In this third part of ethical issues we are going to discuss about a recent topic. The idea of machines taking over health world and doctors being replaced by new medical technology is very controversial. Mostly as everything in this technological advanced world. In this page you will find not only artificial intelligence developing towards medicine but also legal issues, deep learning and final thoughts. 

IBM Watson 

You know what Google, Siri, Alexa are right? Most probably you have all used them at least once in your life, even if only for your own curiosity. Then, there you have an idea of what IBM Watson may be. IBM Watson is a supercomputer which has an optimal performance as a question answering machine. Watson has artificial intelligence AI combined with a advanced analytical software and it is based on cognitive computing. As in most cases, Watson got his name from his creator Thomas J. Watson.  [1] IBM Watson basically knows everything starting from chef, designers, movie director etc. But what, Watson is focused on lately and is health area. IBM has bought up medical image databases and is using deep learning to try to help doctors spot diseases more rapidly. [2] The idea is to turn Watson into a go-to digital doctor assistant. They are more interested for cancer since this is one of the most concerning and harmful diseases nowadays. Watson's ability to sift through troves of the latest medical data - including mountains of clinical trial data, medical journal entries, textbooks, and other literature - is a considerable asset for doctors who have to see as many patients as those in a community setting do. Watson can present cancer care teams with reports ranking the most effective therapies and treatment options. "He" also helps to make the best decision and the highest quality care for the patient.  According to IBM, Watson has been taught to suss out treatment plans for breast, lung, colorectal, cervical, ovarian, and gastric cancers, and will be trained on nine additional cancers this year. [3]


  How IBM Watson works in a nutshell [4]

An overview how Watson can help with cancer treatment [3] ©Fortune Magazine


But, should we trust IBM Watson? Should doctors take opinion from a machine and let them diagnose a patient? Should patients allow all their data being held on their database? 

There is no paper telling or showing (until now at least) that using Watson may be harmful for patients or even cause death for them from unregular data or not correct data. It is normal for a patient not to trust a machine to process their data and be afraid that maybe the machine is not secure and would not prescribe the right treatment for them. But, trust comes with experience and time. And mostly this is not about patients as it is for physicians. IBM Watson makes it easier for doctors to analyze information not only about the patients, but also helps them to be updated with the latest researches and papers written, latest new innovations and drug developments. As it is known, healthcare data is among the most complex and voluminous data produced in the world today. Lying among this huge pile of healthcare data are precious insights that can directly impact and improve the quality of human lives. [4] A doctor reads about a half dozen medical research papers in a month, Meyerson says, whereas Watson can read a half million in about 15 seconds. From that, machine learning (one of the key types of artificial intelligence today) can suggest diagnoses and the most promising course of treatment. [5] Using Watson, doesn't mean that the doctor will take everything that 'he" says or shows as granted. The machine gives only an idea, the decision remains to the physician. When it comes to patients, they should allow their medical information being saved in a huge huge database. All that data is used to help them and the doctors responsible for their treatment. This not only helps the patient during his treatment, but also during it's life betterment after being discharged from the hospital. [6] In 2016, a study made from Human experts at the University of North Carolina School of Medicine who tested Watson by having the AI analyze 1,000 cancer diagnoses, showed that:

  •  In 99 percent of the cases, Watson was able to recommend treatment plans that matched actual suggestions from oncologists. 
  • Because Watson can read and digest thousands of documents in minutes, "he" found treatment options human doctors missed in 30 percent of the cases. 
  • The AI’s processing power allowed it to take into account all of the research papers or clinical trials that the human oncologists might not have read at the time of diagnosis.  [7] 


The magic of IBM Watson data analytics (Example) 

The picture you see on your right shows the process of getting information from IBM Watson for a type of cancer and trying to find a treatment for the patient through the platform, but from another place in the world to another. Let’s take a hypothetical case of a patient in a far corner of Asia who is suffering from a rare form of lung cancer that is genetically linked. The doctors in the hospital where the patient is getting treated may not have the necessary expertise to treat this specific strain of lung cancer, but Watson for Oncology does with help from MSK Cancer Center data. The significance of this app is far-reaching as any doctor from anywhere in the world can access the app by just getting a license for the program and give their patients access to world-class cancer treatment.  [4]

 [4]


IBM Watson Health progress [13]


Artificial Intelligence (Other Opportunities)

IBM Watson is one of the most Big data platforms for healthcare out there. But, there are other big data platforms too which are also trying to help and be useful when it comes to analytics. Some of these machine learning Artificial Intelligence systems created from other companies, similar to IBM Watson are listed down below. 

  1. Microsoft Project Oxford
    Project Oxford is Microsoft’s venture into the world of artificial intelligence and deep learning. It takes in several key areas, including image, facial, text and speech recognition, and hopes to implement the technology into its computer operating systems and smartphone software. [8]
  2. Google DeepMind
    British tech company DeepMind Technologies looking to combine machine learning and the pursuit of neuroscience was bought out by Google in 2014. Renamed Google DeepMind and tasked with building the best general purpose learning algorithms in the industry. Unlike IBM’s Watson, which had a pre-determined purpose, DeepMind is more open-ended, and is also comparatively simpler to use than its competitors because of its deep learning capabilities. [8]
  3. Baidu Minwa
    Produced by Baidu, Google’s Chinese equivalent, Minwa is their landmark project, and mirrors the IBM Watson model, with over 72 processors and 144 graphics processors. Its image recognition capabilities are among some of the best in the world of artificial intelligence. Much like Watson, Minwa is an image recognition engine, and a very powerful one too, with a 36-server node set-up, 6.9 terabytes of host memory and a 0.9 petaflop peak performance. Not unlike Watson, Minwa’s Natural Language Processing capabilities are some of the most impressive in the world, but the whole project was shrouded in disrepute after the most recent Image Classification Challenge, in which Minwa posted a 4.58% error rate, better than its competitors from Google and Microsoft, and better than the average human rate of 5%. However, the company revealed it had fallen foul of the competition rules, therefore nullifying their scores. [8]
  4. OpenText Magellan
    The Canadian enterprise information management (EIM) company unveiled OpenText Magellan, its new flexible AI platform that combines open source machine learning with advanced analysis to acquire, merge, manage, and analyze big data, at its annual Enterprise World 2017 in Toronto. “However, OpenText is taking a decidedly different approach because our platform is not only built to handle massive amounts of structured and unstructured data, but it’s also open source Apache Spark-based so our customers can get the most current insights and updates.” says Adam Howatson, chief marketing officer of EIM. It has a wide range of applications across any industry, such as new drug approval processes in the pharmaceutical industry, optimizing supply chains in retail or industrial, or employee management in human resources. [9]
  5. Sap Leonardo
    The company has expanded the Sap innovation system, to include integrations in machine learning, Internet of Things (IoT), big data, and blockchain technologies, all on its SAP Cloud Platform. This digital innovation platform will offer customers either “an industry-focused approach called industry accelerators, aligned to specific use cases with fixed prices and timelines,” or a flexible approach that allows customers to build whatever they want and need for their industry or business. Sap has also collaborated with Google on March 2017 and includes additional certification of SAP technology and applications on Google Cloud Platform (GCP). [10]

Deep Learning

Nowadays hypothetically there exists a  huge collection of demographic and clinical information related to the diagnosis, treatment, and history of different  diseases and cancer in particular. Much of this data is drawn from electronic, text-based clinical reports that must be manually curated—a time-intensive process—before it can be used in research. Analyzing this data in an effective way can significantly improve the the reliability and the comprehensibility of diagnosis .

Applications of Deep Learning in Oncology

  • Cancer Detection From Gene Expression Data
    Due to the high dimensionality and complexity of these data this is not a trivial task. After decades of research there is still uncertainty in the clinical diagnosis of cancer and the identification of tumor-specific markers. A deep learning approach to cancer detection, and to the identification of genes critical for the diagnosis of breast cancer was presented by Danaee P, Ghaeini R, Hendrix DA[20]. They used Stacked Denoising Autoencoder (SDAE) to deeply extract functional features from high dimensional gene expression profiles. Next, evaluated the performance of the extracted representation through supervised classification models to verify the usefulness of the new features in cancer detection; and after that a set of highly interactive genes identified by analyzing the SDAE connectivity matrices. 
  • Cancer Classification

    A research paper published in Nature by Stanford University researchers has shown that their convolutional neural network (CNN) achieves performance on par with all tested experts when classifying skin cancer.[21]



  • Tumor Segmentation

    Accurate estimation of the relative volume of the subcomponents of a brain tumour is critical for monitoring progression, radiotherapy planning, outcome assessment and follow-up studies. For this, accurate delineation of the tumour is required. Manual segmentation poses significant challenges for human experts, both because of the variability of tumour appearance but also because of the need to consult multiple images from different MRI sequences in order to classify tissue type correctly. That is why an automatic segmentation of MRI images is a subject of interest for many research groups today. 

    In the paper [25] it is proposed an automatic segmentation method based on CNN, exploring small 3×3 kernels. The use of small kernels allows designing a deeper architecture, besides having a positive effect against overfitting, given the fewer number of weights in the network. 

  • Assisting Pathologists in Detecting Cancer

    Royal Philips and LabPON[23] announced the  plans to create a digital database of massive aggregated sets of annotated pathology images and big data utilizing Philips IntelliSite Pathology Solution. The database will provide pathologists with a wealth of clinical information for the development of image analytics algorithms for computational pathology and pathology education, while promoting research and discovery to develop new insights for disease assessment, including cancer. The challenge for executing deep learning techniques is having access to a database with sufficient high volume and high quality data from which to develop the algorithms. As one of the largest pathology laboratories in the Netherlands, LabPON will contribute its repository of approximately 300,000 whole slide images (WSI) they prospectively create each year to the database. This will contain de-identified datasets of annotated cases that are manually commented by the pathologist, and will comprise of a wide variety of tissue and disease types, as well as other pertinent diagnostic information to facilitate deep learning.[24]

     [24]

  • Prognosis Detection

    Prognosis provides an estimate of how serious or advanced the stage of cancer, and hence the chances of survival. Staging systems for cancer are critical for predicting the patient’s prognosis but suffer from limitations. Researchers in South Korea utilized deep learning to develop a prediction model for the prognosis of patients suffering from gastric cancer[22] and undergoing treatment (i.e. gastrectomy). They found that deep learning showed superior survival predictive powers compared to other prediction models. 

    They used a deep neural network model which consisted of 5 layers: input layer, 3 fully connected layer, and output layer with 8 characteristics (age, sex, histology, depth of tumor, number of metastatic and examined lymph node, presence of distant metastasis, and resection extent) of patients . Each layer functioned as the nonlinear weighted sum of lower layer. [22]


Legal Issues 

Many people (including us) would think about: Why would an AI machine have legal issues? It is only a machine. Well, apparently it has (and a lot of them!) The point is that there is also ethical issues so basically we should take AI really seriously if we haven't yet. The idea is who is responsible if a physician relies on a false prediction by a computer program? Who is responsible for any laws that are violated by the AI?  Manufacturers are expected to exercise reasonable care to ensure their products are not dangerous when used as intended. This means testing, analysis, and QA to ensure that common or even uncommon-but-forseeable situations don't create a hazard. [14]  The European Parliament has not any exact rules when it comes to AI machines. The European Parliament accordingly passed a resolution with recommendations to the European Commission on civil law rules on robotics (2015/2103(INL)) on 16 February 2017; the resolution was adopted with 396 votes in favour, 123 against and 85 abstentions.  In this context, the European Parliament is calling on the Commission to consider introducing a specific legal status for intelligent robots in the long term.  The European Parliament notes that the development and use of robotics give rise to a number of tensions and risks relating to human safety, privacy, integrity, dignity, autonomy and data ownership. A majority of MEPs believe that an ethical framework is required for the design and use of robots and AI machines. They would have to comply with the principles of beneficence (robots should act in the best interests of humans), non-maleficence (robots should not harm a human), autonomy (the capacity to make an informed, un-coerced decision about the terms of interaction with robots) and justice with regard to fair distribution of the benefits associated with robotics. Until then, the fault for any harm remains to the manufacturer and user (in our case physicians) of the AI machine. [15]


                                                                                                           [16]

                                                                                                                   [17]


Can Doctors be replaced? 

The right answer given from experts, including here IBM Watson inventors is NO. The idea of this machines is to help healthcare and not take doctors away from the medicine world. Even though humanism thinks different. Unconsciously when a new technology develops especially when it is related to healthcare one of the first thoughts that comes in people minds is about physician replacement. As a human instinct this frightens them. But, there is no point to be scared, other people say. Bernie Meyerson, IBM’s chief innovation officer says: "Watson can be helpful in a lot of industries, such as medicine, which are awash in data, but it can’t replace people. Human brains bring passion to the work, they bring common sense. "  Tim Estes, CEO of AI company  Digital Reasoning, explains: "Does AI really understand cancer? Not like a doctor does. But can it see signs of cancer based on how cancer is talked about? Absolutely.” [4]  Dr. Iain Hennessey, Clinical Director of Innovation at Alder Hey Children’s Hospital and Martijn G.H. Van Oijen, PhD is an Associate Professor at the Academic Medical Center – University of Amsterdam (Practitioners work with Watson everyday) both thought that AI cannot be a substitute for communication. Hennessey highlighted "When we just talk about disruptive innovations in healthcare, we tend to overhype the importance of technologies. When we see them in everyday practice, things become clearer. The fear that technologies such as Watson will replace physicians has no basis."  [11] Many other doctors think the same when it comes to this topic (especially those who work with AI machines like IBM Watson). They say that one of the most important reason that this replacement will not be possible is communication. Medicine world is an art and art has ultimate need for communication and understanding the patients feeling. Maybe the role of physicians will change but they will remain needed to give the last verdict when it comes to everything regarding a patient. Humans need Humans. They need advocacy. And there is currently no algorithm or smartphone app for empathy or understanding. [12] 


                                                                                                              [18]

                                                                                                           [19]

Bibliography:  

  1. http://whatis.techtarget.com/definition/IBM-Watson-supercomputer
  2. https://www.technologyreview.com/s/602744/ibms-watson-is-everywhere-but-what-is-it/
  3. http://fortune.com/2017/02/01/ibm-watson-cancer-florida-hospital/
  4. https://www.edureka.co/blog/hadoop-big-data-in-healthcare
  5. https://www.fastcompany.com/3065339/can-ibms-watson-do-it-all
  6. https://www.thememo.com/2016/11/29/ibm-watson-healthcare-app-ibm-watson-hospital-ibm-watson-cancer/
  7. https://futurism.com/ibms-watson-ai-recommends-same-treatment-as-doctors-in-99-of-cancer-cases/
  8. http://www.kdnuggets.com/2016/02/ai-supercomputers-microsoft-ibm-watson-google-deepmind-baidu.html
  9. http://www.itworldcanada.com/article/opentext-taking-on-ibm-watson-with-new-artificial-intelligence-platform/394763
  10. http://www.itworldcanada.com/article/sap-leonardo-reinvents-the-digital-innovation-system/393252
  11. http://medicalfuturist.com/what-is-using-ibm-watson-in-medicine-like/
  12. http://medicalfuturist.com/can-an-algorithm-diagnose-better-than-a-doctor/
  13. https://m.acc.com/chapters/del/upload/503-Cozen_Artificial_Intelligence-PPTX.pdf
  14. https://techcrunch.com/2017/01/28/artificial-intelligence-and-the-law/
  15. http://www.cms-lawnow.com/ealerts/2017/04/do-robots-have-rights-the-european-parliament-addresses-artificial-intelligence-and-robotics
  16. https://www.ravn.co.uk/artificial-intelligence-law-firms-run/
  17. https://techcrunch.com/2017/01/28/artificial-intelligence-and-the-law/
  18. http://wuwm.com/post/how-your-doctor-feels-about-you-could-affect-your-care#stream/0
  19. https://morebrainpoints.blogspot.de/2012/12/empathy-and-wellness.html
  20. Danaee P, Ghaeini R, Hendrix DA. A DEEP LEARNING APPROACH FOR CANCER DETECTION AND RELEVANT GENE IDENTIFICATION. Pacific Symposium on Biocomputing Pacific Symposium on Biocomputing. 2016;22:219-229.
  21. Andre Esteva et al. Dermatologist-level classification of skin cancer with deep neural networks, Nature 542, 115–118 , 2017
  22. Woo Jin Hyung, et al. Superior prognosis prediction performance of deep learning for gastric cancer compared to Yonsei prognosis prediction model using Cox regression, 2017 Gastrointestinal Cancers Symposium

  23. https://www.labpon.nl/

  24. http://www.philips.com/a-w/about/news/archive/standard/news/press/2017/20170306-philips-and-labpon-plan-to-create-worlds-largest-pathology-database-of-annotated-tissue-images-for-deep-learning.html

  25. Sérgio Pereira et al. Brain Tumor Segmentation Using Convolutional Neural Networks in MRI Images,  IEEE Transactions on Medical Imaging Volume: 35, Issue: 5, 1240 - 1251, May 2016 )