COMPUT COMPUT E E R AIDE R AIDE D DE D DE T T E E CT CT - PowerPoint PPT Presentation

COMPUT COMPUT E E R AIDE R AIDE D DE D DE T T E E CT CT ION OF ION OF PUL PUL MONARY MONARY T UBE RCUL OSIS ON DIGIT AL CHE ST RADIOGRAPHS: A SYST E MAT IC RE VIE W T RI PT I PANDE ADVANCE ADVANCE D T D T B DI B DI AGNOST AGNOST I I CS COURSE CS COURSE MCGI L L SUMME R I NST I T UT E OF I NF E CT I OUS DI SE ASE S AND GL OBAL HE AL T H 23 RD , 2016 JUNE

OUT L I NE  Che st ra dio g ra phy vs Co mpute r a ide d-  Che st ra dio g ra phy vs Co mpute r a ide d- de te c tio n  Study o b je c tive s Study o b je c tive s  Me tho ds  Re sults  Ana lysis  F ina l re ma rks  F uture wo rk

CHE ST RADI OGRAPHY PROS CONS Hig hly se nsitive Hig h c o st e q uipme nt Mo de ra te ly spe c ific I nte r- pro fe ssio na l va ria b ility Use ful sc re e ning to o l I ntra - pro fe ssio na l va ria b ility

COMPUT E R AI DE D DE T E CT I ON  De finitio n:  De finitio n:  Auto ma te d syste m whic h q ua ntifie d va rio us ima g e c ha ra c te ristic s o f c he st ra dio g ra phs ra dio g ra phs  F unc tio ns o f CAD:  Using DXR a nd so ftwa re pro g ra ms to de te c t g p g ra dio g ra phic a b no rma litie s c o mpa tib le with pulmo na ry T B  Ca n e limina te pro b le ms o f de la ys in i t inte rpre ta tio n a nd po o r inte r-re a de r va ria b ility t ti d i t d i b ilit

PRI MARY AND SE CONDARY OBE CT I VE S  Prima ry obje c tive : a sse ss the a c c ura c y fo r de te c tio n o f  Prima ry obje c tive : a sse ss the a c c ura c y fo r de te c tio n o f mic ro b io lo g ic a lly c o nfirme d pulmo na ry T B  Se c onda ry obje c tive : c o mpa ring CAD dia g no sis to tha t o f c linic a l o ffic e rs dia g no sing pulmo na ry T ffi di i l T B i h B via c he st ra dio g ra phs t di h

I NCL USI ON AND E XCL USI ON CRI T E RI A I nc lusio n E xc lusio n • E stima ting dia g no stic • CAD use d fo r dia g no stic a c c ura c y o f CAD so ftwa re mo da litie s o the r tha n CXR • De te c ting pulmo na ry • Did no t use mic ro b io lo g ic a l tub e rc ulo sis (PT B) re fe re nc e sta nda rd • Co mpa ring to a • Studie s invo lving c hildre n mic ro b io lo g ic re fe re nc e sta nda rd

SE ARCH ST RAT E GY Se a rc h limita tio ns: Pub me d  No la ng ua g e re stric tio n  Ja nua ry 1st, 2010 – De c e mb e r 31 st , y 2015 Sc o pus E mb a se Se a rc h te rms:  Va lida te d b y me dic a l lib ra ria n Va lida te d b y me dic a l lib ra ria n  Bro a d se a rc h te rms use d E ng ine e ring villa g e villa g e

Pa nde , Co he n e t a l. 2016 (a c c e pte d) S T SUL RE

DAT A E XT RACT I ON  Ye a r o f pub lic a tio n, g e o g ra phic  Mic ro b io lo g ic re fe re nc e Y f b li ti hi Mi b i l i f lo c a tio n  Dia g no stic a c c ura c y; se nsitivity,  Study de sig n (re tro spe c tive , spe c ific ity, AUC pro spe c tive ) ti )  Studie s c o mpa ring to huma ns:  I nc lusio n/ e xc lusio n c rite ria numb e r o f huma n re a de rs a nd the ir e xpe rie nc e  Me tho d o f g e ne ra ting dig ita l Me tho d o f g e ne ra ting dig ita l  Qua lity a ssura nc e : QUADAS 2 ra dio g ra phs  Na me , ve rsio n a nd me tho d fo r sc o ring CAD sc o ring CAD

GE NE RAL RE SUL T S Study Country Study de sig n Mic robiolog ic re fe re nc e sta nda rd Ma duska re t a l. Za mb ia Re tro spe c tive Sme a r o r c ulture Muyo ye ta e t a l. Za mb ia Pro spe c tive Xpe rt MT B/ RI F Bre uning e r e t a l. T a nza nia Re tro spe c tive Po sitive c ulture re sult Ho g e we g e t a l. E ng la nd Re tro spe c tive Sme a r o r c ulture So uth Afric a Philipse n e t a l. So uth Afric a Pro spe c tive Culture a nd c linic a l c rite ria Pa nde , Co he n e t a l. 2016 (a c c e pte d)

Pa nde , Co he n e t a l. 2016 (a c c e pte d) ANALYSIS

ANALYSIS - CAD VS HUMAN INTERPRETATION Olde r ve rsio n (v 1 07) Olde r ve rsio n (v 1 07) Olde r ve rsio n (v. 1.07) Olde r ve rsio n (v. 1.07) pe rfo rme d simila r to pe rfo rme d simila r to c linic a l o ffic e rs c linic a l o ffic e rs Ne we r ve rsio n (v. 3.08) Ne we r ve rsio n (v 3 08) Ne we r ve rsio n (v 3 08) Ne we r ve rsio n (v. 3.08) pe rfo rme d simila r to pe rfo rme d simila r to e xpe rt ra dio lo g ists e xpe rt ra dio lo g ists • CAD se t to a c hie ve sa me spe c ific ity a s no n-e xpe rt c linic ia n  CAD a c hie ve d sa me se nsitivity a s o the r no n-e xpe rts • hre sho ld se t to a c hie ve sa me se nsitivity a s huma n re a de rs  CAD a c hie ve d T lo we r spe c ific ity tha n ra dio lo g ist BUT g re a te r tha n c linic a l o ffic e r

ANALYSIS- QUADAS-2 RISK OF BIAS RISK OF BIAS APPLICABILITY CONCERNS APPLICABILITY CONCERNS Patient Reference Flow and Patient Reference STUDY Index test Index test selection standard timing selection standard AUTHORS Maduskar et al. 8 U L L U H H L Breuninger et al. 10 H L H H H L H Muyoyeta et al. 9 L L L L U U L Hogeweg et al. 6 H L U U U U L Philipsen et al. 7 L L L L L L L *Legend: L = low, H= high, U= unclear Pa nde , Co he n e t a l. 2016 (a c c e pte d)

F I NAL RE MARK S  I mpo rta nt po int to re me mb e r: e vide nc e b a se to suppo rt the use o f CAD i CAD is q uite limite d (n=5) it li it d ( 5)  Hig h se nsitivity thre sho ld  sub sta ntia l fa ll in spe c ific ity (vic e ve rsa )  CAD so ftwa re a c hie ving hig h se nsitivity a nd spe c ific ity wo uld b e  CAD so ftwa re a c hie ving hig h se nsitivity a nd spe c ific ity wo uld b e ma jo r le a p fo r this te c hno lo g y  Our syste ma tic re vie w ma inly hig hlig hts;  Unc e rta inty in the na sc e nt e vide nc e b a se  He lp g uide the dire c tio n fo r future e va lua tive studie s fo r CAD4T B  Ac c e pte d fo r pub lic a tio n in I  Ac c e pte d fo r pub lic a tio n in I JT JT L L D (Ma y 2016) D (Ma y 2016)

F UT URE WORK  Study de sig n:  Minimize risk o f b ia s using pro spe c tive de sig ns  Pre -spe c ifie d thre sho ld sc o re s  Mic ro b io lo g ic a lly de fine d c a se de finitio ns  Ge ne ra liza b ility:  Dive rse se tting s a nd po pula tio ns

WHO-CAD RE VI E W  I nc luding unpub lishe d studie s to inc re a se the study da ta b a se  Re vise d da ta e xtra c tio n fo rm  Pe rfo rming a me ta -a na lysis a s we ll P f i t l i ll

ACK NOWL E DGME NT S  Co -re se a rc he r:  F I ND:  Cha d Co he n  Dr. Cla udia De nking e r  Mc Gill Unive rsity:  Dr F  Dr. F a iz Ahma d K a iz Ahma d K ha n ha n  Dr. Ma dhuka r Pa i  Na z T Na z T o ra b i (me dic a l lib ra ria n) o ra b i (me dic a l lib ra ria n)

YOU! YOU! ONS? HANK HANK I ST QUE T T