Measuring Coding Accuracy Artificial Intelligence in Medicine National Cancer Institute.

Slides:



Advertisements
Similar presentations
Office Links - Sharing Data in Microsoft Office A Mixed Bag of Treasures Chester N. Barkan Registrar Long Island University, C.W.Post Campus.
Advertisements

Surgical Pathology of Wide Local Excision of Breast
Breast Pathology Helge Stalsberg MD University Hospital of North Norway.
COMPUTATIONAL INTELLIGENCE FOR THE DETECTION AND CLASSIFICATION OF MALIGNANT LESIONS IN SCREENING MAMMOGRAPHY DATA E. Panourgias,
THE DEVELOPMENT OF SYNOPTIC REPORTS FROM FREE TEXT CONTENT OF ARCHIVAL PATHOLOGY REPORTS GENERATED IN THE ANATOMIC PATHOLOGY LABORATORY INFORMATION SYSTEM.
Classifications and CASCOT Ritva Ellison Institute for Employment Research University of Warwick.
Breast Sanjaya Adikari Department of Anatomy.
MCQs On Breast Imaging:
Thorax Breasts.
Thorax Breasts.
Microscopically Yours: A Glimpse at our Cells, in Sickness and in Health Nina C. Zanetti Siena College Department of Biology.
Breast Pathology Dr. M. Griffin.
ADVANCED MICROSOFT POWERPOINT Lesson 6 – Creating Tables and Charts
Chapter 4 Essential Concepts in Molecular Pathology Companion site for Molecular Pathology Author: William B. Coleman and Gregory J. Tsongalis.
6- RENAL CARCINOMA. Renal cell carcinoma occupying the lower renal pole Cross-section of kidney shows a well circumscribed yellowish tumor mass occupying.
AJCC Staging Moments AJCC TNM Staging 7th Edition Breast Case #2 Contributors: Stephen B. Edge, MD Roswell Park Cancer Institute, Buffalo, New York David.
Morphology of breast cancer
Case Study 63: Cancer of the Female Breast
Clinico-Pathological Conference (CPC) Meet Karpagam Medical College Hospital
AJCC Staging Moments AJCC TNM Staging 7th Edition Breast Case #1 Contributors: Stephen B. Edge, MD Roswell Park Cancer Institute, Buffalo, New York David.
MAMMOGRAPHY - Pt 2 EQUIPMENT LECTURE & more….. RTEC 255 Week # 3 D. CHARMAN, M.Ed.,R.T.(R,M)
CANCER BREAST OVERVIEW Dr. Ehab M.Oraby. INTRODUCTION  Breast is a modified sweat gland between skin and pectoral fascia.
Using the Georgia Online Assessment System(OAS) We will lead the nation in improving student achievement. Kathy Cox, State Superintendent of Schools.
Ductal Carcinoma In Situ Shahla Masood, M.D. Professor of Pathology University of Florida College of Medicine - Jacksonville Chief of Pathology and Laboratory.
GUI development with Matlab: GUI Front Panel Components 1 GUI front panel components In this section, we will look at -GUI front panel components -Programming.
AJCC 6 TH EDITION STAGING OF BREAST CARCINOMA. AJCC NODE STAGING -16 CATEGORIES pNX – 1 option pN0 – 5 options; null,(i-),(i+),(mol-),(mol+) pN1 – 4.
Sector of a Circle Section  Sector – portion of the area of a circle area of sector = arc measure area of circle 360 Definitions & Formulas.
CREATING TEMPLATES CREATING CUSTOM CHARACTERS IMPORTING BATCH DATA SAVING DATA & TEMPLATES CREATING SERIES DATA PRINTING THE DATA.
Touchstone Automation’s DART ™ (Data Analysis and Reporting Tool)
RENAL TUMORS Renal BlockPathology Dept, KSU Renal Practical III.
Pathology.
© 2009 Bentley Systems, Incorporated Earthworks for InRoads V8i.
CHAPTER 7 LESSON C Creating Database Reports. Lesson C Objectives  Display image data in a report  Manually create queries and data links  Create summary.
W. Scott Campbell, Ph.D., MBA University of Nebraska Medical Center
From Qualitative to Quality Impact Heather Bryant, MD, PhD Health System Use Summit February, 2016.
George Cernile, Manager A.I. Technology Group Artificial Intelligence In Medicine Inc. Advancements in Automated Synoptic Reporting.
Exploring Taverna engine Aleksandra Pawlik materials by Katy Wolstencroft University of Manchester.
Electronic CAP Cancer Checklists and Cancer Registries – A Pilot Project 2009 NAACCR Conference Ken Gerlach, MPH, CTR Castine Verrill, MS, CTR CDC-National.
Advanced Taverna Aleksandra Pawlik University of Manchester materials by Katy Wolstencroft, Aleksandra Pawlik, Alan Williams
PATHOLOGY OF NECK DISSECTION. VIEW FROM DEEP ASPECT OF NECK DISSECTION.
Taverna allows you to automatically iterate through large data sets. This section introduces you to some of the more advanced configuration options for.
Diseases of the prostate Osvaldo Rubinstein, MD. Normal urinary bladder with right and left ureters.
CHAPTER 7 LESSON B Creating Database Reports. Lesson B Objectives  Describe the components of a report  Modify report components  Modify the format.
Plotting in Excel KY San Jose State University Engineering 10.
SNOMED CT and Surgical Pathology
Female Reproductive Anatomy Breasts
W. Scott Campbell, MBA, PhD James R. Campbell, MD
Invasive breast carcinoma
Requirements – Essential To Robust Product Design
SNOMED CT and Surgical Pathology
OptiSystem applications: BER analysis of BPSK with RS encoding
W. Scott Campbell, MBA, PhD University of Nebraska Medical Center
GO! with Microsoft Office 2016
Invasive breast carcinoma
NEOPLASIA (Malignant Tumors)
W. Scott Campbell, MBA, PhD University of Nebraska Medical Center
Thomas Kirchner, Thomas Brabletz  The American Journal of Pathology 
Dr. Sura Obay Al-Dewachi
MD Online IEP System Instructional Series – PD Activity #7
TRAINING OF FOCAL POINTS ON THE CountrySTAT/FENIX SYSTEM
Current Status of Breast Ultrasound
Data Upload & Management
Can I Get That as a PDF or Excel File
Avoiding Pitfalls in Mammographic Interpretation
Percutaneous Computed Tomography-Guided Coaxial Core Biopsy for Small Pulmonary Lesions with Ground-Glass Attenuation  Chia-Hung Lu, MD, FRCR, Cheng-Hsiang.
Digital Fundamentals Floyd Chapter 4 Tenth Edition
Microsoft Excel 2007 – Level 2
(histological continuity)
Presentation transcript:

Measuring Coding Accuracy Artificial Intelligence in Medicine National Cancer Institute

Project This project was funded in part by Contract Number 263-MQ from the National Cancer Institute Participating registries Kentucky Cancer Registry Los Angeles Cancer Registry Atlanta Cancer Registry New Jersey Cancer Registry

Objective Develop a software tool that measures the accuracy of an automated coding system against a reference data set. Sub-tasks Define a coding accuracy model. Create a software tool that accepts input from any automated coding system to produce accuracy data.

Automated coding CLINICAL HISTORY/MACROSCOPY Right mastectomy and axillary tissue. A right mastectomy specimen with overlying skin measuring 220mm x 85mm and underlying breast tissue measuring 220mm x 100mm x 70mm. The axillary tail measures 125 x 60mm. The nipple is slightly retracted and located centrally. The superior margin is painted red, the inferior margin painted green and the deep cut margin is painted blue. Cut sections of the underlying breast tissue shows an ill-defined grey white yellow lesion with patchy areas of haemorrhage measuring 35 x 35 x 35mm located immediately below the nipple, 20mm from the inferior margin, 45mm from the deep cut margin, 50mm from the superior margin, 85mm from the medial margin and 100mm from the lateral cut margin. A1 - nipple, B1 - upper outer quadrant, C1 - upper inner quadrant, D1 - lower outer quadrant, E1 - lower inner quadrant, F1, G1 - tumour composite blocks, H1, I1 - tumour composite blocks, J1 - deep cut margin, K1 - superior margin, L1 – inferior margin, M4 - lymph nodes, N4 - lymph nodes, O - 3 serial slices, lymph node, P - 3 lymph nodes. MICROSCOPY This right mastectomy specimen demonstrates an invasive ductal carcinoma with the following pathological features: TUMOUR HISTOLOGY & GRADE The tumour is of an infiltrating poorly differentiated ductal carcinoma of non-otherwise specified type. The tumour is poorly defined and extremely infiltrative, comprising poorly-formed tubules, nests or strands of cuboidal tumour cells displaying high grade nuclei. The tumour cells are set within fibrotic desmoplastic stroma. Many lactiferous ducts are entrapped within the tumour. Frequent tumour mitoses are seen. Microcalcification is seen in some neoplastic tubules. Tumour grade (Modified Bloom-Richardson Scoring System): Tubular formation: 3 Nuclear atypia: 3 Tumour mitoses: 2 Total score: 8 (Grade III) TUMOUR LOCATION, SIZE AND EXTENT The tumour is located 5mm below the nipple and has a macroscopic size of 35mm across. The border of the tumour is poorly circumscribed and infiltrative. INTRA-LYMPHOVASCULAR OR PERINEURAL TUMOUR PERMEATION Focal intralymphatic tumour permeation is noted. No perineural tumour invasion is seen in sections submitted. M M M C50.9 C77.9

What is measured? Sensitivity, specificity, reducibility and confidence Of a single code, either (topography) or (morphology) Of a pair of codes (topography, morphology)

Notation: Adenocarcinoma “M-81403” as the subject code NotationThe report was coded by the: Reference methodAutomatic method X : XAs X (M-81403) As X (M-81403) X : OAs X (M-81403) As other than X (M-82113) O : XAs other than X (M-82003) As X (M-81403) O : OAs other than X (M-82003) As other than X (M-82003) X : X+OAs X (M-81403) As X plus other codes (M-81403, M-80103) O : X+OAs other than X (M-82003) As X plus other codes (M-81403, M-82003) 6 Possibilities

Definitions NotationThe report was coded by the: Reference methodAutomatic method X : XAs X X : OAs XAs other than X O : XAs other than XAs X O : OAs other than X X : X+OAs XAs X plus other codes O : X+OAs other than XAs X plus other codes Note: The labeled areas include only the portions that are bounded by the arcs, not the entire circle. Q O:O B X:O A X:X R O:X C X:X+O S O:X+O Venn Diagram

Definitions NotationThe report was coded by the: Reference methodAutomatic method X : XAs XAs X only X : OAs XAs other than X O : XAs other than XAs X only O : OAs other than X X : X+OAs XAs X plus other codes O : X+OAs other than XAs X plus other codes Note: The labeled areas include only the portions that are bounded by the arcs, not the entire circle. Q O:O B X:O A X:X R O:X C X:X+O S O:X+O Venn Diagram

Definitions NotationThe report was coded by the: Reference methodAutomatic method X : XAs XAs X only X : OAs XAs other than X O : XAs other than XAs X only O : OAs other than X X : X+OAs XAs X plus other codes O : X+OAs other than XAs X plus other codes Venn Diagram

Definitions NotationThe report was coded by the: Reference methodAutomatic method X : XAs XAs X only X : OAs XAs other than X O : XAs other than XAs X only O : OAs other than X X : X+OAs XAs X plus other codes O : X+OAs other than XAs X plus other codes Note: The labeled areas include only the portions that are bounded by the arcs, not the entire circle. Q O:O B X:O A X:X R O:X C X:X+O S O:X+O Venn Diagram

Definitions NotationThe report was coded by the: Reference methodAutomatic method X : XAs XAs X only X : OAs XAs other than X O : XAs other than XAs X only O : OAs other than X X : X+OAs XAs X plus other codes O : X+OAs other than XAs X plus other codes Note: The labeled areas include only the portions that are bounded by the arcs, not the entire circle. Q O:O B X:O A X:X R O:X C X:X+O S O:X+O Venn Diagram

Definitions NotationThe report was coded by the: Reference methodAutomatic method X : XAs XAs X only X : OAs XAs other than X O : XAs other than XAs X only O : OAs other than X X : X+OAs XAs X plus other codes O : X+OAs other than XAs X plus other codes Note: The labeled areas include only the portions that are bounded by the arcs, not the entire circle. Q O:O B X:O A X:X R O:X C X:X+O S O:X+O Venn Diagram

Definitions NotationThe report was coded by the: Reference methodAutomatic method X : XAs XAs X only X : OAs XAs other than X O : XAs other than XAs X only O : OAs other than X X : X+OAs XAs X plus other codes O : X+OAs other than XAs X plus other codes Note: The labeled areas include only the portions that are bounded by the arcs, not the entire circle. Q O:O B X:O A X:X R O:X C X:X+O S O:X+O Venn Diagram

Definitions Note: The labeled areas include only the portions that are bounded by the arcs, not the entire circle. Q O:O B X:O A X:X R O:X C X:X+O S O:X+O Coding accuracy measures Confidence = A / (A+R) How much confidence can we place in the result. Reducibility = (A+R) / (A+R+C+S) How often is the subject code the only code identified when the code is identified. Specificity = Q / (Q+R+S) How often is the subject code not returned in those reports where it is not the reference code. Sensitivity = (A+C) / (A+B+C) How often is the reference code returned in those reports where it is the subject code.

Definitions Note: The labeled areas include only the portions that are bounded by the arcs, not the entire circle. Q O:O B X:O A X:X R O:X C X:X+O S O:X+O Coding accuracy measures Confidence = A / (A+R) How much confidence can we place in the result. Reducibility = (A+R) / (A+R+C+S) How often is the subject code the only code identified when the code is identified. Specificity = Q / (Q+R+S) How often is the subject code not returned in those reports where it is not the reference code. Sensitivity = (A+C) / (A+B+C) How often is the reference code returned in those reports where it is the subject code.

Definitions Note: The labeled areas include only the portions that are bounded by the arcs, not the entire circle. Q O:O B X:O A X:X R O:X C X:X+O S O:X+O Coding accuracy measures Confidence = A / (A+R) How much confidence can we place in the result. Reducibility = (A+R) / (A+R+C+S) How often is the subject code the only code identified when the code is identified. Specificity = Q / (Q+R+S) How often is the subject code not returned in those reports where it is not the reference code. Sensitivity = (A+C) / (A+B+C) How often is the reference code returned in those reports where it is the subject code.

Definitions Note: The labeled areas include only the portions that are bounded by the arcs, not the entire circle. Q O:O B X:O A X:X R O:X C X:X+O S O:X+O Coding accuracy measures Confidence = A / (A+R) How much confidence can we place in the result. Reducibility = (A+R) / (A+R+C+S) How often is the subject code the only code identified when the code is identified. Specificity = Q / (Q+R+S) How often is the subject code not returned in those reports where it is not the reference code. Sensitivity = (A+C) / (A+B+C) How often is the reference code returned in those reports where it is the subject code.

Data Flow Input Data Comparison, Matching and Analysis SQL Data Base With Query Assist Output Spread Sheets & Graphs SQL Database CODAC Source Data Coded Pathology Reports Reference Codes Automated Coding System Machine Generated Codes Accuracy Calculations Display Accuracy Data Input Data Comparison, Matching and Analysis SQL Data Base With Query Assist Output Spread Sheets & Graphs SQL Database CODAC Source Data Coded Pathology Reports Reference Codes Automated Coding System Machine Generated Codes Accuracy Calculations Display Accuracy Data

Software inputs

CODAC Front End

Software description Written in C#, uses latest.NET technology Runs on Standard Pentium workstation Imports and exports (CSV). Files can be edited with use text editor or Excel Optional Links to SQL database engine The performance of any automated coding system can be measured by using the specified data format

Software operation We ran pathology reports through the software. Software automatically calculates accuracy parameters by comparing reference data to test data.

Example of high confidence M (Adenocarcinoma) Sensitivity 0.82 Specificity 0.90 Reducibility 0.08 Confidence 0.87 Reference count 2647=15%

Example of high confidence M-81403,C61.9 (Adenocarcinoma, Prostate) Sensitivity 0.87 Specificity 0.99 Reducibility 0.06 Confidence 1.00 Reference count 1008=6%

Example of low confidence C44.9 (Skin) Sensitivity.57 Specificity.76 Reducibility.05 Confidence.03 Reference count 67 =.4%

Morphology Accuracy Plots

Code Pairs

An Experiment Modify AutoCode to produce output as follows: Take the largest morphology value Take the smallest topography value Example: Reduce M M C17.0 C16.9 C17.9 To M C16.9

Morphology – MinMax rule

Morphology – Before & After

Code Pairs – Before & After

Wrap Up Created a coding accuracy measurement system Applied to AIM’s AutoCode, but can be used to measure any coding system. Software available to public domain

Topography

Code Pairs – Min Max Rule