It is reported that pattern analysis, and not simplified algorithms, improves the diagnostic accuracy by 5-30% with respect to simple clinical observation. However the evaluation of the many morphological characteristics of PSLs by ELM is often extremely complex and subjective: the pattern analysis is based on the recognition of certainly morphological and chromatic characteristics whose objective definition is really problematic and it is left to the training and experience of the individual observers. In one side, there are difficulties due to the subjective laws of perception. On the other side, there are semantic problems arising from certain primitives of language difficult to be standardized and described with unambiguous terms. In recent years, we have witnessed an attempt to move away from qualitative "common sense and experience guided instant recognition of patterns" diagnoses towards the more or less mechanic, time-consuming, assignment of "scores" to certain visual characteristics of the lesion. To obviate these problems certain research groups have developed equipments and methods that enable objective evaluation of the many parameters that have emerged with decades of experience. In 1991 we started to develop a dedicated software able to automatically measure a certain number of objective variables able to describe the pigmented skin lesions's aspect. One year later we developed an artificial network able to provide a diagnostic responce to the expert as an aid to the early detection of melanoma.
The couple skin-lesion can be numerically seen as an "object", i.e. the lesion, over a "background", i.e. the healthy skin. Our software, DBdermo/DDAX3 evaluates a lesion through a series of variables employing sophisticated pattern recognition techniques oriented to the diagnosis of melanoma. In our approach we developed a digital aided analyzer able to explain the reasons of the aided diagnosis showing results by samples. In the last 15 years we published many scientific papers related to the results obtained using this new technology.
In 2001 we engineered a new integrated System employing high resolution digital live cameras. In 2004 we developed an innovative software, named DDAX3, able to interface the Heine Delta20 and the Dermlite 3Gen dermatoscopic adapter for a cheap but efficient and robust solution.
The main goal of the aided diagnosis is to develop machines that will help clinicians in their daily practice. For this reason, the instrument must be easy to use, fast and not based on subjective evaluation. DB-Mips and DDAX3 employ robust and efficient classifiers. The results obtained during the last decade by independent clinicians who used the DB-Mips are reported on the bibliography. Please note, from the bibliography, how the results obtained by different groups of clinicians show increased and coherent accuracies over the years. Subjective , non-reproducible algorithms, have been replaced by objective and reliable variables.
Telemedicine is a growing field for the application of the new technologies and deals mainly with the remote process of remote diagnosis, performed through proper media. In the early diagnosis context the diagnostic support comes from the patients' information, the series of images of their lesions, plus global-view pictures taken in order to document the general situation. The diagnostic process is based on many factors involving questions to the patient, touching of some lesions, looking at detailed views of a particular site as well as of the image of the lesion. Much of this important information is not available on the remote consultation site. For this reason, we stress the fact that the image quality is more important than in the normal examination context. In many of our studies, we demonstrated the possibility for clinicians to perform an accurate remote diagnosis without significant differences if compared with local check up. Some innovative software like DDAX3, infact, performs the automatic parameters evaluation and sends them together with the image through a simple visual interface. Image quality, objectivity and the remote observer's experience provides a reliable diagnosis.
|
|---|
