Hello, everyone. It is my great pleasure today to give this presentation about Line Field Confocal Optical Credence Tomography, or LCOCT, for the Skin Cancer Imaging course for the American Academy of Dermatology. My name is Mariano Suppa. I am an Italian dermatologist working in Brussels, Belgium. Those are my conflicts of interest. In particular, I was able to use free of charge a prototype of LCOCT in my department to perform clinical research. So let's get down to business. What is LCOCT? It is a new imaging technique combining the technical advantages of conventional OCT and confocal microscopy with line illumination and detection. It is characterized then by high resolution and at the same time high penetration. High resolution like confocal microscopy and high penetration like conventional OCT. The first prototype was able to produce only vertical imaging, such as in histopathology, with a high level of details. Then later on, the machine was able to produce also horizontal imaging like in confocal microscopy. And then, of course, the device is able to recombine vertical and horizontal to give a three-dimensional reconstruction of the skin, as you can see here. And then a very important addition in 2020, the addition of an integrated dermoscopic camera, a great characteristic because it enables us to know at every single moment where we are on the lesion we are examining. Take this as an example. So this is a basal cell carcinoma. Here you can see an ovoid nest and then the corresponding LCOCT. You see the ovoid nest corresponds to this BCC lobule. We are able to do the borders with our cursor of the BCC lobule and they nicely correspond to dermoscopy. Then we change the filtering and we are able to appreciate cytological details inside the lobule with a level of precision which was, believe me, unprecedented in the OCT field until now. And again, we are able to go along the borders and this nicely corresponds to dermoscopy here. So great characteristics and the ultimate addition was the addition of artificial intelligence very recently. And what can artificial intelligence do applied to LCOCT? On the one hand, it can help us quantifying skin structures which are otherwise impossible to quantify. And on the other hand, it can directly help us with the diagnosis by producing probability diagnostic scores. There is a lot to talk about LCOCT and my time today is limited. I will just tell you that the vast majority of things that you need to know about the machine and especially about the role of the machine for melanocytic and non-melanocytic lesions is summarized in these two papers. But today I will just be very practical and I will just tell you what the machine can do in different types of skin cancer. Let's start with non-melanoma skin cancer, non-melanocytic lesions then. So the principal indication of LCOCT is basal cell carcinoma. We describe many criteria under LCOCT of basal cell carcinoma, but the most important one, such as in histopathology, is the presence of lobules, of BCC lobules, which you can see here. This is the epidermis. This is the dermis compressed by the presence of this big lobule and the lobule is characterized by this very peculiar pattern, which we named the millefeuille pattern because it resembles the shape of the eponymous French delicacy. The shape and the distribution and the connection of the lobules to the epidermis points the diagnosis towards the different BCC subtypes. Here you can see a superficial BCC, lobules connected to the epidermis, so stratum corneum here, epidermis, lobules of BCC connected to the epidermis and then the underlying dermis. In nodular BCC you see that the lobules are separated from the epidermis and in infiltrative BCC the lobules are separated from the epidermis and at the same time they have a different shape, they are branched, a more convoluted, complicated shape. Here you can see the differential diagnosis of nodular BCC. You see a nodular BCC here, a lobule in the dermis with millefeuille pattern, a sebaceous hyperplasia. Here you see again a lobule separated from the epidermis but the quality of the lobule inside is different, loss of signal after the first layers and the first layers are granular, so this corresponds to the presence of sebocytes at the top of the layers and here you have a dermal nebus in which you have a very different pattern, so no millefeuille like in nodular BCC. The differential diagnosis of superficial BCC, here you can see a superficial BCC, lobule attached to the epidermis and then an actinic keratosis in which you might think that there are lobules but in reality it is just the epidermis, a typical epidermis protruding towards the dermis and here you have a clearly invasive squamous cell carcinoma in which you see the protrusion of the epidermis and a frank invasion here. Artificial intelligence applies to basal cell carcinoma in the sense of the ability of the machine to give us the probability scores of diagnosis. Look at that, this is a lesion which might or might not be basal cell carcinoma, we apply LCOCT and what do we see? The expert eye actually sees lobules attached to the epidermis but if you're not an expert and you want help from the artificial intelligence, you can just switch it on on the machine, the machine tells you that it is a probability 100% of basal cell carcinoma and it also tells you where the machine is looking to tell you this is a basal cell carcinoma, so very helpful not just to have the diagnosis but also to learn where you should look to make the diagnosis. Another example, it might or might not be a basal cell carcinoma and artificial intelligence, zero probability of basal cell carcinoma, I can tell you this is an actinic keratosis and another example, again BCC or non-BCC, here the non-expert eye might think I see a lobule so it must be BCC but actually you can see a granular pattern here and loss of signal at the bottom of the lesion and indeed the artificial intelligence tells you this is not BCC and it was a sebaceous hyperplasia. And actually now we have demonstrated that the addition of artificial intelligence is very helpful in increasing the diagnostic performance because we performed a study which will be very soon published with 43 European dermatologists on 200 equivocal lesions, the lesions were evaluated half with the help of LCOCT and the other half with the additional help of artificial intelligence. After a washout period of two to four weeks we did exactly the same but switching the presence of artificial intelligence and we found that the addition of artificial intelligence can take the diagnostic accuracy very high up to 93 percent with a significant gain of precision and this gain in diagnostic precision is particularly evident for novices of course and the same goes for the confidence level of the diagnosis and this is the new layout of artificial intelligence for BCC which you will be able to find in the machine very soon on the market. What about actinic keratosis and squamous cell carcinoma? Basically they have at the epidermal level more or less the same characteristic atypia, cellular atypia and pleomorphism of the epidermis, so different shapes and sizes of the keratinocytes. Those changes are just more evident in squamous cell carcinoma than in actinic keratosis. The difference is the dermal-epidermal junction which is preserved in actinic keratosis and disrupted in invasive squamous cell carcinoma. This is a three-dimensional example of an actinic keratosis. Look at the epidermis, plenty of atypical cells, different size and shapes of the keratinocytes, nicely corresponding to histopathology and the presence all over the lesion of a well-defined dermal-epidermal junction such as in the histology corresponding. Here if you put the machine at the periphery of the lesion you see an area of actinic keratosis slightly protruding towards the dermis. Here more towards the center you see an area of bigger protrusion of the epidermis towards the underlying dermis and so this is an in-situ area of squamous cell carcinoma and when you put the probe frankly inside the lesion you see a clear-cut squamous cell carcinoma because you see a clear invasion of the dermis. LCOCT can also help in the distinction of different subtypes of actinic keratosis and we demonstrated that by a comparison with histopathology as you can see here. And even more interestingly, LCOCT also nicely compares with the histological score of severity of actinic keratosis which is called protrusion score or pro-score. Basically the more the lesion protrudes towards the dermis the more at risk the lesion is. But I can tell you that you really need a very expert eye to calculate the pro-score on just simple LCOCT images of actinic keratosis. So what can we do to increase our ability? Again artificial intelligence but in this case it will help us in the quantification of skin structures which are otherwise impossible or very difficult to quantify. For the protrusion score the undulation of the dermal-epidermal junction as you can see here and you can do this even live so at every second you have the calculation of your pro-score. Very nice feature. And also another help of artificial intelligence is to calculate the atypia via an atypia score of the keratinocyte. We published this on Nature Scientific Reports. As you can see here the machine is able to tell us what are the nuclei of the keratinocytes which are atypical and they are colored in red. Of course more prominent in an area of frank actinic keratosis but still present nearby in an area of field of cancerization. And you can also compare different actinic keratosis so the more red you see the more the actinic keratosis will be atypical or vice versa you can actually select the blue color so low level of atypia and you can see immediately what is the actinic keratosis with the less atypia in three dimensions. Also the machine is very helpful to monitor non-invasive treatment of melanoma skin cancer for example imiquimod. Here presence of a BCC lobule at the baseline and absence of the same lobule after the end of the treatment. Same goes for actinic keratosis and tirbanibulin. Baseline frank actinic keratosis, a lot of inflammation after eight days and apoptosis and complete resolution after two months. And the same with tirbanibulin you can do with the help of artificial intelligence as it was shown here by these german authors. And also the same goes with cryotherapy, help of artificial intelligence and actinic keratosis at the baseline a lot of undulation of the dermal epidermal junction and the same lesion after the healing of cryotherapy. The lesion is way more flat and the pro score goes down. But also and those are still unpublished data with cryosurgery. This is with my friend Paola Pasquale who came to my department to teach me cryosurgery. This is a cryosurgery of a basal cell carcinoma on the nose. It works for BCC and also Bowen actinic keratosis. Here you see the baseline of a superficial BCC, lobule attached to the epidermis, nodular BCC and Bowen disease. This is what happens after 32 minutes after the cryosurgery. Basically a lot of inflammation coming in and with a very clear cut cleavage of the lesion and three months after complete resolution. So you can non-invasively follow non-invasive treatments of melanoma skin cancer. And another very nice characteristic, the presurgical delineation of margins. So this is a very difficult to define basal cell carcinoma. This was my first margin guess based on my dermoscopy. I checked this with LCOCT and the machine told me here, be careful, you still have basal cell carcinoma. And this is the presence of basal cell carcinoma in that particular area. You can see it here with the help of artificial intelligence. So what did I do? I just widened my margins only in the incriminated part if you want and then I removed the lesion following these margins and the surgery was with free excision margins. Another example for a recurrence, this is a recurrence of a previously excised basal cell carcinoma, which we demonstrated with the help of artificial intelligence and applied to LCOCT again, as you can see here. And so what did I do? I just wanted to check that apart from this recurrence, also the rest of the scar was fine. And so I designed my margins and I checked them with LCOCT. And so here you can see all the way that I did with my examination. And again, here the cursor is moving and when it's yellow, it means that the machine found presence of basal cell carcinoma. So a very nice way to check what you do before surgery in a reliable way. What about melanocytic lesions? What can we see? I told you LCOCT has a vertical imaging and an horizontal imaging at the same time. So of course, in the vertical imaging, you can see criteria which are very similar to histopathology, like in this nevus. And the horizontal view, you can see criteria which are very similar to those of confocal microscopy, which were discussed before. And the same goes for nevi, in situ melanoma, like here, vertical and horizontal, and invasive melanoma. Again, vertical, like histopathology, and horizontal, like confocal microscopy. The criteria which we described with LCOCT are many, but we found that some criteria are very important to differentiate melanoma from nevi. And those are the presence of pagetoid cells in the epidermis, as you can see here. The presence of an irregular architecture of the epidermis, as you can see here, compared to a regular architecture. But also the disruption of the dermal-epidermal junction, which is conserved here in a nevus and disrupted in a melanoma. And finally, the presence of a particular criteria, which is called clefting, like the clefting that you can see in basal cell carcinoma. And this is a new criterion, which you will hear a lot in the future, I think. And the newest addition, which I was telling you about, the integrated dermoscopy, now is also colocalized, meaning that you can take a dermoscopic picture before the examination, then you do your LCOCT, this is the integrated dermoscopy, and the machine tells you immediately where you are on your lesion, which is, of course, of great importance, not just for the BCC margins, as I showed you just before, but also to know in which portion of the lesion you are for a melanocytic lesion, and especially if you want to go and look what lies behind a particular and maybe worrisome dermoscopic criterion. And then once you are in the area of interest, you can take a cube of lesion and then you can go and check what you have done. For example, here in this particular melanocytic lesion, I took the three-dimensional stacks in five areas. What I was telling you just now, the interest of go and see what lies particular dermoscopic criteria, is already published. So, we published a study investigating the correlation between particular dermoscopic criteria of melanocytic lesions and LCOCT. And we found nice correlates of everything. And so, pigment network, atypical pigment network, brown dots and globules, structureless areas, and also other melanocytic criteria are all described with LCOCT. And this is very important because it will teach you to look at dermoscopy in a different way and just stop to think about morphological criteria, but just start thinking directly, oh, I think I'm seeing cells at the dermal-epidermal junction, or maybe I'm seeing nests at the dermal-epidermal junction or deep down the dermis and so on. So, it's a nice bridge between what we can see morphologically during our clinical routine and the reality of histopathology. And the same we did with basal cell carcinoma. This is a first report about a particular criterion, but we are coming up with a new study encompassing the entire spectrum of basal cell carcinoma criteria explained by LCOCT. And for example, here you have a leaf-like area, what we call a leaf-like area in a superficial BCC. And you know that this particular criterion here corresponds to this BCC lobule attached to the epidermis. So, you know the next time you see this criterion that you are in front of a superficial BCC. And this other point of the lesion corresponds very nicely to this other superficial basal cell carcinoma lobule, therefore attached to the epidermis. So, in conclusion, LCOCT is a new, exciting technique, in my opinion, that combines the technical advantages of confocal microscopy and conventional OCT, and all that with integrated and colocalized dermoscopy. It displays a very high level of similarity with histopathology. It now integrates artificial intelligence, and this is very exciting. And in my opinion, it has a clinical impact for both non-melanocytic and melanocytic lesions, as well as other tumoral and non-tumoral conditions which we couldn't treat today. And it is important not just in the diagnosis of these lesions, but also in the determination of their prognosis, in particular for the subtype discrimination, but also in the determination of the presurgical margins, and in helping us to monitor non-invasive treatments. And with that, I thank you very much for your attention. Bye.

LCOCT

skin cancer

confocal microscopy

artificial intelligence

diagnostic accuracy

non-invasive treatments