CAD as CAD can

Rinckside 2011; 22,9: 17-18.

 wo years ago my banker told me, beaming with joy, that his bank had helped a medical company go public and that he had bought some shares for me. "I have seen what they are doing, it's just fantastic: computer-aided detection and computer-aided diagnosis."

He was hardly stoppable on the phone when he explained to me what CAD can do and what the newcomer to the stock exchange will do: assisting any kind of doctor or paramedic in the interpretation of digital images, including CT, MRI, and PET. Like a pro, he talked about image processing, artificial in­tel­li­gence, mam­mo­graphy, me­di­cal screening, and lung cancer. And they had something, some completely new technology, that nobody else had.

"After several years doing re­search in auto­mated in­for­ma­tion ex­trac­tion from MR images, others and I had come to the conclusion that such methods are unreliable in clinical routine. A break­through would be good …"

They must have invented something I didn't know about, I thought. I hadn't followed closely the development of CAD for 15 years. After several years doing research in automated information extraction from MR images, others and I had come to the conclusion that such methods are unreliable in clinical routine. A break­through would be good for the new com­pany — and hope­fully good for me too.

   Some nights ago, I stumbled across an interesting and entertaining software program on the Internet. It was also a kind of computer-assisted diagnosis, created by Dmi­try Chest­nykh, a 28-year-old Russian software programmer [1].

The algorithms of the program analyze the writing styles of people and ascribe them to an author. Works from 50 of the main authors of the English language have been fed into the system and can be compared to text samples. I tried it.

First, I typed in part of a column I had writ­ten last year. I received a diagnosis: "You write like H.P. Love­craft." I had never heard of Love­craft, but found out that he was an American author of fantastic and macabre shorts stories, and master of the Gothic tale of terror. It's an interesting description, and it challenged me.

Copy and paste is easy. So I pasted more columns into the "How do I write?" form. The re­spon­ses changed from Lovec­raft to Arthur C. Clarke, back and forth. Clarke was another science fiction writer. Then I copied and pasted two pages of Ernest Hemingway's prose from his story "A clean and well-lighted place." It was written like Ernest He­ming­way. I tried He­ming­way again; he was now writ­ing like James Joyce.

Joseph Conrad writes like Agatha Christie, Agatha Christie writes like Isaac Asimov or per­haps H.P. Love­craft. Henry Wads­worth Long­fellow's poem "A Psalm of Life" was writ­ten in the style of Char­les Dickens.

I got more curious and pasted one of my columns translated into Russian in Cy­ril­lic into the form. In Rus­sian, I write in the style of Doug­las Adams, who was British. Per­so­nal­ly, I don't believe that I write sci­ence fict­ion or hor­ror columns. Some of them might be pa­rab­les.

The computer-assisted diagnostic program for li­te­ra­ture works at random.

   CAD in mam­mo­graphy, however, does not work at random. In a study pub­li­shed some weeks ago, Dr. Joshua J. Fen­ton of the Uni­ver­si­ty of Ca­li­for­nia, Davis, and col­leagues, ana­lyzed data from more than 1.6 million film-screen­ing mam­mo­grams carried out between 1998 and 2006. They found: "CAD use during film-screen screen­ing mam­mo­graphy in the Uni­ted Sta­tes is asso­ciated with de­creas­ed spe­ci­fi­ci­ty but not with im­prove­ment in the de­tec­tion rate or pro­gnos­tic cha­ra­cte­ri­stics of in­va­sive breast cancer [2]."

An accompanying editorial published in the same issue of the Journal of the National Cancer Institute is worthwhile reading — for me it is the paper of the month [3]. It's "CAD as CAD can." The author, Dr. Donald A. Berry from the M.D. Anderson Cancer Center in Houston, sums up as follows:

"An argument for the use of CAD with film or digital mam­mo­grams is that it will get better over time. Fine. Re­searchers and devi­ce com­pa­nies should work to make the soft­ware ever better. But this should happen in an ex­peri­men­tal sett­ing and not while expos­ing mil­lions of women to a tech­no­logy that may be more harm­ful than it is bene­fi­cial."

   When I recently checked my share port­folio, the value of the CAD com­pany had dropped from 50 euros (around $70 U.S.) to 3 euros (around $4 U.S.). "Their pro­grams seem not exact­ly to work as thought," my ban­ker told me. "It's a write-off." Indeed.

References

1. Chestnykh D. I write like. http://iwl.me. Accessed 14 September 2011.
2. Fenton JJ, Abraham L, Taplin SH, et al. Effectiveness of computer-aided detection in community mammography practice. J Natl Cancer Inst. 2011; 103(15): 1152-1161.
3. Berry DA. Computer-assisted detection and screening mammography: Where's the beef? J Natl Cancer Inst. 2011; 103(15): 1139-1141.

Citation: Rinck PA. CAD as CAD can. Rinckside 2011; 22,9: 17-18.

A digest version of this column was published as:
Rinck PA. CAD as CAD can.
Aunt Minnie Europe. Maverinck. 19 September 2011.