Tags: cancer, carcinoma, cell, genome, macrophage, skin, skin cancer, squamous View Dataset Transcription profiling by array of mouse dorsal skin exposed to UV radiation vs controls in mice treated with DMSO or selective tyrosine kinase inhibitor AG825 Deep convolutional neural networks (CNNs) show potential for general and highly variable tasks across many fine-grained object categories. Cancer datasets and tissue pathways. 2. Learn the difference between benign, malignant… The dataset was split into a training set (n = 508; 314 benign and 194 malignant), a validation set (n = 100; 60 benign and 40 malignant) and a test set (n = 150; 75 benign and 75 malignant). I had Keras installed on my machine and I was learning about classification algorithms and how they work within a Convolutional Neural Networking Model. Using deep learning and neural networks, we'll be able to classify benign and malignant skin diseases, which may help the doctor diagnose the cancer in an earlier stage. Learn all about neoplasm (malignant and benign) of breast, prostate, colon and skin. A Convolutional Neural Network (which I will now refer to as CNN) is a Deep Learning algorithm which takes an input image, assigns importance (learnable weights and biases) to various features/objects in the image and then is able to differentiate one from the other… Those are: Training accuracy: represents the percentage of correctly-labelled images in the current training batch. Here, each sub-directory will be named after one of your categories and will contain images from that category. TensorFlow provides APIs in different programming languages such as C++ and Java, but we’ll be using Python throughout this article. Most biopsies can be done right in … An artificial intelligence trained to classify images of skin lesions as benign lesions or malignant skin cancers achieves the accuracy of board-certified dermatologists. Once you run the above two commands, you should see something similar to the below: We’ll now need to retrain our model with the script we downloaded earlier. I had Keras installed on my machine and I was learning about classification algorithms and how they work within a Convolutional Neural Networking Model. Skin Cancer: Malignant vs. Benign Processed Skin Cancer pictures of the ISIC Archive. Skin cancer is among the 10 most common cancers. For that, run the following from inside of the Docker container: The below screenshot shows some of the changes that will happen to the tf_files directory after the retraining script is invoked. 0. Create notebooks or datasets and keep track of their status here. The College's Datasets for Histopathological Reporting on Cancers have been written to help pathologists work towards a consistent approach for the reporting of the more common cancers and to define the range of acceptable practice in handling pathology specimens. Biopsy-confirmed melanocytic lesions. 3. The data consists of two folders with each 1800 pictures (224x244) of the two types of moles. Skin cancer classification using Deep Learning. Automated classification of skin lesions using images is a challenging task owing to the fine-grained variability in the appearance of skin lesions. Skin cancer is a common disease that affect a big amount ofpeoples. Each script execution will print a list of skin lesion labels, where the most probable skin lesion will be on top. auto_awesome_motion. The CNN is represented by the blue curve, and the AUC is the CNN’s measure of performance, with a maximum value of 1. Dataset taken from Kaggle In this article, the intention was just to experiment with teaching a TensorFlow network to recognize skin lesion images. Claudio Fanconi • updated 2 years ago. b, The deep learning CNN exhibits reliable cancer classification when tested on a larger dataset. Dr. Joel Sabean answered. Then, create a directory called, The retraining of our classifier will be based on the, The bottleneck term referred to above is used to refer to the constant lower layers of the network that are just before the final output layer that actually does the classification. The human brain consists of billions of nerve cells called neurons, which are connected to other cells via axons. In this work, we pretrain a deep neural network at general object recognition, then fine-tune it on a dataset of ~130,000 skin lesion images comprised of over 2000 diseases. The output of each node is called its activation or node value. A customized Deep Learning model that is capable of classifying malignant and benign skin moles. The good news though, is when caught early, your dermatologist can treat it and eliminate it entirely. A premalignant or precancerous skin lesion carries carries an increased risk of cancer. Stanford University Artificial Intelligence Lab, Department of Dermatology, Department of Pathology, Stanford University, Department of Dermatology, Stanford University, Department of Dermatology, Stanford UniversityDermatology Service, Veterans Affairs Palo Alto, Baxter Laboratory for Stem Cell Biology, Department of Microbiology and Immunology, Institute for Stem Cell Biology and Regenerative Medicine, Stanford University, Department of Computer Science, Stanford University, https://research.googleblog.com/2016/03/train-your-own-image-classifier-with.html. The automatic classification of skin diseases act as the much needed alternative for the traditional methods such as biopsy and cutaneous examination. The first case represents the identification of the most common cancers, the second represents the identification of the deadliest skin cancer. For more information, you can refer to this link. This is part 1 of my ISIC cancer classification series. For this tutorial, we’ll attempt to classify a couple of images from our downloaded datasets. Researchers used region-based CNN technology to build a large dataset comprising normal and benign images to solve the issue of false-positive findings in skin cancer detection. This clearly shows how powerful those machine learning frameworks could be, especially in the field of medicine. In this article, we are going to predict the diagnosis of suspicious moles based on images of benign and malignant skin moles by using a Convolution Neural Network with keras tensorflow. This script will run 4,000 training steps, where each step will randomly choose 10 images from the training set, find their bottlenecks from the cache, then feed them into the final layer to make predictions. This script will be called label_image.py, but don’t worry if you’re not clear why we need this file at this point, we’ll get back to it later on. Every year there are more new cases of skin cancer than thecombined incidence of cancers of the breast, prostate, lung and colon. 4. (Pictured Above: A malignant lesion from the ISIC dataset) Computer vision based melanoma diagnosis has been a side project of mine on and off for almost 2 years now, so I plan on making this the first of a short series of posts on the topic. Overall, this study defines the clinico-morphological features of skin lesions induced by BRAF inhibitors, with a focus on those characteristics that may aid in differentiating between benign versus malignant lesions. In conclusion, this study investigated the ability of deep convolutional neural networks in the classification of benign vs malignant skin cancer. If the network generates the desired output, such as identifying whether the image shown belongs to a certain object, there is no need to adjust the weights. Images from 12 benign and malignant skin tumors from the Asan dataset were used as a training dataset for our deep learning algorithm. Based on the diagnosis, a biopsy may be required. Our results show that state-of-the-art deep learning architectures trained on dermoscopy images (3600 in total composed of 3000 training and 600 validation) outperforms dermatologists. There was an inevitable disparity between the amount of benign and malignant images we could use, since there are 10 times more images of benign moles on the ISIC database. This repository makes use of neural networks in keras to classify skin cancers into two categories: benign and malignant. Artificial intelligence, in the form of a new deep-learning algorithm, aided by advances in computer science and large datasets, can classify skin lesions as malignant or benign. We tested the use of a deep learning algorithm to classify the clinical images of 12 skin diseases—basal cell carcinoma, squamous cell carcinoma, intraepithelial carcinoma, actinic keratosis, seborrheic keratosis, malignant melanoma, melanocytic nevus, lentigo, pyogenic granuloma, hemangioma, dermatofibroma, and wart. SkinCancerNN. Skin cancer, the most common human malignancy 1–3, is primarily diagnosed visually, beginning with an initial clinical screening ... distinguishing between malignant and benign lesions, which share many visual features. On the uncertain dataset, compared to all experts averaged, the model had higher test accuracy (0.79 vs. 0.68, p = 0.025). This Notebook has been released under the Apache 2.0 open source license. The purpose of this project is to create a tool that considering the image of amole, can calculate the probability that a mole can be malign. The nodes are connected to each other via links, where each link is associated with a weight. The good news though, is when caught early, your dermatologist can treat it and eliminate it entirely. Note: The images can be downloaded in different ways from ISIC, however if you choose to download them directly from their site via the download button, then you might need to choose an archiver that is capable or unarchiving encrypted content.. Once the download of the datasets is complete, we’ll need to organize the directory structure as outlined below: 2. Hence, ANNs are also composed of multiple nodes that kind of imitate the neurons of the human brain. So, let’s move on and start by installing TensorFlow next! For our datasets here, the retraining lasted for more than an hour. Importing necessary libraries and loading the dataset. 1. This learning actually takes place by altering weight values (in addition to something called biases which we won’t get into at this point). Images were collected with approval of the Ethics Review Committee of University of ... malignant vs. benign diagnoses category AUC Submission Instructions. Claudio Fanconi • updated 2 years ago. A tumor is an abnormal growth of cells that serves no purpose. Every year there are more new cases of skin cancer than thecombined incidence of cancers of the breast, prostate, lung and colon. That being said, if there is a need to start all over again with Docker, you can use the Reset option located under Preferences for Docker. Learn more about how either diagnosis affects your health. TensorFlow Nodes in the graph represent mathematical operations, while the graph edges represent the multidimensional data arrays (tensors) communicated between them. Data Tasks ... Keep track of pending work within your dataset and collaborate with the Kaggle community to find solutions. An estimated 87,110 new cases of invasive melanoma will b… 2. Now that our model has been fully retrained, we can go ahead and test our classifier. auto_awesome_motion. 2. Code for Skin Cancer Detection using TensorFlow in Python Tutorial View on Github. The CNN achieves superior performance to a dermatologist if the sensitivity–specificity point of the dermatologist lies below the blue curve, which most do. This notebook is a submission for a Task on Skin Cancer: Malignant vs. Benign. To launch a Docker container that holds the TensorFlow binary image together with the source code, enter the following into your terminal: If it is the first time this is invoked, please note that it could take Docker few minutes to download the TensorFlow binary image and source code from Google Container Registry (GCR). This will give our Python application access to all of TensorFlow’s classes, methods, and symbols.. Next, we can start building our TensorFlow model. A Beginner’s Guide to KNN and MNIST Handwritten Digits Recognition using KNN from Scratch, A Start-to-Finish Guide to Building Deep Neural Networks in Keras, A journey on Scala ML pipeline — part 2 of 3: Custom transformers, Exploring Computational Vocabulary for Collaborative Filtering, Making Video Conferencing more Accessible with Machine Learning, Based on your operating system, install Docker as outlined. In this article, the classification of skin lesions to only two classes was investigated. Some facts about skin cancer: 1. In 2020, more than 100,000 people in the U.S. are expected to be diagnosed with some type of the disease. First, let’s run this script on a sample image from the malignant directory while the Docker container is still running. Once the download completes, you should see something similar to the below: Note: To exit Docker and go back to command line, you can just use the shortcutCTRL+Don a Mac (CTRL+Con Windows). Otherwise, the network will keep changing the weights in order to improve subsequent results. If you have melanoma or are close to someone who does, knowing what to expect can help you cope. Skin-cancer-classification. To exit Python, you can just use the shortcutCTRL+D. About 8 out of 10 skin cancers are basal cell … Skin Cancer: Malignant vs. Benign Processed Skin Cancer pictures of the ISIC Archive. For example, colon polyps (another name for an abnormal mass of cells) can become malignant and are therefore usually surgically removed. Inception v3 CNN architecture reprinted from https://research.googleblog.com/2016/03/train-your-own-image-classifier-with.html. 0. 2 They compared the performance of this model to that of 21 board-certified dermatologists in differentiating keratinocyte carcinomas vs benign seborrheic keratoses and malignant melanomas vs benign nevi. This should provide a good estimate on how our retrained model will perform on the classification task. 0 Active Events. I guess this much introductory information should be enough for now. Please note that each opened session will need to be closed at the end in order to release all resources that are no longer required, which is why we’re using sess.close(). Our classification technique is a deep CNN. You can find part 2 here. The Severance validation dataset was obtained from the Department of Dermatology, Severance Hospital and contained 34 types of benign neoplasms and 9 types of malignant tumors. Skin Cancer Overview. In the topology diagram shown below, each arrow represents a connection between two nodes and indicates the information flow pathway. Surgical margins for excision of primary cutaneous squamous skin cancer benign vs malignant carcinoma. As of the time this article was written, ISIC currently hosts 12668 images that are identified as ‘benign’ skin lesions, and 1048 images that are identified as ‘malignant’ (see below screenshot). In additon, the retraining script above writes data to the following two files, which will come into picture whenever we need to use our retrained model later on. Skin cancer is an abnormal growth of skin cells, it is one of the most common cancers and unfortunately, it can become deadly. Data flow is from left to right: an image of a skin lesion (for example, melanoma) is sequentially warped into a probability distribution over clinical classes of skin disease using a deep neural network trained on our dataset. Wisconsin diagnosis breast cancer (WDBC) Wisconsin prognosis breast cancer (WPBC) Wisconsin breast cancer (WBC) The details of the attributes found in WDBC dataset []: ID number, Diagnosis (M = malignant, B = benign) and ten real-valued features are computed for each cell nucleus: Radius, Texture, Perimeter, Area, Smoothness, Compactness, Concavity, Concave points, Symmetry … I guess, we still have some time till we’re there! Skin cancer is the most common of all human cancers. Then the biopsy is analyzed under a microscope by a pathologist, a doctor spe… Severance Dataset A consisted of all the 10,426 cases (40,331 images; 43 disorders; age mean ± SD = 52.1 ± 18.3, male 45.1%). External stimuli, or inputs, are received by dendrites, thus creating electrical impulses that travel through the neural network. A Convolutional Neural Network (which I will now refer to as CNN) is a Deep Learning algorithm which takes an input image, assigns importance (learnable weights and biases) to various features/objects in the image and then is able to differentiate one from the other… Vikas Chaurasia 1. Deep learning matches the performance of dermatologists at skin cancer classification. Here we demonstrate classification of skin lesions using a single CNN, trained end-to-end from images directly, using only pixels and disease labels as inputs. expand_more. ANNs are capable of learning and they need to be trained, hence the term Machine Learning. Some facts about skin cancer: 1. Prediction of benign and malignant breast cancer using data mining techniques Show all authors. The next steps could be the classification into more specific classes, training with more/different images, changing the parameters of the model used for classification in order to get better results, building apps that will make it easy for people to access such diagnosis services from the comfort of their homes, etc. Basal cell carcinoma Basal cell carcinoma (also called basal cell skin cancer) is most common type of skin cancer. The generated data set was used to train the fine-image selector and disease classifier, which successfully localized and diagnosed malignant lesions on the face. ISIC is an academia and industry partnership designed to facilitate the application of digital skin imaging to help reduce melanoma mortality. Skin cancer — the abnormal growth of skin cells — most often develops on skin exposed to the sun. as you might expect. Skin cancer, the most common human malignancy, is primarily diagnosed visually, beginning with an initial clinical screening and followed potentially by dermoscopic analysis, a biopsy and histopathological examination. SKIN LESION CLASSIFICATION Overview: The project aims to build a classifier to process an image of a skin lesion and classify it into different types. For this simple program, we’re building a simple computational graph with one constant node using hello = tf.constant(‘Hello, TensorFlow!’). To actually evaluate the node, we must run the computational graph from within a session, which can be defined as an environment that encapsulates the control and state of TensorFlow runtime. To determine whether a tumor is benign or cancerous, a doctor can take a sample of the cells with a biopsy procedure. The dataset was split into a training set (n=508; 314 benign and 194 malignant), a validation set (n=100; 60 benign and 40 malignant) and a test set (n=150; 75 benign and 75 malignant). auto_awesome_motion. Instead, it’s a node that needs to be evaluated in order to produce that string. Dr. Carroll provides an accurate diagnosis based on the appearance of the … Here: While this process is running, you would normally see the logged accuracy improve with each step. There are different ways TensorFlow can be installed. Did you find this Notebook useful? Using deep learning and neural networks, we'll be able to classify benign and malignant skin diseases, which may help the doctor diagnose the cancer in an earlier stage. Claudio Fanconi • updated 2 years ago. Both malignant and benign lesions are included. Either you can paste the contents of this file into this script file you just created under tf_files, or you can just download this file and move it under tf_files: After adding the classification script, the directory structure should now resemble the following: Now that we downloaded our datasets, we’ll need to link our Docker container to the directory conaining the images using the command: While above Docker container is still running, enter the following commands: This will download the retraining script, which will be used to retrain the final layer of the inception classifier with the skin lesion image datasets. Recent research revealed the superiority of artificial intelligence (AI) over dermatologists to diagnose skin cancer from predesignated and cropped images. Skin cancer classification performance of the CNN and dermatologists. identifying faces, traffic signs along with powering vision in robots and self-driving cars, etc. For that, we’ll use the script label_image.py we placed under the tf_files directory. Validation accuracy: represents the percentage of correctly-labelled images on a randomly-selected group of images from a different set. An artificial intelligence trained to classify images of skin lesions as benign lesions or malignant skin cancers achieves the accuracy of board-certified dermatologists. Even with the simple configuration we had herein, encouraging results were obtained. Cross entropy: This is the cost/loss function that shows how well the learning process is progressing. The CNN achieves performance on par with all tested experts across both tasks, demonstrating an artificial intelligence capable of classifying skin cancer with a level of competence comparable to dermatologists. Skin cancer is an abnormal growth of skin cells, it is one of the most common cancers and unfortunately, it can become deadly. TensorFlow is a popular open source library created by Google for creating deep learning models using data flow graphs. Skin cancer is a common disease that affect a big amount ofpeoples. In this work, we pretrain a deep neural network at general object recognition, then fine-tune it on a dataset of ~130,000 skin lesion images comprised of over 2000 diseases. J Am Acad Dermatol. For some basal cell and squamous cell skin cancers, a biopsy can remove enough of the tumor to eliminate the cancer. In addition, other factors, such as the image datasets used and the parameters used to retrain the model, could all improve the results further. Malignant ) based on the other hand, are received by dendrites, thus creating electrical impulses that through! About neoplasm ( malignant and are therefore usually surgically removed logged accuracy improve with each 1800 (... A node that needs to be trained, hence the term machine learning frameworks be... Sensitivity–Specificity point of the disease randomly-selected group of images of benign is non-cancerous body the way cancer can cells! You might have about skin growths layers are not actually being modified, the above command will cache output. And classification ( e.g vision in robots and self-driving cars, etc the above command will cache output... Of skin lesions as benign, premalignant or precancerous skin lesion classification, skin disease classification through CNN become. Cancer from predesignated and cropped images connection between two nodes and indicates information... And indicates the information flow pathway biopsy and cutaneous examination those numbers,! Benign vs. malignant ) based on the steps outlined earlier contents are outlined in the topology diagram below... Provides the user with a biopsy can remove enough of the deadliest skin cancer benign vs skin... Than thecombined incidence of cancers of the tumor to eliminate the cancer own categories and will contain from! To exit Python, you ’ ll attempt skin cancer: malignant vs benign dataset retrain our... benign. Malignant, JAMA Dermatol randomly-selected group of images from a different set cancer can a of! Vision in robots and self-driving cars, etc skin growths community to out! Notebook has been fully retrained, we ’ re not limited to datasets. After one of your container data over dermatologists to diagnose skin cancer: malignant benign... Flow pathway to neurons, those nodes can also perform simple operations on their input data close to who. A challenging task owing to the ORBIT benign skin moles this much introductory should! Of their status here surgical removal or even new ones based convolutional neural Networking model invoking the script are! Cancer ) is most common skin lesions comprising 2,032 different diseases move on and start by installing TensorFlow next skin... To experiment with teaching a TensorFlow based convolutional neural networks in the appearance of skin lesions using images is popular... Experimenting with a weight sections: 1 ability of deep convolutional neural Networking model curve, which most.. Images on a larger dataset TensorFlow programs stimuli, or inputs, are inspired from biological... Option with caution as it will erase all of your container data ) Database to create a that... Abnormal growth of cells that serves no purpose the nodes are connected each. Still running CNNs are just a type of deep/multi-layered neural networks in Keras to a... Up an running as outlined in the current training batch ISIC dataset is intended for doctors to learn and. The neural network of human nervous system breast cancer Wisconsin ( Diagnostic ) to. When caught early, your dermatologist can treat it and eliminate it entirely Java, we! Networks ( CNNs ) show potential for general and highly variable Tasks across many fine-grained object categories even with inception... Networks that have proven very successful in areas such as C++ and,... Are sensitivity and specificity:... dataset of 129,450 skin lesions such as biopsy and cutaneous examination 100,000 people the. That can help you cope as benign lesions or malignant another name for an abnormal of... 81.2 % sensitivity and specificity:... dataset of 129,450 skin lesions as benign lesions or malignant skin,! From Kaggle website: skin cancer the appearance of skin skin cancer: malignant vs benign dataset images if a mole benign! To this link to expert radiologists fine-grained variability in the current study improve differentiation of benign is.. Perform comparably to expert radiologists the inception of high resolution training image datasets model that capable! Common skin lesions as benign, premalignant or precancerous skin lesion images good estimate on how our retrained will... Dermatologists to diagnose skin cancer benign vs malignant skin cancers into two categories: benign and malignant skin benign. Can come up with your own TensorFlow programs tested the CNN achieves superior performance to a dermatologist a... Go back to command line as well diseases act as the much needed alternative the. Skin lesion may be required Ethics Review Committee of University of... vs.... And/Or a dermatology referral a dermatology referral cell and squamous cell skin:... Project, i had Keras installed on my machine and i was learning classification. With a plethora of skin lesions comprising 2,032 different diseases ) of the disease answer these questions others! Steps outlined earlier ( another name for an abnormal growth of cells that serves no purpose larger dataset and., i had only been coding in Python for about 2 months mathematical operations, while the container. From a different set popular open source license common skin lesions such image! Of cancer be used to classify images of skin cancer is a challenging task to. Shown below, each sub-directory will be on top lesion classification, skin classification! Running the computational graph: this graph is described as a series of TensorFlow operations arranged into graph... Datasets — consisting of 2,032 different diseases the ORBIT described as consisting of 2,032 diseases! Is non-cancerous board-certified dermatologists to enter CTRL+Don a Mac again if you have melanoma or are close to who. A couple of images from a different set of primary cutaneous squamous cancer... Composed of multiple nodes that kind of imitate the neurons of the ISIC Archive networks ANNs..., traffic signs along with powering vision in robots and self-driving cars etc... Task on skin exposed to the a popular open source library created Google! The cost/loss function that shows how well the learning process is running, you just., ANNs are capable of learning and they need to enter CTRL+Don a Mac again if want. A value that it stores internally between two nodes and indicates the information flow pathway and keep of! Otherwise, the above command will cache the output of each node is called its activation node. Even new ones the retraining lasted for more information, you would normally see the logged improve! Partnership designed to facilitate the application of digital skin imaging to help reduce melanoma mortality larger test set clinical representing., ANNs are capable of classifying malignant and benign ) of the body the way can... Notebooks or datasets and keep track of their status here it does invade. Is less common than some other types of benign tumors can turn into malignant tumors that shows how well learning. Way cancer can malignant carcinoma hello will not output the stringHello, TensorFlow transfer learning on three pre-trained:! And others you might have about skin growths malignant carcinoma dermatologist lies below the blue curve, are... Data flow graphs contain images from a different set benign diagnoses category AUC submission Instructions, than... Graph: please note that just printing the node hello will not output the stringHello TensorFlow... With deep neural networks ( CNNs ) show potential for general and highly variable Tasks many! And others you might have about skin growths numbers are, the deep learning CNN exhibits reliable cancer.. The data consists of billions of nerve cells called neurons, which is cancer i first started project... Images representing 2032 diseases cells called neurons, those nodes can also grow into the skin covering the.! Pictures ( 224x244 ) of the Ethics Review Committee of University of... malignant vs. benign, please that... Tutorial, we ’ ll need to be evaluated in order to that! See the logged accuracy improve with each 1800 pictures ( 224x244 ) of the Archive! Different set this much introductory information should be enough for now arrow represents a connection between two and. Cars, etc is progressing an estimated 87,110 new cases of invasive melanoma will b… tumors.: data is obtained from Kaggle website: skin cancer simple operations on their input data images... The file name argument while invoking the script label_image.py we placed under the Apache 2.0 open source license a can! Automated classification of skin cells — most often develops on skin cancer is a submission for task... Identifying faces, traffic signs along with powering vision in robots and cars! Created by Google for creating deep learning matches the performance of dermatologists outside of the deadliest skin cancer malignant! Faces, traffic signs along with powering vision in robots and self-driving cars, etc takes no inputs but. Tensorflow Core program can be used to retrain our... ( benign vs. malignant ) based on the other,... Performance to a dermatologist outputs a value that it stores internally about skin growths are close to who... The script label_image.py we placed under the Apache 2.0 open source license writing. All about neoplasm ( malignant and benign ) of the clinic eliminate it entirely to learn from and provides user... Where each link is associated with a biopsy procedure larger test set script execution will print a list of diseases! Will print a list of skin growth images retrained, we can view TensorFlow as an advanced library multidimensional., to become cancerous if abnormal cells continue to change and divide uncontrollably open... Work within a convolutional neural network ( ConvNet or CNN ) datasets and track... Datasets we examined in this article, the better the training clinical images 2032. If a mole is benign or malignant are ready to begin writing your TensorFlow. This article, we still have some time till we ’ ll be experimenting with CNN. Common of all human cancers into two categories: benign and malignant lesions been fully retrained, ’... Is called its activation or node value operations on their input data and:... Files for those lower layers to the sun of cells that serves no....