37].

Conclusion

Warts are common viral skin infections caused by Human Papillomavirus (HPV), presenting with variable morphology and distribution that often complicate accurate diagnosis. This study demonstrates that an Artificial Neural Network (ANN), particularly attention-based frameworks, can effectively analyze dermoscopic images to differentiate warts from healthy skin and provide consistent diagnostic support. By minimizing interobserver variability and streamlining dermoscopic evaluation, AI has the potential to improve early recognition, guide treatment decisions and enhance patient monitoring. Although the current model primarily focuses on binary classification warts versus non-wart skin future developments should incorporate a broader range of cutaneous viral and non-viral lesions to strengthen differential diagnosis. Integrating dermoscopic findings with clinical metadata, lesion history and patient-specific risk factors could further enhance diagnostic precision and support individualized management. In summary, AI-driven dermoscopic analysis holds promise for improving diagnostic reliability, reducing unnecessary procedures and optimizing therapeutic strategies in patients with warts. Broader clinical adoption of such systems could also reduce healthcare burden by enabling earlier intervention and more efficient use of dermatology resources.

Conflict of Interest

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Financial Disclosure

This research did not receive any grant from funding agencies in the public, commercial or not-for-profit sectors.

Acknowledgment

Acknowledge those who provided support during the study.

Consent To Participate

The authors certify that they have obtained all appropriate patient consent.

Data Availability and Consent of Patient

Data is available for the journal. Informed consents were not necessary for this paper.

Author’s Contribution

All authors contributed equally in this paper.

References

  1. Bruggink SC. Warts transmitted in families and schools: a prospective cohort. J Infect Dis. 2013;207(9):1429-38.
  2. Sterling JC, Handfield-Jones S, Hudson PM. Guidelines for the management of cutaneous warts. Br J Dermatol. 2001;144(1):4-11.
  3. Lipke MM. An armamentarium of wart treatments. Clin Med Res. 2006;4(4):273-93.
  4. De Koning MN. HPV types present in verrucae vulgares: A meta-analysis. J Clin Virol. 2009;44(3):145-9.
  5. Garland SM, Steben M. Genital warts. Best Pract Res Clin Obstet Gynaecol. 2014;28(7):1063-73.
  6. Massing AM, Epstein WL. Natural history of warts: A two-year study. Arch Dermatol. 1963;87:306-10.
  7. Lallas A. Dermoscopy in general dermatology. Dermatol Clin. 2013;31(4):679-94.
  8. Tschandl P. Diagnostic accuracy of dermatoscopy for cutaneous warts. Br J Dermatol. 2018;178(4):877-83.
  9. Esteva A. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542(7639):115-8.
  10. Navarrete-Dechent C. Artificial intelligence and dermoscopy: From research to clinical practice. Dermatol Clin. 2021;39(4):587-96.
  11. Lipke MM. An armamentarium of wart treatments. Clin Med Res. 2006;4(4):273-93.
  12. Garland SM, Steben M. Genital warts. Best Pract Res Clin Obstet Gynaecol. 2014;28(7):1063-73.
  13. Bruggink SC. Impact of warts on quality of life in primary schoolchildren. Br J Dermatol. 2013;168(1):172-7.
  14. Berman B. Verrucae in immunocompromised patients. J Am Acad Dermatol. 1992;26(2):210-7.
  15. Doorbar J. The biology and life-cycle of human papillomaviruses. Vaccine. 2012;30(Suppl 5):F55-70.
  16. Esteva A. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542(7639):115-8.
  17. Tschandl P. Expert-level diagnosis of nonpigmented skin cancer by combined convolutional neural networks. JAMA Dermatol. 2019;155(1):58-65.
  18. Tschandl P. Diagnostic accuracy of dermatoscopy for cutaneous warts. Br J Dermatol. 2018;178(4):877-83.
  19. Zalaudek I. Dermoscopy of common warts: new insights and perspectives. Dermatology. 2006;212(1):56-62.
  20. Brinker TJ. Deep learning outperformed 136 of 157 dermatologists in a head-to-head dermoscopic melanoma image classification task. Eur J Cancer. 2019;113:47-54.
  21. Lallas A, Apalla Z, Argenziano G. Dermoscopic patterns of common warts: A prospective study on 100 patients. J Am Acad Dermatol. 2013;68(3):381-7.
  22. Nami N, Happle R, Argenziano G. Accuracy of dermoscopy in the diagnosis of viral warts: A multicenter study. Dermatology. 2014;229(3):185-90.
  23. Kittler H, Marghoob AA, Argenziano G. Diagnostic utility of dermoscopy in non-melanocytic skin tumors. Semin Cutan Med Surg. 2009;28(3):142-9.
  24. Tschandl P, Rinner C, Apalla Z. Human-computer collaboration for skin cancer recognition. Nat Med. 2020;26(8):1229-34.
  25. Winkler JK, Fink C, Toberer F. Artificial intelligence in dermatopathology: Classification and prediction. Br J Dermatol. 2019;181(4):767-9.
  26. Esteva A, Kuprel B, Novoa RA. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542(7639):115-8.
  27. Sterling JC, Gibbs S, Haque Hussain SS, Mohd Mustapa MF, Handfield-Jones SE. British Association of Dermatologists’ guidelines for the management of cutaneous warts 2014. Br J Dermatol. 2014;171(4):696-712.
  28. de Koning MN, Quint WG, Bruggink SC, Gussekloo J, Feltkamp MC, Bouwes Bavinck JN. High prevalence of cutaneous warts in elementary school children and the role of HPV transmission. Br J Dermatol. 2015;172(1):196-201.
  29. Cubie HA. Diseases associated with human papillomavirus infection. Virology. 2013;445(1-2):21-34.
  30. Bruggink SC, Gussekloo J, Berger MY, Zaaijer K, Assendelft WJJ, de Waal MWM. Cryotherapy with liquid nitrogen versus salicylic acid for cutaneous warts in primary care: randomized controlled trial. CMAJ. 2010;182(15):1624-30.
  31. Zaballos P, Lallas A, Argenziano G. Dermoscopy of infectious skin disorders. Dermatol Clin. 2018;36(4):349-60.
  32. Lallas A, Apalla Z, Argenziano G. Dermoscopic patterns of warts and their mimickers: Clinical and histopathologic correlates. Dermatol Clin. 2013;31(4):629-35.
  33. Bruggink SC, Eekhof JAH, Egberts PF, van Blijswijk SCE, Assendelft WJJ, Gussekloo J. Warts transmitted in families and schools: a prospective cohort. Pediatrics. 2013;131(5):928-34.
  34. Esteva A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542(7639):115-8.
  35. Tschandl P, Codella N, Akay BN. Comparison of the accuracy of human readers versus machine-learning algorithms for pigmented skin lesion classification: an open, web-based, international, diagnostic study. Lancet Oncol. 2019;20(7):938-47.
  36. Winkler JK, Fink C, Toberer F, Enk A, Deinlein T, Hofmann-Wellenhof R. Association between surgical skin markings in dermoscopic images and performance of a deep learning algorithm in melanoma recognition. JAMA Dermatol. 2019;155(10):1135-41.
  37. Du-Harpur X, Watt FM, Luscombe NM, Lynch MD. What is AI? Applications of artificial intelligence to dermatology. Br J Dermatol. 2020;183(3):423-30.

Mahajabeen Madarkar1*, D Purshotam B1, Muskan Jain2

1Professor and Head of the Dermatology Department, S R Patil Medical College, Badagandi, Bagalkot, India

2Himalayan Institute of Medical Sciences, Jollygrant, Dehradun, Uttarakhand, India

*Correspondence author: Mahajabeen Madarkar, Associate Professor and Head of the Department, SR Patil Medical College, Badagandi, Bagalkot, India;
Email: mahajabeenmadarkar@gmail.com

Mahajabeen Madarkar1*, D Purshotam B1, Muskan Jain2

1Professor and Head of the Dermatology Department, S R Patil Medical College, Badagandi, Bagalkot, India

2Himalayan Institute of Medical Sciences, Jollygrant, Dehradun, Uttarakhand, India

*Correspondence author: Mahajabeen Madarkar, Associate Professor and Head of the Department, SR Patil Medical College, Badagandi, Bagalkot, India;
Email: mahajabeenmadarkar@gmail.com

Copyright© 2025 by Madarkar M. All rights reserved. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Citation: Madarkar M, et al. Artificial Intelligence: Assisted Dermoscopy for the Diagnosis of Warts. Jour Clin Med Res. 2025;6(3):1-10.

Download PDF