 You as a doctor, give hope and health to many each day. There are many more to give hope and health to. And with Fujifilm, we are doing that together. Thank you everybody. And so we're going to veer a bit away from imaging to tumor biology and a brief outline of my talk is as follows. I will give you a brief primer into understanding breast cancer biology and just a little bit about what the breast imager needs to know. Of course, I'm going beyond my brief and my knowledge here is what I have gained from my pathologist and friends and friends at molecular biology. We will see how breast imaging is linked to tumor biology. What is the practical implication of these associations and how this has led to the developing interest in radiogenomics and the quest for imaging biomarkers that can meaningfully impact patient care? So the heart of the problem, the conundrum we face is that breast cancer is a heterogeneous disease. And so the presentation is varied. It depends upon the age, the environment, the lifestyle, and the treatment responses are varied from one individual to the next and prognosis to is varied. We are all familiar with the TNM staging, which we have used since decades for staging the tumors and for guiding our treatment. From our pathologist friends, we also get information about the histological grade for which they use a grading scheme called the Bloom Richardson Nottingham scheme, where they look at three features of the tumor. They being tubule formation, nuclear pleomorphism and mitotic activity. Each of these are assigned points. The lower the tubule formation, more the pleomorphism, more the mitotic activity, more aggressive the tumor is, higher the points are assigned. These points are summed up to reach a score and accordingly the tumor is graded from one to three. So let us look at a clinical scenario. You have three patients here, Lina, Mina and Nina. All of them have T2N0M0 disease that is tumor size somewhere between two to three centimeters. So they are what we call early stage invasive breast cancer. So here is a poll question. What would be the standard treatment protocol for these patients? Would we do imaging followed by biopsy, conservative breast surgery, chemotherapy or radiotherapy? And our choices are true or false. Is the poll on? Yes. Do we have the answers? So majority of you think that this is going to be the uniform treatment that we can offer to all these patients. But what we have learned from our experiences is that we cannot use this one size fits all approach. Traditional broad based chemotherapy is less effective in treating all our patients and sometimes we have off-site target side effects which are often more detrimental to the patient than the disease itself. Some patients will do well, some patients will recur in say even less than two years. And that is what takes us to precision medicine where we tailor the treatment according to the genetic makeup of the tumor and the tumor environment. And this personalized treatment will dictate our treatment choices, how we screen these patients and the prognosis of these patients. So coming back to our clinical scenario, here we have Lina who has an aggressive what we call her to enriched tumor. And she will not receive surgery up front. She will undergo neo adjuvant chemotherapy even though her disease is early stage. And we have good targeted therapy available for her in the form of anti her two monoclonal antibodies. And we see that she responds very well to this treatment. She after completion of her chemotherapy, she undergoes surgery. And the final histopath report says that there is a pathological complete response and that she has no residual tumor. So what started out as an aggressive disease, we end up with a very good extremely good prognosis for this lady. Mina on the other hand has a low grade hormone sensitive tumor. She will undergo surgery first. And we have ways of telling that she does not need chemotherapy following her surgery. And all that she needs is oral medication with anti estrogen drugs like tamoxifen or selective estrogen receptor modulators. And finally, Nina here has what we call a triple negative high grade breast tumor. She has a positive family history of breast cancer in her mother and ovarian cancer in a maternal aunt. She's advised genetic counseling. It turns out that she has a BRCA1 gene mutation. And she's advised risk reduction strategies in the form of bilateral mastectomies and opherectomy. So what is it that causes these tumors in these three different patients to behave in a different manner? It is because of the genetic makeup of the cell, the information of which is carried within the DNA. The DNA by a process of transcription gives you the MR produces the mRNA, which carries the blueprint for the expression of specific proteins. And these proteins in turn lead to triggering of metabolic pathways that lead to sustained cell proliferation that can make the tumor evade growth suppressors, activate invasion, induce angiogenesis, and resist cell death. It was in the year 2000 that a significant development happened. And that was the development of multi gene panel assays, which we are now quite familiar with thanks to COVID and also known as next generation sequencing. And it was in the year 2000 that Peru et al published this paper on the molecular portrait of breast cancer, where they did multi gene sequence assays on 78 breast tumors and they could identify hundreds of genes that were associated with cancers in these patients. And they found that they could broadly classify these cancers into four molecular subtypes. They were the luminal A, the luminal B, the herto enraged, and the basal-like tumors. Now to understand this, we need to go back to the basic functioning unit of the breast parenchyma, which is the terminal ductal lobular unit, the assinous, which is the milk producing unit of the breast. And this is lined by a two cell layer. The inner layer is the luminal cell, which lines the lumen outer basal myo epithelial layer. Now this luminal cell is responsible for milk production. It is ER sensitive, and it has estrogen receptors on its surface. Cancers that originate from these luminal cells are what we call as the luminal cancers. And because they arise from the luminal cell, they will recapitulate the gene profile of the luminal cells that is CK8 and 18, and also express estrogen progesterone and hormone activity genes. They can be further classified into luminal A, which has higher ER expression, less proliferative, and a very favorable response. And luminal B, which has less ER expression, more proliferative and slightly more aggressive. So as we said, these cancers have estrogen and progesterone on their cell surface. So when the estrogen attaches itself to these receptor sites, they will change the gene expression activity, which leads to expression of proteins that trigger metabolic pathways that cause uncontrolled cell proliferation. Now we can use this information to treat these cancers. With the help of anti-estrogen drugs like tamoxifen or aromatic inhibitors, they will compete with the estrogen for these receptor sites. They will block the activity of the estrogen and as a result, the cells will stop growing. The second group of tumors that we are, the HER2-enriched tumors, these are more aggressive and have a high proliferation. Fortunately, they are less common, about 10 to 15% of all breast cancers. What is peculiar to them is the overexpression of the HER2 receptor protein, that is the human epitoemyl growth factor, which in turn leads to sustained signaling of the nucleus to undergo cell proliferation. And finally, we have the basal cells, which we say the cell arrives from these basal myoepithelium cells. Again, they will show a high expression of the basal genes, that is CK5, 14, 17, and high proliferation genes. They are more aggressive and they have a poorer prognosis. Now, genetic assays are not easily available and so we have surrogate markers in the form of special stains that we use at immunohistochemistry and fluorescent in situ hybridization. So, a luminal A cancer will stain positively for the estrogen and progesterone receptors, but not for the HER2 receptor. A luminal B will stain less intensely for estrogen progesterone and may or may not stain for HER2 receptor. The HER2 enriched will not stain for the estrogen progesterone receptors, but stain strongly for the HER2 receptors. And finally, the basal cell will not stain for any of these receptors and these are what we call the triple negative cancers. Now, the HER2 status report comes as reads as 0 to 3 where 0 and 1 is considered negative, 2 is equivocal and 3 is strongly positive. So, if it is equivocal, we do another test which is the fluorescent inside the fish test and HER2 enriched tumors will take up this fluorescent stain. We also have a marker for cell proliferation which is called the key 67. A higher key 67 score indicates a more aggressive tumor and we use this to differentiate luminal A from luminal B. So, with a cutoff value of around 15 to 20%, so hormone positive tumors showing this high score are classified as luminal B. Now, we are aware that it is our genetic makeup that determines our facial features, the skin color, tone and the hair texture, etc. Similarly, it is the tumor genotype that will determine its imaging phenotype. So, here you have a 66-year-old lady who's come for a routine screening. We see this small irregular mass here. We can see speculations that are also beautifully seen on these DBT images and this is the ultrasound where we see a taller than wild again lesion with the speculations are well seen. There is distill acoustic shadowing. So, here is a poll question. What do you think? Which cancer is molecular subtype? Do you think this cancer is luminal A, luminal B, HER2 enriched or the basal type? Can we have the poll, please? Luminal A, B, HER2 enriched or the basal cell? Excellent. So, most of you think this is the luminal A and that was right. These are slow-growing indolent tumors. At biopsy, this was a low-grade tumor, strongly ER positive with a K67 score of 10%. And this is how luminal cancers present. They are very slow indolent tumors. And because of their slow growth, they may, if you had a mammogram that was of this lady, say from one or two years ago, maybe the lesion was there and it was overlooked on the mammogram because they do grow very slowly. And because of their indolent growth pattern, they incite a host reaction, which leads to this desmoplastic reaction causing these speculations and this dense acoustic shadowing. Now, here you have a 36-year-old lady who's presented with a left breast lump. You can see the asymmetric density in her left breast. On the DBT images, we see this irregular mass here and she also had pleomorphic microcalcification in a segmental distribution. On the ultrasound, there is a micro-lobulated lesion with parallel orientation. The disease was multifocal and there was also an abnormal enlarged auxiliary node. Again, can we guess on the molecular subtype? Luminal A, B, hereto-enriched or the basal subtype? Do we have the answers, Mithusha? So, most of you thought this was a luminal B. So, let's get back to this case. So, what we see here is at biopsy, this was a hormone negative and a hereto-enriched tumor and that is how often hereto-enriched tumors present. They present with calcification. They often have multifocal disease. They often are large tumors and also may have auxiliary nodes at presentation. Again, we have targeted therapy available for her. So, she was given neo-adjuvant chemotherapy with Trastuzumab along with the regular chemotherapy drug. And we see that we follow these patients during their NACT. We see that there is excellent resolution of her tumor, which is now no longer we are able to discern it on the DBT images. The calcifications appear to have reduced in number, though they persist. So, we did a wire localization and bracketing of these calcifications for the surgeon to guide them for the surgery. Here's a specimen radiograph showing the calcifications and in the specimen. And at final histopath, there was no residual tumor, no DCIs and no disease in the nodes. So, what she had achieved was again a pathological complete response. Here is a 33-year-old lady who presents with a circumscribe lump in her upper outer quadrant. Here is her DBT and her ultrasound image, extremely circumscribed, very hypoequic, almost cystic-appearing mass. What do you think this tumor was? Here is our poll question again. Is this a complicated cyst? Is this a fibroedinoma? Is it an IDC or is it metastasis? Excellent. So, most of you thought this was an IDC. Indeed, it was an at-biopsy. Sorry, I didn't show you this image that showed that this lesion was very vascular. It's not a cystic mass. And at-biopsy, this was a high-grade IDC and triple negative breast cancer. Now, the problem is that there is a variable agreement between these surrogate markers and the formal gene testing. And the agreement can be as low as 41%. And most genetic defects will not have surrogate markers identified. And therefore, there will be some imprecision to our diagnosis if we are to rely entirely on this surrogate markers, which will lead to underdiagnosis of lethal cancers overdiagnosis and over-treatment of indolent cancers. And therefore, gene expression profile assays will help us in resolving this issue. At the moment, they are available only for ER positive tumors. The commercial ones that are now available are the oncotide DX, the mammoprint, the PAM 50. All of them look at different number of genes by different methods. And what they're basically looking at is for ER positivity and the proliferation genes. What they tell you is about the risk of recurrence and the risk of metastases and the treatment decision that we can take from these tests is whether the patient needs adjuvant chemotherapy or radiation. So let's look at two such assays, the oncotide DX, which evaluates the tumor for 21 genes. And what it tells us is about the risk of ipsilateral breast cancer within 10 years for this lady. And if the score is low, that is below 11, then she has no risk of developing ipsilateral breast cancer. And the treatment choice that we can make is that we can avoid chemotherapy in this lady and she need only to be put on endocrine or anti-hormone therapy. In fact, chemotherapy will only cause her harm because it's not going to do any good for her tumor as such. On the other hand, Mammaprint here also analyzes about 70 genes. What it tells you is about the risk of metastases in five years. And again, if the score is low, we can make this choice of avoiding chemotherapy. Now, the impact from these biomarkers have been so significant in the patient treatment and prognosis that they are now been included in the eighth edition of the AJCC tumor staging, known as the prognostic staging. So if you have a T2N0M0 tumor, according to the seventh edition of the AJCC, which was an anatomic TNM staging, this tumor would have been classified as 2A. But with an oncotype DX score of 11, we can now downgrade this tumor to 1A. Now, the limitations of these gene assays is the high turnaround time. It takes about a couple of months for the results to be available. They are often only available at off-site or offshore labs. And the cost of this tests are also high, running into a few thousands. And therefore, there is a strong demand to develop other alternatives, like imaging biomarkers and surrogates. We have just seen how the biomedical images are a product of the processes that occur at the genetic and the molecular level. And this has led to this recent and extreme interest in these two fields of radiology, that is, radiomics and radiogenomics. What is radiomics? It has evolved because of the immense progress that we have made in data analytics, artificial intelligence, and machine learning. What we can now do is extract a large amount of qualitative and quantitative features from a large volume of digital images and amalgamate it with the patient clinical data. Now, what are these features that we extract? It could be the by-rads morphological features, various aspects of the tumor kinetics and things like texture analysis of the tumor. And this is done by various processes like histogram analysis, ROI identification, segmentation, et cetera. So when the information from radiomics is combined with the genetic data, that is what we call as radiogenomics. Now, radiogenomics is a system biology approach. So from the tumor sample, we can get information about the genes. There are also now assays available to get information about the mRNA transcripts, the proteins, and the metabolites. And then there is all this information that we can get from the images. Now, the advantages of imaging is that you're looking at the tumor as a whole. While in tumor sampling, we're only seeing a small portion of the tumor. So we only have a snapshot of the whole picture. And intratumoral heterogeneity is also known by which we mean part of the tumor may be hormone sensitive and part of it may be her to enrich. And your result will depend on from where you're sampling this tumor. So all this information is integrated by computational analysis and mathematics and the computer can build a model. It can generate computer algorithms. Hypotheses can be generated, which can be tested and validated, which leads to more precise diagnosis, prediction, and prognosis. All this information is available as an open source in centralized repositories like the cancer imaging archives and the cancer genomic atlas. So let's see some applications. Yamalmoto et al. as early as 2012, they studied 10 patients. They looked at 21 qualitative imaging features of MR and they could correlate with over 71% of the 52,000 genes that they studied. The same group also developed algorithm that could detect tumor enhancing heterogeneity that linked to immune-related genes that we see in interferon-rich triple negative breast cancer. Now, these are the cancers that would escape the normal line of chemotherapy. And so if we have this information available to us prior, we can start targeted therapy or specific therapy in these individuals earlier. And this has great significance for the patient prognosis. This is a more recent paper by Zoo et al. in 2020 where they could create computer models by wavelet texture analysis where they could differentiate and predict tumors that could achieve PCR and differentiate it from tumors that would not achieve PCR. Researchers have also been able to find imaging features linking them to molecular subtypes. For example, homogeneous tumor enhancement has 100% negative predictive value for luminal B cancers. Or if you look at the initial rapid-phase enhancement, her-to-enrich tumors will have a steeper slope of enhancement as compared to other tumors where all their researchers have also found imaging correlates that link with the oncotype DX and help in predicting risk of recurrence. So in conclusion, radiogenomics is a promising field with potential to capitalize on the rapid growth in data analytics. And the deep well-spring of breast cancer genetic knowledge. The limitations, however, is that the images that are used are vastly that we obtained from the MRI. So there is a selection bias. There is a positive of the genetic information, for example,