METODO E APPARECCHIATURA DI MAPPATURA TRIDIMENSIONALE DI UNA PORZIONE DELLA CUTE DI UN PAZIENTE ; Skin tridimensional mapping method and device to support the melanoma diagnosys

Digital medical imaging (DMI) involves a wide range of data acquisition and processing techniques. Recent developments not only significantly improve the diagnostic performance but also offer new hopes of treatment for many diseases. Melanoma is among the leading causes of death due to cancer and its incidence has increased over the last decades. However, accurate and timely diagnoses allow a 5-year survival rate significantly higher than 90%. The clinician makes his own judgment on the risk that the pigmented lesion under observation is a melanoma based on several well established dermoscopic parameters. This assessment is error prone due to subjective evaluation of each parameter and to the intrinsic diagnostic limitation of the procedure, especially in the cases of patients having a large number of moles. Therefore, unbiased objective assessment tools and new parameters are eagerly needed to improve diagnostic accuracy. Digital photogrammetry allows the high-resolution reconstruction of a three-dimensional (3D) structure from a series of partially overlapping images taken from different positions. The resulting point cloud, as well as the corresponding 3D model, is photorealistic and faithful to the reality less than a scale factor. The Structure-from-Motion photogrammetry (SfM) is a recent development that comes from introduction of computer vision methods in photogrammetry [3]. Such a technique, which enables a fast data acquisition and a highly automated data processing, is particularly suitable for DMI. Since the innovative studies in the early 2000s, SfM is used in DMI, for example in postural analysis, orthopedics, ophthalmology, dermatology, dentistry/maxillofacial surgery, craniofacial analysis, biplanar radiography, forensics. In dermatology, SfM is used e.g. for the low-cost remote diagnosis of suspect moles in deprived areas. Digital Infrared Thermal Imaging (DITI) allows a reconstruction of the surface temperature profile of an observed object from the electromagnetic radiation emitted by this object in the thermal IR band (1-10 μm), detected by means of an IR camera. Nowadays, this technique is largely used in DMI. A melanoma is typically characterized by a higher temperature with respect to the surrounding healthy tissue. This is because the metabolism is higher in the lesion area and an increase in blood supply occurs to support the tumor growth [7]. Performance of dynamic DITI, in which the skin is temporarily heated or cooled to study the process of recovery of the initial temperature, is significantly better than the static DITI one, which is based instead on the analysis of the temperature difference of a suspect mole with respect to the surrounding skin [8]. In general, some methods for the quantitative non-destructive laboratory testing of materials, e.g. the evaluation of running window temperature contrast [9], can be innovatively used for the diagnosis of melanoma. There are interesting examples of “lock-in” DITI in dermatology, where the thermal stimulation of the skin is cyclic and methods of processing of periodic signals can thus be used [10]. Although an IR camera is an expensive device, there are also examples of low cost systems for melanoma recognition [11]. All these facts suggest the fusion of data in the visible and thermal IR bands to provide an SfM-based 3D model of the patient skin surface textured with visible and IR radiometric information. Some interesting examples of integration between SfM and other medical DMI techniques are already available [12]. However, the registration of IR images on images in visible band is not trivial and is still an open topic [13]. The proposed Research Project (RP) aims at developing an innovative DMI system where SfM and DITI are integrated to provide a high accuracy, reliable, multimodal 3D model of all or part of the body surface in order to assist the dermatologist in the diagnosis of melanoma. The emphasis is on high performance in terms of speed and success in the correct diagnosis, user-friendliness, affordable cost for a typical health system and complete compatibility with the databases currently used in DMI (the output data will be provided in DICOM format [14]). An important innovation factor is the integration of SfM and DITI in a ready for use, low cost device specific for the melanoma detection, filling a gap in DMI because the currently available solutions are very expensive systems or are prototypes that cannot be directly used by a physician. An application for an European patent on the proposed system is included in the RP. In the updated formulation of the Project, the classification of nevus is performed using deep learning techniques (deep learning) such as convolutional neural networks convolutional neural networks, CNN). However, in order to maximize the information provided to the doctor in the case of nevus recognized as possible melanomas, the diagnostic indices corresponding to the indicators (i) - (vii) are calculated, as well as, in the case of multitemporal observations, their variations. CNN data (structure, weights of neurons) are contained in the diagnostic database of skin lesions (skin lesion diagnostic database, SLDD).