Skin cancer is the most commonly diagnosed cancer in the United States, affecting one in five Americans during their lifetime. While basal cell carcinomas (BCCs) and squamous cell carcinomas (SCCs) account for most cases, melanoma remains the deadliest form.
Early detection still depends heavily on clinician experience and visual inspection from a specialist. As dermatologist shortages and long referral delays persist, primary care physicians are increasingly responsible for identifying suspicious lesions, often without the tools, training or time needed to do so with confidence and accuracy.
In recent years, a new wave of diagnostic technologies has emerged. Advances in machine learning, AI and novel optical spectroscopic techniques are reshaping how clinicians evaluate lesions and make referral decisions.
Emerging Trends Shaping Skin Cancer Diagnostics
As the diagnostic landscape evolves, several patterns are emerging across both clinical research and market innovation. These trends likely reflect broader shifts in how healthcare systems are approaching early cancer detection, clinical support and the integration of AI-enabled tools.
Improvements And Limitations In AI-Based Image Analysis
AI models trained on dermoscopic and smartphone images have made impressive strides, with several studies showing dermatologist-level performance under controlled conditions with an early such study published in Nature in 2017. However, translating this accuracy into everyday practice has proven to be elusive.
Lighting inconsistencies, variation in image quality and differences in skin tone may limit the utility of image-based systems. For example, Memorial Sloan Kettering researchers published a study showing that commercially available image-based skin cancer tools had a mean sensitivity for melanoma of only 28%; none of these tools are authorized for sale by FDA, however, some are still able to be accessed in the US and others in Europe through app stores. For many primary care environments, where clinicians lack dermoscopy training and operate under tight visit times, image-dependent tools still require more standardization before they can be scaled reliably.
Optical Spectroscopy: A More Objective Approach
Noninvasive optical technologies are increasingly used in innovative skin cancer diagnostics, driven by their ability to capture subcellular structural and biochemical information that visual tools cannot.
Techniques such as Elastic Scattering Spectroscopy (ESS) and Raman spectroscopy generate objective tissue signatures unaffected by user skill, lighting conditions or melanin content, ameliorating long-standing health equity and consistency gaps in dermatologic evaluation.
Furthermore, these spectroscopy-based methods offer rapid and repeatable assessment, reducing the need for unnecessary referrals and biopsies and supporting earlier detection across diverse patient populations.
Clinical Workflow Integration
Regardless of accuracy, a diagnostic tool will not gain traction if it slows clinicians down. Primary care visits often last just minutes, and new technologies must fit within those constraints. Solutions that deliver results in seconds, require minimal training and integrate naturally into existing clinical workflows will naturally outperform more complex alternatives.
As health systems continue shifting toward value-based care, leaders are prioritizing innovations that enhance clinician efficiency and confidence without adding operational burden.
The Path Forward
While many emerging diagnostic tools reflect the broader evolution of AI-enabled decision support, I believe the development and validation of ESS offers particularly useful lessons for healthcare leaders.
Designing Tools For Real Clinical Environments
Diagnostic technology must match the realities of frontline practice. In primary care settings, clinicians vary in dermatologic training, and decision making often depends as much on confidence as on clinical findings.
My own company’s study, which evaluated our tool’s ESS-based assessment of suspicious lesions, highlights the importance of real-world variability. The study utilized 22 primary
care sites and over 1,000 patients, helping to validate it in the same chaotic, heterogeneous environments in which clinicians actually practice.
Even in these conditions, the study reported an overall sensitivity of 95.5% and a negative predictive value of 96.6%. Rather than evaluating technology under idealized conditions, new diagnostic tools increasingly need to demonstrate resilience under the real-world variability that clinicians navigate on a daily basis.
Balancing High Sensitivity With Operational Impact
That emphasis on sensitivity shaped our tool’s performance profile. The approach evaluated in the trial was designed to minimize the risk of a malignant lesion being overlooked. While these figures demonstrate strong alignment with safety-focused design, they also highlight a broader consideration for healthcare leaders adopting emerging diagnostic tools: high sensitivity often influences downstream operational patterns.
Technologies that identify most cancers may also increase the number of high-risk findings requiring additional assessment, whether biopsy, imaging or referral. This dynamic isn’t unique to spectroscopy; it appears across many categories of AI-enabled diagnostics where early detection is prioritized.
The strategic question for organizations isn’t simply whether a tool is accurate, but how its performance characteristics interact with staffing, referral capacity, risk tolerance and patient flow. Leaders implementing new technologies increasingly evaluate them not as standalone solutions but as contributors to a broader clinical ecosystem in which safety, efficiency and resource allocation must be balanced thoughtfully.
Supporting Clinical Judgment Through Quantitative Inputs
Findings from a companion clinical utility study of our tool illustrated how structured, objective data streams can strengthen clinical judgment in managing suspicious lesions, particularly when clinicians face uncertainty.
When 108 primary care physicians evaluated 100 digitally presented lesions with and without ESS output, diagnostic sensitivity increased from 71.1% to 81.7% and management (referral) sensitivity rose from 82.0% to 91.4% with device data available. This corresponded to a halving of the false-negative rate. As a result, confidence also shifted. The proportion of high-confidence management decisions increased from 36.8% to 53.4%.
These results show that tools that provide interpretable, structured information tend to strengthen decision quality in settings where training, experience and case complexity vary. Unlike fully automated systems, spectroscopy-based outputs function as cognitive scaffolding, helping clinicians calibrate uncertainty while maintaining control of final decisions.
For healthcare leaders, the broader takeaway is how augmentative technologies that enhance accuracy rather than replace expertise can prove the most adaptable in real-world clinical environments.