DNA breakthrough helps police trace ‘Beast of Bondi’

In Associate Professor Ewan Millar’s vision for the future of pathology in NSW, the skills, experience and expertise of Anatomical Pathologists is backed by powerful technology.

In the near future, Anatom­i­cal Pathol­o­gists – doc­tors who study body tis­sue and cells to diag­nose dis­eases like can­cer – will scru­ti­nise dig­i­tal slides rather than tra­di­tion­al glass slides to rou­tine­ly search for the small­est abnor­mal­i­ties in body tis­sue samples.

Ewan, a Senior Staff Spe­cial­ist Histopathol­o­gist with NSW Health Pathol­o­gy, based at St George Hos­pi­tal, believes it is like­ly Arti­fi­cial Intel­li­gence (AI) will take on the role of ‘Pathologist’s assis­tant’, pick­ing up on sub­tle pix­el pat­terns that flag tis­sue sam­ples for the doctor’s atten­tion and fur­ther inves­ti­ga­tion. Ulti­mate­ly, com­bin­ing the pathologist’s exper­tise and skill with advanced tech could rev­o­lu­tionise the diag­no­sis and under­stand­ing of dis­eases, and improve patient care and outcomes.

With enough devel­op­ment, the AI would also go one step fur­ther to iden­ti­fy bio­log­i­cal mark­ers, or ‘bio­mark­ers’, which indi­cate changes in tis­sue on a mol­e­c­u­lar lev­el, not vis­i­ble to the naked eye. These bio­mark­ers will not only help diag­nose dis­ease, such as can­cer, but help doc­tors under­stand a patient’s poten­tial response to tar­get­ed treatment.

It’s an ambi­tious vision, but one Ewan’s hap­py to chip away at thanks to a Researcher Exchange and Devel­op­ment with­in Indus­try (REDI) Fellowship.

Under the fel­low­ship, which aims to con­nect researchers with indus­try, Ewan is in the midst of a two-year, part-time project with Paige, a New York-based glob­al leader in Arti­fi­cial Intel­li­gence-based diag­nos­tic pathol­o­gy soft­ware. They mar­ket the first and only AI prod­uct to detect prostate can­cer approved by the US Food and Drug Admin­is­tra­tion (FDA). The AI is trained to iden­ti­fy sub­tle pix­el pat­terns in the tis­sue to accu­rate­ly pre­dict the pres­ence of prostate can­cer, and even map where in the prostate it is located.

Under the REDI Fel­low­ship, Ewan is assist­ing Paige to devel­op sim­i­lar AI prod­ucts for breast can­cer and breast can­cer lymph node spread. It is a dream role Ewan relishes.

“This REDI Fel­low­ship oppor­tu­ni­ty to work with Paige’s team of experts in AI is unri­valled,” Ewan said. “AI can real­ly trans­form the way pathol­o­gists work in the future, includ­ing right here in NSW.”

While the words ‘Arti­fi­cial Intel­li­gence’ can be anx­i­ety-induc­ing in med­ical cir­cles, Ewan explains AI is com­pli­men­ta­ry to tra­di­tion­al Anatom­i­cal Pathol­o­gy and will not replace human pathologists.

“It’s like hav­ing a sec­ond pair of eyes, or a good trainee pathol­o­gist by your side,” he explains.

“Pathol­o­gists can look at hun­dreds of slides a day. AI can run algo­rithms that dou­ble check the slides and can flag spe­cif­ic areas on the slides for the pathol­o­gist to con­cen­trate on. By mak­ing the work­flows more effi­cient, pathol­o­gists can con­cen­trate on the most mean­ing­ful parts of their role.”

While AI can over­come the chal­lenge of spot­ting abnor­mal­i­ties which are sim­ply too dif­fi­cult to detect with the naked eye, Ewan said it can also short-cut time-con­sum­ing, man­u­al tasks such as esti­mat­ing the loca­tion of small, dif­fi­cult to find tumours.

“Find­ing can­cer­ous tis­sue can be time-con­sum­ing,” Ewan admits. “Some work can be like look­ing for a nee­dle in a haystack, but the AI can be trained to recog­nise pix­el pat­terns. For exam­ple, pathol­o­gists may look at up to 75 slides from one patient’s prostate core biop­sies to find a tumour no more than 1mm in diam­e­ter. The AI is capa­ble of review­ing the slides with­in min­utes, com­pared to the time it takes to review slides manually.”

For now, Ewan is putting in the hard yards, train­ing the AI soft­ware to even­tu­al­ly give Anatom­i­cal Pathol­o­gists access to a pow­er­ful, accu­rate tool.

“We go through iter­a­tive cycles – train, test, review, work out what’s wrong, and repeat until the AI per­for­mance improves,” Ewan explains. “AI soft­ware must be close to 90 per cent accu­rate before it’s ready for fur­ther testing.”

While Ewan’s lend­ing his exper­tise to sev­er­al projects, most of his time is spent with the Bio­mark­er team devel­op­ing the next break­through in can­cer diagnostics.

From his home in Syd­ney, in the com­fort of his kids’ old gam­ing room, Ewan applies his pathol­o­gist eye to images used to train and val­i­date Paige’s bio­mark­er AI soft­ware. He reviews the dig­i­tal slides and the results that the AI algo­rithms pro­duce, then works close­ly with engi­neers to iden­ti­fy detec­tion errors and painstak­ing­ly improve the soft­ware per­for­mance slide by slide. Every can­cer type requires its own unique AI algo­rithm, which needs hun­dreds or thou­sands of slides to suc­cess­ful­ly train the software.

Using this same approach, Paige recent­ly built start-up algo­rithms capa­ble of detect­ing dig­i­tal bio­mark­ers for tumour muta­tions in breast and prostate can­cer. It’s hoped, in the future, dig­i­tal bio­mark­ers could be used to screen tumour slides for spe­cif­ic mol­e­c­u­lar fea­tures with­in just a few min­utes, to deter­mine a person’s response to tar­get­ed treat­ments. This new method for mon­i­tor­ing dis­ease tumour pro­gres­sion and treat­ment effi­ca­cy could poten­tial­ly replace some exist­ing expen­sive and time-con­sum­ing mol­e­c­u­lar tests.

The expe­ri­ence Ewan is gain­ing will pro­vide crit­i­cal knowl­edge to ensure dig­i­tal trans­for­ma­tion plan­ning, under­way at Anatom­i­cal Pathol­o­gy for NSW Health Pathol­o­gy, is AI-enabled for the future. It also pro­vides impor­tant indus­try expe­ri­ence to fur­ther sup­port his con­tri­bu­tion to AI researchers at UNSW Com­put­er Sci­ence and Engi­neer­ing, where sev­er­al PhD stu­dents are cur­rent­ly work­ing on breast can­cer pathol­o­gy AI projects.

Ewan’s REDI Fel­low­ship is sup­port­ed and fund­ed by MTP­Con­nect’s $32 mil­lion Researcher Exchange and Devel­op­ment with­in Indus­try (REDI) ini­tia­tive made pos­si­ble by the Med­ical Research Future Fund (MRFF), pro­vid­ing indus­try expe­ri­ences and skills devel­op­ment for ear­ly and mid-career researchers, clin­i­cians, and inno­va­tors to devel­op an indus­try ready work­force keep­ing pace with the demands of a rapid­ly chang­ing sector.