Cutaneous malignant melanoma is a rapidly growing health problem in the U.S.. Its incidence climbs yearly. Some 27,000 Americans developed melanoma in 1989, and 6,000 will die of the disease. Early detection and treatment are essential for cure; there is no effective treatment for invasive or disseminated disease. Melanomas invading subcutaneous tissue have a 5-year survival of only 44%. In contrast, cutaneous melanoma is 100% curable if detected and removed in the early radial-growth phase of the disease. Unfortunately, current methods of melanoma screening are often inadequate. Patients with the dysplastic nevus syndrome and/or many moles are at highest risk, but are also those most difficult to assess. Documenting changes in nevi and detecting the presence of new moles are the two keys to effective screening. The important changes of melanoma in the radial growth phase are well known, and occur over months or years. They include slow, radial enlargement of a precursor lesion, irregularities of the lesion border, loss of skin markings, and color changes. Elevation of the lesion generally occurs only after the dangerous vertical-growth phase of development has begun. Approximately 90% of melanomas demonstrate these features of radial-growth. Many of the remaining nodular melanomas occur de novo, and must be directly identified as being new skin lesions. This project will improve monitoring for cutaneous melanoma by developing a precise, objective means of detecting skin changes suggestive of early melanoma. Large areas of a patient's skin will be optically imaged with a video camera and the electronic images stored in digital form. The patient is then re-imaged at a later date, and the two sets of images are compared by computer. Changes, particularly those suggestive of early-phase malignant melanoma, are then presented for physician examination. This is a new approach to computerized melanoma detection. First, it is designed to examine relatively large areas of skin in the course of a typical examination. It will therefore detect completely new or unrecognized lesions and flag them for careful examination. Second, it is designed to be a screening tool rather than a diagnostic device. Leaving diagnosis to the physician addresses the specific problems of current screening programs, yet maximizes expected usage and sensitivity to atypical presentations of disease, while minimizing cost. Third, it is readily adaptable for many other uses such as surveying large or patchy lesions, counting skin lesions, and objectively measuring non-melanoma skin diseases such psoriasis. These applications will enhance the system's usefulness, adaptability and distribution.
| Ngan, P M; Coombs, B D (1994) Segmentation of intensity basins in gray-scale images. Comput Biomed Res 27:39-44 |
