Appendix 7.1) In the UK the final cancer treatment decision is made at a Multidisciplinary Team Meeting (MDT). This includes surgeons, radiologists, pathologists and oncologists, both medical and radiotherapeutic, who review every aspect of each patient’s current data. The MDT decides on the treatment modalities to be used and the sequence in which they are deployed. This constitutes a treatment vector, hence the term Strategic Therapeutic Vector, STV. For example, the strategic decision the use chemotherapy then surgery then radiotherapy is represented by the vector CTY -> SGY -> RTY. There are over 20 such vectors built into the system but, additional vectors can be developed by the user as required and appropriate.
Immunotherapy is nominally included in the list of single agent treatment modalities but, at present no such treatments are routinely available. However, in designing a system it is easier to include extra “space” for something that will but needed in the future rather than adding that space at a later date to an already finished article. And, that day is not far in the future for immunotherapy. The analogy is drawn between designing the architecture of a building with an empty extra floor where the furniture can be added later, rather than having to build an extension to the finished building for the additional furniture.
Appendix 7.2) The charts itemized in iNOTZ Medical content under headings 4.13, 4.15.1 and 4.20 are customized for two of our β-test sites. Although some of these items will be common to all groups, we will individually customize these during set-up according to needs.
Appendix 7.3) Turing (1950) predicted that computers would have a capacity of 10**9 bits by the year 2000 and would be able to solve many machine intelligence problems. He argued that, irrespective of what happens within the “black box”, what really matters is the output. This has developed into the Turing Test for artificial intelligence which he called “The Imitation Game” and goes something like this.
If the output from a computer is placed before a judge who is not able to distinguish it from that produced by a human using the same input, then who can decide if the computer is or, is not, intelligent.
7.3.1) The iNOTZ “Imitation Game”
The following two paragraphs namely, HISTORY of PRESENT COMPLAINT and EXAMINATION FINDINGS were generated from just the menu items by our Digital Linguistics system. The items in bold were selected from a total of 27 menus and the unselected items are depicted in italics. The remaining text, including grammar, was generated by the system
HISTORY OF PRESENT COMPLAINT
This patient presented with symptoms that suggested a primary left breast tumor. She complained of malaise, lethargy, weight loss, a lump in the breast and a lump in the axilla but, she did not complain of changes to the skin, nipple alterations, altered sensation, changes in shape or altered consistency. The non-specific general symptoms of malaise, lethargy and weight loss had been present for about six months, more than six months and about twelve months respectively. The weight loss amounted to approximately 10 lbs: 4.54 kilos. The lump at the primary site had been present for about four months and the mass in the axilla had been noted for less than 1 month.
On examination a lump was found in the upper outer quadrant of the left breast that measured between 2.1 – 3.5 cm and on palpation this was hard, craggy and irregular but not indurated or inflammatory. The lump itself was painless but the surrounding breast was tender. The mass was fully mobile within a medium to large breast with no spread to local structures. On ipsilateral axillary examination there were 2 lymph nodes that were fully mobile with the larger estimated to be about 2.0 cms in diameter. There were no nodes in any of the other gland bearing areas and examination of the contralateral breast revealed no abnormalities. Finally, there was no purely clinical suspicion of metastatic disease.
The ICD10 code is C50.422, ICDo10 coding with the currently available data is C50.4-M~~~~/3/~ and the clinical TNM stage is T2 N1 M0 which translates to STAGE IIB.
Clearly, the above computer output passes the Turing test as this is indistinguishable from a dictated note that a human would produce. Moreover, the ICD codes and the TNM staging are generated automatically, in the background, from the relevant selected menu items. However, the ICDo10 code is not complete as the system has not yet reached the point in the narrative at which the tumor type and grade can be input. These missing data are represented by the tildes (~) that will be substituted later by the missing data when the relevant morphology menus are encountered.
7.3.2) The binary data record
The binary record of the primary data units (PDUs, see section 220.127.116.11 and 18.104.22.168) derived from the menu items that generated the text in 7.3.1 are shown below, where the bits set to 1 and 0 represent the selected and unselected menu items respectively.
7.3.3) Binary data structure
Some of the above data identify the record types as shown in the following schematic where the nomenclature is “borrowed” from DNA structure:
The promotor (yellow, housekeeping) occupies 16 bits and contains identification of the record, the number of words ahead, and the bit, at which the next promotor has to be inserted. The coding region (red) contains the bits that have to be converted to text. The linker region (blue) is a redundant sequence as there is not sufficient space to insert the 16 bits of the next promoter in the current word.
7.3.4) The promotor structure
The promotor is a composite 16-bit word containing three data items as shown below:
1) A total of 1023 records (blue, bits 0 through 9) can be accommodated starting at ZERO for present complaint
2) The next promotor (red, bits 10 through 14) is inserted at N words-ahead of the current position, a recursive problem (see below section 7.3.5). Because of the compact nature of the data this should not be more than 31 words ahead of “here”.
3) Bit 15 (yellow) is reserved for the bit in “word-ahead” at which the next promotor is inserted. If zero it starts at zero. If one it starts at bit 16.
7.3.5) The recursive problem
Recursion may be described as travelling to a destination without a map and you do not know the distance. Clearly, the distance can only be known on arrival.
The digital data are stored in an array of 32-bit words called bitsSet (see section 7.3.2). Each record type has a numerical value assigned to bits 0 through 9 in the promotor with bits 10 through 15 reserved, and left blank, for the word in which the next promotor is to be inserted.
The coding sequence of the first record always starts at bit 16 in bitsSet the positions identified by variables bitPtr and wrdPtr, (bit pointer and word pointer) , that are initially set to 16 and zero respectively.
The length of the record, in bits and words (bitPtr and wrdPtr), is determined from hidden menu items illustrated with the symptoms menu of section 22.214.171.124. The hidden item for this menu is [a1c21 (2H]
The number of menu-assigned bits is always located in position 2 (counting starts at character zero on the left). This is 12 for the above menu even though it contains only 10 symptoms which leaves 2 bits for any subsequent additional symptoms that need to be added. Every time a menu is encountered the bitPtr is incremented by the number of bits assigned. If this is greater than or equal to 32 the wrdPtr is incremented by one and the bitPtr is reset to bitPtr-32. Hence, the system keeps count of the record length in bits and words and these data are used to calculate and update the position of the next promotor in bits 10 through 15 that were “left blank” at the start of the current record.