Oncosifter: A Customized Approach to Cancer Information Authors : Ketan Mane & Sidharth Thakur Presenter: Yueyu Fu We are introducing a system called – “Oncosifter”, which acts as a search engine for cancer related information.

Overview of Oncosifter Modules of Interaction Conclusion Motivation Overview of Oncosifter Modules of Interaction Conclusion Here is an outline of the presentation on the system

Overview of Oncosifter Modules of Interaction Conclusion Motivation Overview of Oncosifter Modules of Interaction Conclusion Starting with “Motivation” …What motivated us to design the system ?

The Susan G. Komen Breast Cancer Foundation http://www.komen.org/bci/ Oncosifter: Motivation Volumes of medical data is available on lab research, latest news, treatment & diagnosis. In information age – data available in electronic format and accessible over internet. Various websites provide information only related to specific topics Eg: Oral Cancer, Breast Cancer Thus a need to make all topics available at one location. The Oral Cancer Foundation http://www.oralcancer.org We see that to find a cure for cancer, a constant effort is being put in by the research labs, hospitals and cancer centers. The research generates scores of data on latest research development, research, different treatment & diagnosis available and other information. All this information is available over the internet. However, there exists websites that are dedicated to just one specific cancer like – Oral cancer and Breast Cancer. Sometimes a user might not be familiar with all websites. Also may be too much of browsing needs to be done to find relevant information. Thus, this generates a need to have all information easily accessible at one location. The Susan G. Komen Breast Cancer Foundation http://www.komen.org/bci/

Oncosifter: Motivation Websites have data available at different levels/ locations Medical Jargon Usage Need to: - Make data easily available - Standardize data access procedures - Focus on non-medical community Sometimes the information is available at one location but is embedded deeper. Also we find that information is available at different sections/ areas of the website. In general, it sometimes becomes difficult to find relevant topic of user interest. A medical jargon is used , thus making it knowledgeable to only a medical community This all introduces a need to : Make data easily available Introduce fast means of access to the data Focus on non-medical community, who would be also interested in seeking information on a specific cancer All the above things motivated us in designing a system that will try to address these concerns.

Overview of Oncosifter Modules of Interaction Conclusion Motivation Overview of Oncosifter Modules of Interaction Conclusion Next is the general overview of the system

Oncosifter: Overview “Onco” means Cancer Oncosifter is developed to filter cancer related news and medical information Limited medical jargon is used – with focus on non-medical user groups Latest News and Diagnosis & treatment information is provided at same location Information access procedures are standardized Personalization of news through customization In Oncosifter – the word “Onco” means Cancer Oncosifter filters cancer news and medical diagnosis & treatment information on different cancers With focus on the non – medical users, who would also be using the system, use of medical language is done only wherever there is no substitute. To shorten the learning curve in adapting to the system, the access methods are standardized Information personalization feature for news is also introduced

Oncosifter: Overview System Design: System implemented in Perl-CGI Information Sources: Medlineplus - latest cancer news Cancer.gov - diagnosis and treatment information Consistent interaction styles for data access Different modules of interfaces developed Keyword Based Search Interface Directory Interface Hierarchical Visualization Interface Personalization Search Interface The whole system is developed using Perl-CGI Information is available through public websites – Medline and Cancer.gov Medline – is used to acquire information on latest cancer news and Cancer.gov – is used to get diagnosis & treatment information Interaction style – Mouse click is standardized for the system Different user-friendly interface modules are developed for information access, they are as follows: Keyword Based Search Interface Directory Interface Hierarchical Visualization Interface Personalization Search Interface

Different Modules of interaction in Oncosifter Oncosifter: Overview The diagram gives an overview of the different modules available in Oncosifter at a glance. Different Modules of interaction in Oncosifter

Overview of Oncosifter Modules of Interaction Conclusion Motivation Overview of Oncosifter Modules of Interaction Conclusion Next is the general overview of the system

Keyword-Based Search Interface Oncosifter: Modules of Interaction Keyword Based Search Query submission is through a text-box Query - Controlled vocabulary match is performed to retrieve information In Oncosifter: Query – keyword match performed to acquire URL-addendum Data is dynamically accessed Information filters applied to parse relevant data for display First we discuss about the Keyword based search Module A standard text box is used for query submission Usually the standard procedure is to perform Query-Controlled vocabulary match to fetch the information However, in Oncosifter – in the backend processing is done to acquire URL – addendums These are used to acquire information dynamically Information filters are applied using regular expressions to parse the required information Keyword-Based Search Interface

Oncosifter: Modules of Interaction The figure shows a Interface and the results page Figure: Keyword based search interface and corresponding results page

Directory Search Interface Oncosifter: Modules of Interaction Directory Search Provides overview of the different types of cancer Categorization of cancers – for simplified search Common terms used by non-medical community are used to identify cancers Multiple cancers in the same category name are concatenated in display Second Module is : the Directory Search Module It gives a general overview of the system Cancer are categorized to simply search Non-medical vocabulary is used For results, if the cancer belongs to the same category name then … information is concatenated in the same display page Directory Search Interface

Oncosifter: Modules of Interaction The figure shows a directory search interface and its display page. In this we see at the top that there are 2 types of Aids related cancer… the results of these cancer are available on the same result page. Figure: Directory search interface and corresponding results page

Hierarchical Visualization Interface: Body Locations/ Systems Oncosifter: Modules of Interaction Hierarchical Visualization Interface Graphical visualization reveal structure in data Cancer categories represent hierarchical tree data structure Hyperbolic tree used in to display cancer categories Category classification of cancer is easily available Minimum interaction needed to get information Third is the Hierarchical Visualization Interface: Here is attempt to use different approach of using visualization Hyperbolic tree is used owing to the hierarchical nature of the information It provides a clear layout of the cancer types and different cancers in each cancer type Hierarchical Visualization Interface: Body Locations/ Systems

Oncosifter: Modules of Interaction Here is a diagram, the small boxes identifies the clicked regions and associated change in the layout. The last one is the results page showing the information that the user wants Figure: Hierarchical search interface and corresponding results page

Keyword-Based Search Interface Oncosifter: Modules of Interaction Personalization Search Interface Feature to customize news retrieval topics Rating scale provided to set the users profile Highly rated cancer news are displayed at the top Users discretion to change the profile at any instance Article rating used to promote system learning and change users profile automatically The fourth and final interface shows – Sign in feature for existing Members and non-members The user would create a profile based on his interest The final results are shown according to the user-interest ratings The profile can be changed at any instance With the rating scale on the display page, automatic track of user interest is kept based on the feedback Keyword-Based Search Interface

Oncosifter: Modules of Interaction Diagram shows how the personalization feature can be accessed , followed by the rating scale to determine the user interest. Then the results page showing the related news on the chosen topics.. Finally the article display Figure: Personalization search interface and corresponding results page

Overview of Oncosifter Modules of Interaction Conclusion Motivation Overview of Oncosifter Modules of Interaction Conclusion Finally we conclude by stating the features that we successfully incorporated in Oncosifter

Oncosifter: Conclusion Oncosifter System: Latest news and diagnosis & treatment information is obtained Different user-friendly modules of interaction are available Only relevant information is displayed Multiple cancers of the same type are displayed together Consistency in layout Similar interaction styles Ability to build personal profiles and customize it based on feedback Reduces cognitive load on user Latest information is available User-friendly modules of interaction are available Required information is easily available Cognitive load on the user is reduced by maintaining consistency in interaction styles and layout Personal profile creation feature to filter related news only

Oncosifter Acknowledgements We would like to thank Dr. Javed Mostafa for providing valuable insight during the design process

Oncosifter: References 1. Furnas, G.W. Landauer, T.K, Gomez L.M and Susan Dumais, S.T. (1987), The vocabulary problem in human system communication. Commun. ACM, 30(11): 964 – 971 2. Gaines, B. R. and Shaw, M.L.G, (1989), Comparing the conceptual system of experts, In Eleventh International Conference on Artificial Intelligence, 633 – 638 3. Robertson, G. G., Card, S. K., Mackinlay, J. D., (1993). Information Visualization using 3D Interactive Animations, Commun. ACM, 36(4), 57-71 4. Foltz, P. W., Dumais, S. T., (1992), Personalized Information Delivery: An Analysis of Information Filtering Methods, Commun. ACM, 35(12), 51 – 60 5. Shneiderman, B. (1997). Human factors of interactive software. In Designing the User Interface: Strategies for Effective Human-Computer Interaction , Addison- Wesley, 1-37. 6. J. M. Mostafa, S. Mukhopadhyay, W. Lam and M. Palakal, (1997), A Multilevel Approach to Intelligent Information Filtering: Model, System and Evaluation, ACM Transaction of Information System, 15(4).

Oncosifter: References Thank You