Data Detec'ves Enhancing the inves'ga've process with visualiza'ons
Hunter Whitney San Francisco
November 13, 2013
Think of a detec.ve trying to solve a case where the answer may lie hidden within stacks of evidence, .me is limited, and there are serious consequences for any errors.
From healthcare workers to security professionals, this same basic scenario applies to many people who work with data.
While there’s no single approach to finding answers in an array of data, UX design can help support various inves.ga.ve processes.
• PuCng the pieces together • Weighing the evidence • Evidence handling • Collabora.ng
Data as evidence
Pu8ng the Pieces Together
Users oFen have to work with evidence that is: • Fragmentary, mixed, and incomplete • Delivered in an asynchronous, non-‐sequen.al manner
Visual juxtaposi'on of different kinds of data
EEG
MRI MRI (DTI)
Electrical ac.vity data
Anatomical Features Fiber Tracts (Connec.vity)
Hunter Whitney and Jeff Chang, MD Please see hSp://vimeo.com/album/2489932
Visual juxtaposi'on of different kinds of data
EEG
MRI MRI (DTI)
Electrical ac.vity data
Anatomical Features Fiber Tracts (Connec.vity)
Hunter Whitney and Jeff Chang, ER Radiologist Please see hSp://vimeo.com/album/2489932
MRI + DTI
DTI
Visual transi'ons of different kinds of data
Hunter Whitney and Jeff Chang, ER Radiologist Please see hSp://vimeo.com/album/2489932
Weighing the evidence
A piece of evidence does not exist in isola'on. It should be considered in context (but the context can rapidly change): • The rela.ve meaning and significance of any piece of data
may change as other available informa.on comes in • A piece of evidence may be true but irrelevant or
poten.ally misleading if not well framed
THE CASE A previously healthy 12-‐year-‐old boy presents to an emergency department in southern Ohio in October with a chief complaint of fever and rash. The pa'ent has had subjec've fever for 3 days with weakness, malaise, anorexia, and decreased ac'vity. He had emesis followed by diarrhea on the second day. He complains of diffuse myalgias and migratory arthralgias, which began in the right shoulder and have progressed to involve the cervical spine and all joints of the extremi'es bilaterally, but sparing the hips. A rash developed bilaterally on his inner thighs at the onset of illness, then spread to involve his forearms, abdomen, trunk, and hands, sequen'ally. In the ED, his temperature is 37.3C, blood pressure is 136/70 mmHg, pulse is 104 beats per minute, and respiratory rate is 24 breaths per minute. The pa.ent's past medical history is significant only for aSen.on-‐deficit disorder (ADD) without hyperac.vity, and his medica.ons include acetaminophen, melatonin, and methylphenidate. He has no known drug allergies, and his immuniza.ons are current. Physical examina.on reveals erythematous lesions on the palate and pain with ac.ve and passive range of mo.on in all extremi.es. There is no joint swelling, erythema, or warmth. A red maculopapular rash with scaSered vesicular lesions is present on the thighs, forearms, trunk, and palms. A rapid group A streptococcal an.gen test on a throat swab is nega.ve. The pa.ent is Tanner stage 4 for genital development. He has no discharge or lesions. The pa.ent is prescribed ibuprofen and discharged from the ED with a diagnosis of hand, foot, and mouth disease. Worsening symptoms The pa.ent returns to the ED 2 days later with persistent fever, worsening arthralgia, and new swelling in the leF knee. He is unable to ambulate and has decreased oral intake. Both wrists and mul.ple finger joints have become painful. The pain is greatest in the right wrist and leF knee. Further ques.oning reveals that the pa.ent traveled to southern Kentucky about 4 weeks ago for a Labor Day camping trip with his family. He recalls having found .cks on his clothing but not on his skin. He denies alcohol or drug use. The family has numerous pets including 1 hermit crab, 1 African clawed frog, 2 lizards, 1 domes.c white rat, 1 python, and a tank of tropical fish. They had 2 dogs, but both recently died of unknown causes. The pa.ent denies any animal bites. He reports that he has a girlfriend, but he denies any sexual ac.vity. His temperature is 38.9C, blood pressure is 100/72 mmHg, pulse is 110 beats per minute, and respiratory rate is 30 breaths per minute. The rash has progressed, with an increased number of well-‐demarcated, macular, papular, and vesicular lesions involving the extremi.es, especially the palms and soles. The lesions are blanching and nonpruri.c. The leF knee is warm, nonerythematous, and swollen with ballotable fluid.
Further ques.oning reveals that the pa.ent traveled to southern Kentucky about 4 weeks ago for a Labor Day camping trip with his family. He recalls having found .cks on his clothing but not on his skin. He denies alcohol or drug use. He reports that he has a girlfriend, but denies any sexual ac.vity. The family has numerous pets including 1 hermit crab, 1 African clawed frog, 2 lizards, 1 domes.c white rat, 1 python, and a tank of tropical fish. They had 2 dogs, but both recently died of unknown causes. The pa.ent denies any animal bites.
Many clues, but which are relevant?
Accelera'ng the diagnos'c process with visualiza'ons
• Sugges.ng possibili.es • Process of elimina.on • Visual emphasis of promising direc.ons • Dynamic weigh.ng • Transparency of the process
Differen'al Diagnosis: the full range of poten.al causes, or combina.ons of causes, for a set of symptoms
Reaves, L; Hamilton, S; Vossmeyer, M; and Brady, R. Fever, rash and migratory polyarthralgia. Contemporary Pediatrics. Mar 2010;27(3):24-‐38.
Juvenile Idiopathic Arthritis
Postinfectious arthritis
Differen'al Diagnosis (Alternate layout 1)
Henoch-Schönlein purpura
Kawasaki disease
Infectious
Rheumatologic
Familial Mediterranean Fever
Genetics
Septic arthritis
Lyme disease
Bacterial
Enteroviruses
Parvovirus Viral
+5 -‐4
-‐6
History and Physical
Labs
Radiology
Fever, rash, joint pain
-‐3
+3
-‐3
more
Rock Mountain Spotted Fever
Hunter Whitney and Veena R Kumar, Pediatric Emergency physician
Diagnos'c Dilemma: when the most likely causes of a pa.ent’s symptoms have been ruled out and a defini.ve diagnosis has not been established
Differen'al Diagnosis (Alternate layout 1)
Infectious
Septic arthritis
Lyme disease Bacterial
Viral
+5 -‐4
History and Physical
Labs
Radiology
Fever, rash, joint pain, handles/sleeps with pet rat oFen, rat bit both dogs
-‐3
+3 Rock Mountain Spotted Fever
Meningococal meningitis
Gonorrhea
Secondary syphilis
Leptospirosis
Rat-bite fever
+2
CBC normal, Rapid strep neg., blood culture neg., joint fluid culture posi.ve for Strep. moniliformis
Chest X-‐ray nega.ve
Hunter Whitney and Veena R Kumar, Pediatric Emergency physician
Further ques.oning reveals that the pa.ent traveled to southern Kentucky about 4 weeks ago for a Labor Day camping trip with his family. He recalls having found .cks on his clothing but not on his skin. He denies alcohol or drug use. He reports that he has a girlfriend, but denies any sexual ac.vity. The family has numerous pets including 1 hermit crab, 1 African clawed frog, 2 lizards, 1 domes'c white rat, 1 python, and a tank of tropical fish. They had 2 dogs, but both recently died of unknown causes. The pa.ent denies any animal bites.
Many clues, but which were relevant?
Evidence Handling
• Visualiza.ons are just one part of inves.ga.on and analysis
• All the decisions star.ng from data collec.on, storage, organiza.on, sta.s.cal summariza.ons, modeling, etc. will influence the analysis and interpreta.on
When possible:
• Enable easy annota.ons and access to source material
• Provide mul.ple views/ways of looking at the data
No ma^er how good the visualiza'on, the outcome of the inves'ga'on depends on the integrity of every step of the process
From Data Insights: New Ways to Visualize and Make Sense of Data by Hunter Whitney (Morgan Kaufmann Publishers, November 2012)
Collabora'ng
• Willing to acknowledge the limits of your own exper.se and to ask for help when necessary
• Visualiza.ons can enhance: • Partnering with machines: detec.ves have their Watson • Collabora.ng with other people
• There’s a lot of experimenta.on and a growing awareness of poten.al uses
• Visualizing data is nothing new, but is now evolving rapidly
• Many different kinds of creators and users are coming to the table with varying exper.se, perspec.ves, and goals
• Approaches to data visualiza.ons are definitely not one size fits all. That’s why the UX design dimension is so important!
Problem Solving with Emerging Tools
Slide 1: hSp://i.huffpost.com/gen/1375606/thumbs/o-‐DETECTIVE-‐facebook.jpg Slide 2: Slide 3: hSp://www.theguardian.com/tv-‐and-‐radio/tvandradioblog+medical-‐drama Slide 4: http://images4.wikia.nocookie.net/__cb20091004222608/flashforward/images/1/1e/MI_wall.jpg Slide 5: http://www.cafepress.com/+forensic_evidence_large_poster,698750731 Slides 6 and 14: http://www.cs.rit.edu/~rlc/Courses/IS/Projects/20131/Proj2/Clue_rules.html Slide 26: http://www.gonemovies.com/WWW/Hoofd/A/PhotoLarge.php?Keuze=BrettJeremy, http://www.digitaltrends.com/computing/ibm-watson-supercomputer-goes-to-college/
Supplementary Image Credits:
Top Related