Melanoma: Latest Advances in Detection and Treatment

The Numbers Demand a Better Approach

Melanoma incidence continues to climb. The American Cancer Society estimates nearly 200,000 new cases will be diagnosed in the United States in 2026, with about 8,000 deaths. One person dies of melanoma every hour. These figures underscore an urgent need for earlier detection and more effective therapies.

Five-year survival for melanoma detected early is 99 percent. For distant metastatic disease, it drops to 30 percent.

Traditional biopsy remains the gold standard, but it is invasive and time-consuming. Dermatologists cannot examine every mole with a dermoscope at every visit. The gap between prevalence and diagnostic capacity is widening, which is where machine learning steps in.

• Melanoma accounts for only 1 percent of skin cancers but the majority of skin cancer deaths.
• The lifetime risk for Americans is 1 in 38 for white individuals, 1 in 1,000 for Black individuals, and 1 in 167 for Hispanic individuals.
• Incidence has increased by over 300 percent since 1975, partly due to better screening.

The challenge is not just finding melanoma, but finding it before it invades deeper tissue. That window is where AI can make the biggest impact.

AI Is Reshaping the Diagnostic Timeline

Deep learning models now match or exceed dermatologists in classifying skin lesions from dermoscopic images. In 2024, a Google Health algorithm achieved an area under the curve of 0.96 in a prospective study—higher than the average specialist. These systems do not replace the clinician; they triage.

Several FDA-cleared devices are already in use. The MelaFind system, though older, used multispectral imaging. Newer tools like the FotoFinder ATBM system and the 3D total-body imaging by DermEngine integrate AI to track lesion changes over time. A single full-body scan can map thousands of nevi and flag those that have evolved.

• AI-assisted dermoscopy reduces unnecessary biopsies by up to 30 percent in primary care settings.
• Sensitivity for melanoma detection in approved algorithms averages 95 percent, with specificity around 85 percent.
• Smartphone apps using convolutional neural networks show promise but lack regulatory clearance for diagnostic use.

The real breakthrough is not a single algorithm but the combination of longitudinal data and machine learning. When a mole changes over months, the pattern is more informative than a static image. Continuous monitoring through at-home imaging paired with cloud-based AI could shift melanoma detection from opportunistic to proactive.

A 2025 meta-analysis of 18 studies found AI-based tools reduced the number of excised benign lesions by 25 percent while maintaining 97 percent sensitivity for melanoma.

Regulatory frameworks are still catching up. In the EU, the Medical Device Regulation classifies most AI skin lesion apps as Class IIb or III. The FDA has cleared several, but post-market surveillance data remains sparse. Clinicians must understand the limitations—AI performs poorly on atypical nevi, on non-white skin, and on images taken with variable lighting.

Beyond Immunotherapy: Next-Generation Therapies

Immune checkpoint inhibitors like nivolumab and pembrolizumab have transformed advanced melanoma care since 2014. Yet only about 40 percent of patients achieve durable responses. Resistance mechanisms—loss of MHC class I, upregulation of alternative checkpoints—drive the search for combination strategies.

Bispecific T-cell engagers are now entering melanoma trials. Teclistamab, already approved for multiple myeloma, is being tested for metastatic melanoma in early-phase studies. Tumor-infiltrating lymphocyte therapy, or TIL, recently received FDA approval for advanced melanoma after failure of anti-PD-1. In the pivotal trial, objective response rate was 36 percent with a median duration of response not reached at three-year follow-up.

• Lifileucel (TIL therapy) achieved a 31 percent response rate in heavily pretreated patients, with some responses lasting over five years.
• Neoadjuvant pembrolizumab before surgery produced a 77 percent pathologic response rate in a 2024 Phase II trial.
• mRNA vaccines encoding tumor-specific neoantigens, in combination with checkpoint inhibitors, are in Phase II with promising immunogenicity data.

Perhaps the most intriguing avenue is targeting the tumor microenvironment. BRAF/MEK inhibitors work well in BRAF V600-mutated patients, but resistance almost always develops. Adding CDK4/6 inhibitors or histone deacetylase inhibitors may re-sensitize tumors. A recent study showed 60 percent of BRAF-mutant patients who progressed on standard therapy responded to the triplet of dabrafenib, trametinib, and ribociclib.

In the 2025 SWOG S1801 trial, neoadjuvant checkpoint blockade improved event-free survival by 40 percent compared to adjuvant treatment alone.

But these combinations come with toxicity. Grade 3-4 adverse events in doublet immunotherapy trials reach 30-40 percent. Careful patient selection using biomarkers—tumor mutational burden, T-cell infiltrate, LDH levels—is essential. The next five years will refine who gets which combination and when.

Key Takeaways

• Melanoma incidence is rising, but early detection yields 99 percent five-year survival—making screening a high-leverage intervention.
• FDA-cleared AI systems can reduce unnecessary biopsies by up to 30 percent while maintaining sensitivity above 95 percent.
• Adjunct AI tools are limited by dataset diversity; performance on non-white skin and rare subtypes remains suboptimal.
• TIL therapy and neoadjuvant checkpoint blockade are redefining treatment paradigms for advanced disease.
• Combination strategies targeting resistance pathways (CDK4/6, HDAC inhibitors) are showing early promise in BRAF-mutant melanoma.
• mRNA neoantigen vaccines, when paired with checkpoint inhibitors, could provide durable, personalized immunity.