Key Information

Register
Submit
The challenge is finished.
Show Deadlines

Challenge Overview

INTRODUCTION
This is the second in a series of challenges to develop a desktop web application to enable artificial intelligence aided dental diagnosis and treatment planning.  We will be using IBM’s Watson to provide the computer learning aspect of the application.

USE CASE
Dental offices and individual practitioners who wish to increase diagnostic accuracy and create evidence based treatment plans.  Staff should be able to upload radiographs, pocket depth probing scores, and high definition photographs for Watson to review and aid the dental professional in diagnosis.  Once the diagnoses are established and confirmed, a treatment plan would be co-created with Watson.

REQUIREMENTS
For this challenge you will be making improvements to the existing codebase from the first challenge in the series as described below.  We will continue to use IBM Watson for analyzing images and providing recommended diagnosis.

Watson Training
-- Watson should be able to recognize the following diagnosis.
    -- Broken Molars
    -- Ceramic Crown
    -- Composite Filling
    -- Crown Molar Tooth
    -- Healthy Interproximal
    -- Healthy Molar
    -- Healthy Pre Molar
    -- Interproximal Decay
    -- Metal Filling
    -- Metal Filling With Decay
    -- Metal Fillings No Decay
    -- PFM Crown With Root Canal Pre Molar
    -- Root Canaled Teeth
    -- Root Canaled Teeth With Abscess
    -- 3 Unit Bridges
-- Watson has been trained on these already
    -- Class names are the above descriptions, lowercase with underscores
    -- Ex “root_canaled_teeth”, “healthy_molar”
-- Currently the system is retraining Watson after every diagnosis
    -- Eliminate this step, but retain functionality
-- Add functionality for retraining based on confirmed and corrected diagnosis
    -- This would take all images for the selected class and upload them as a zip to Watson for retraining
    -- This will be an admin feature, but can be anywhere for the MVP

Universal Screen Changes
-- Remove treatment plans for this iteration of the MVP
    -- Retain functionality as it will be added back in the future

-- Clean up overall display of all views

Diagnosis Creation Screen
-- Add fields for linking a diagnosis to a patient
    -- First name, last name, date of birth, patient ID
-- Currently all images uploaded with a diagnosis are saved and displayed separately
    -- A diagnosis should be a collection of tooth images
    -- The diagnosis should be associated with the patient and grouped by upload date

-- Clean up overall display

Diagnosis Confirmation Screen
-- Corrected Diagnosis filed should be a drop down with the values from Watson Training
    -- Value of the drop down should be the class name  (ex. “healthy_molar”)
-- User should need to click a button to confirm a diagnosis - e.g. it should not be automatic
-- Include the score returned from Watson (not currently displayed)

-- Clean up overall display
-- Display 1 row per image analyzed in a table
    -- Include a drop down to specify the tooth #
    -- Include a drop down to indicate if the image is an X-ray
    -- Include the ability to insert additional rows for teeth not included in the upload batch
        -- This is intended for tracking purposes
        -- Tooth should be automatically marked as not present
-- Include 4 additional rows for periodontal diagnosis (1 per mouth quadrant)
    -- See examples here: https://www.pinterest.com/marrimo/perio/
-- Include 1 additional row for oral cancer screening
    -- Drop down for Oral Pathology (multiple possible)
        -- Leukoplakia
        -- Erythroplakia
        -- Erythroleukoplakia
        -- Snuff Lesion
        -- Varix
        -- Aphthous Ulcer
        -- Herpes Lesion
        -- Fistula
        -- Irritation Fibroma
        -- Hyperkeratosis
        -- Mucocele
        -- Papiloma
        -- Pyogenic Granuloma
        -- Giant Cell Granuloma
        -- Geographic Tongue
        -- Lichens Planus
        -- Candidiasis
        -- ANUG
        -- ANUP
        -- Pericoronitis
        -- Linea Alba
        -- Verrucous Leukoplakia
    -- Drop down for Sites Description (multiple permissible)
        -- Left
        -- Right
        -- Medial
        -- Distal
        -- Anterior
        -- Posterior
        -- Middle
        -- Doral
        -- Lingual
        -- Sub-lingual
        -- Lateral
        -- Buccal
    -- Drop down for Tissue Type
        -- Mucosa
        -- Gingiva
        -- Tongue
        -- Floor of mouth
        -- Palate
    
-- Text input for Adjacent To (tooth numbers)

Diagnosis Detail Screen
-- Display all results as set from the diagnosis confirmation screen above
-- Add the ability to delete individual images/rows from the data
-- Add the ability to upload new images and add to the diagnosis
-- Include patient data as described on diagnosis creation
-- Clean up overall display

Main Screen
-- Change the display to show patient grouped diagnosis, as in the changes indicated for the diagnosis creation screen
    -- Display patient name, date of birth, patient ID, date of diagnosis
-- Selecting an item from the list will still navigate to the Diagnosis Detail

-- Clean up overall display

RESOURCES
Best Practices for custom classifiers in Watson Visual Recognition
Classifier Training Guidelines
PDF describing the goals of the system is located in the challenge forums

SETUP
-- See the challenge forums for details on adding yourself to the BlueDot Dental GitLab Group
-- Fork the repository and use the tc2 branch for the basis of development
-- Add harrywynn as a member of your forked repository with Reporter access

 

Final Submission Guidelines

-- Any sample data required to run your submission
-- Upload your source project and supporting files as a zip archive

Reliability Rating and Bonus

For challenges that have a reliability bonus, the bonus depends on the reliability rating at the moment of registration for that project. A participant with no previous projects is considered to have no reliability rating, and therefore gets no bonus. Reliability bonus does not apply to Digital Run winnings. Since reliability rating is based on the past 15 projects, it can only have 15 discrete values.
Read more.

REVIEW STYLE:

Final Review:

Community Review Board
?

Approval:

User Sign-Off
?

CHALLENGE LINKS:

Review Scorecard

?