Our focus is on translation of a novel microscopy approach, femtosecond pump-probe microscopy, which our preliminary data shows can determine which (nominally) early-stage primary melanomas are instead metastatic cancer. This is important because the current clinical “gold standard” of staging (histopathology and sentinel lymph node biopsy (SLNB)) can assign early (non-metastatic) stages to tumors which are in reality metastatic cancers, which delays treatment and costs lives. In fact, more people die from melanoma after initial Stage I tumor diagnosis than after diagnosis of any higher grade. Today adjuvant therapies have made great strides (on late-stage tumors) but all FDA-approved therapies are restricted to metastatic (stage III or IV) melanomas because significant treatment-related adverse events are common. We believe adjuvant therapies could be more effective, and less toxic, if applied to supposedly early stage tumors which have already generated undetected metastases. Such “early adjuvant therapy” could have great benefits for disease control, reduced toxicity, and reduced health costs, but this requires a good marker for deciding which early-stage patients should go into such therapy, and existing markers have limited value. Our preliminary results show that we can identify such patients with pump-probe microscopy. The ultimate goal is routine identification of incorrectly classified early-stage lesions, at least from stages IIB/C and preferably from earlier stages as well, so the patient can be treated to interrupt disease progression. Specific Aim 1 focuses on optimizing multi-parameter pump-probe imaging to concentrate the clinically relevant contrast. The apparatus redesign features modulation schemes that keep the applied power constant while retaining complete control over pulse polarization and delays, plus detection schemes with angular resolution. Demonstrations start with melanin in model systems and melanoma cells and move on to biopsies, characterizing directional and polarization components of the pump-probe decay to maximize signal correlation with chemical or cellular melanin degradation. Specific Aim 2 focuses on maximizing diagnostic utility using patient biopsies from the Duke Biorepository to retrospectively diagnose metastatic melanoma and test the performance of the improved clinically relevant contrast. This work is closely connected to pathology, as we view the technology as complementing existing diagnostic protocols. Based on our very encouraging preliminary results, machine learning algorithms will pay a large role in our assessment of diagnostic utility. We expect to demonstrate that for at least Stage IIB/IIC tumors (7000 cases a year) pump-probe imaging can reliably segment this population and identify the patients who almost certainly need treatment beyond excision.