about.md (11806B)
1 2 ## husband and father 3 4 Talented, beautiful people surround me. 5 The [now page](/now) 6 usually has the most current updates in this arena. 7 8 9 ## physician 10 11 Oncology fellow, Vanderbilt 2022-2025. 12 Internal medicine resident, Vanderbilt 2020-2022. 13 Medical school, Cleveland Clinic Lerner College of Medicine 14 of Case Western Reserve University 2015-2020. 15 16 I enjoy treating people with any cancer, 17 but have a particular affinity for 18 head-and-neck and upper GI cancers 19 (together, 2/3 of the "aerodigestive" cancers, the other 1/3 being lung). 20 There are many right ways to be a doctor, 21 but the way I understand my role is this: 22 to be with people in their suffering and their joy (suffering is everywhere, but there is a special joy known only in the oncology clinic), 23 to see and know the soul within the body (whatever the tumor or the scalpel or the radiation beam or the drug has done to that body), 24 to cure when I can (and deal with the fallout from that cure), 25 and help always (to the end and beyond). 26 27 There are glimmers of hope and serious advances in recent years, 28 but these remain horrible diseases, 29 many of which have few treatment options 30 once they've reached an advanced stage, 31 and the treatments we do have tend to be quite difficult to tolerate. 32 Most cancers can be disfiguring, 33 but these especially so, 34 whether from the tumors themselves or from the therapies. 35 Aerodigestive cancers also disproportionately affect 36 folks who live out in the country, 37 or in the cities but with few resources, 38 and these folks are my folks. 39 40 We need to help each other, and there is much to do. 41 42 43 ## educator 44 45 46 Harvard Macy Institute faculty, 2018-2020. 47 Health Care Education 2.0. 48 49 <https://vimbook.org>, 2020-present. 50 The Vanderbilt Internal Medicine Handbook 51 was started by Mike Neuss, MD/PhD 52 in the late 2010s when he was a resident. 53 It's an incredible resource, 54 primarily envisioned as a physical book 55 to keep in your white coat pocket 56 for quick, authoritative reference. 57 It had a website when I came to Vandy, 58 but the UX... left something to be desired. 59 I rebuilt it into its current state 60 (website and infrastructure only - each section has its own author(s)), 61 and help maintain the back end. 62 63 Chase Webber is the faculty support, and has been amazing. 64 65 It's used globally, 66 and one of my goals is to make it easier for smaller, 67 particularly international programs to have their own versions. 68 It uses only free and open-source software, 69 and fits comfortably into pretty much any free server 70 despite being a fairly large book with multimedia. 71 72 The main problem is that updating the content, 73 while it is just in Markdown (a very simple text format you can learn in ~7 minutes), 74 feels a little too tech-y to be comfortable for people who haven't spent time in a code text editor. 75 What it really needs is funding, 76 and a part-time developer with protected time. 77 78 79 ## data scientist 80 81 [Google Scholar profile](https://scholar.google.com/citations?user=Ng5AgXAAAAAJ) 82 83 *If you are interested data science consulting, [contact me](/contact). 84 Current rate is listed at that link.* 85 86 *Major projects are listed below in reverse chronological order (roughly), 87 newest projects at the top.* 88 89 90 ### overview 91 92 My work has morphed over the years, 93 as everyone's does, 94 but the consistent thread and drive throughout 95 has been on coaxing large, messy, complex data 96 to tell us a story about ourselves, 97 about all of us as societies and neighborhoods 98 as well as each one of us individually, 99 to empower us to speak our own sequels. 100 101 To translate that into buzzwords, 102 for your bingo game: 103 I'm an oncologist 104 who uses explainable artificial intelligence 105 among other techniques 106 to diagnose and address healthcare disparities, 107 including democratizing personalized medicine 108 and pursuing synergies in the 109 global academic-industrial complex. 110 111 I was trained as an anthropologist, 112 where reductionism is an insult, 113 at the same time that I was trained as a scientist, 114 where reductionism is the central conceit. 115 Despite the way I set up that last sentence, 116 there is no true conflict, 117 as the goal for most scientists and anthropologists is the same: 118 to make things at least a little better for someone, 119 but hopefully a lot better for everyone. 120 Modern anthropology is inherently activist, 121 far from the crusty sepia-toned image of a staid researcher 122 sitting just outside the village campfire furiously scribbling in a notebook, 123 and so is modern medical research 124 (we all read Tuskegee and are appalled and want to do better, 125 though precious few become [Paul Farmer](https://en.wikipedia.org/wiki/Paul_Farmer)). 126 127 Data science was an unexpected boon, 128 a set of tools that lets me deal with 129 staggering complexity 130 in a disciplined way, 131 to a degree unifying the 132 anthropologist's drive to let the data be itself (messy, human) 133 with the scientist's drive to simplify. 134 135 I became acquainted with using code to model the world 136 when I was doing physics research 137 (equivalent circuit modeling 138 and scanning laser doppler vibrometry 139 on a Nigerian-style clay pot drum, 140 total hoot, 141 I [presented](https://doi.org/10.1121/1.3654998) it 142 at the Acoustical Society of America Annual Meeting 143 and we published it in 144 [JASA](https://doi.org/10.1121/1.4789892)), 145 and the research-centric medical school at Cleveland Clinic 146 gave me the better part of five years 147 and a supportive environment 148 to dive in deeper to computer science, machine learning, etc. 149 150 Now that have the freedom to choose my own direction, 151 data science projects that were informed by anthropology 152 but had centroids in other disciplines have become 153 squarely within the overlap of my personal Venn diagram. 154 155 Additionally, as I have moved from academia to the worlds of private practice and industry, 156 I am increasingly inspired toward using these interests to hit cancer where it needs it the most: 157 in the clinic of the community oncologist, where the vast majority of cancer patients receive care, 158 by bringing timely insights from every relevant axis to bear on every visit. 159 160 [Information almost wants to be free](https://en.wikipedia.org/wiki/Information_wants_to_be_free), 161 and there is now so much of it, 162 that a lot can be done for free or at a low cost. 163 (I'm take special glee in byproducts, 164 "digital waste," 165 some say, 166 but there is gold in them there hills - 167 [Shigeru Ban](https://en.wikipedia.org/wiki/Shigeru_Ban) is my hero). 168 There are abstracts from conferences aplenty describing the cutting edge of everything, 169 usually seen as jots and tiddles on the march of science forward but forgotten when the journal article gets published, 170 and lab values and imaging results and clinic notes generated in the usual course of patient care, 171 used once or a few times and then forgotten and discarded in the wake of what comes next - 172 in all of this digital refuse I believe there are insights to be gained, efficiencies to be achieved, lives to be saved. 173 174 --- 175 176 177 ### disparities in cancer research across space and time 178 179 This ongoing project starts with a custom metadata archive 180 of all the abstracts from the two big annual cancer meetings 181 (ASH and ASCO - as a starting point), 182 made machine-readable and hence more easily accessible en masse. 183 Here's a [rough prelim](https://ash-abstracts.vercel.app/abstracts_small/abstracts) 184 of what that looks like, 185 with an interactive map and sorting/search/export 186 (you can hit JSON or CSV endpoints trivially to get nicely formatted data, 187 as opposed to the hulking web scrapers 188 and HTML wranglers I wrote to get the data the first time). 189 Analyses are planned on the authors and affiliations 190 (global mapping, with a time component), 191 and how those overlap with disease states 192 (e.g. how much sickle cell research comes out of Sub-Saharan Africa? 193 We know it's bad, but nobody has quantified how bad - 194 the idea is to have a platform for answering these questions easily), 195 before branching out into things like 196 conflicts of interest and various analyses using NLP, etc. 197 198 --- 199 200 201 ### Cleveland Clinic Center for Clinical Artificial Intelligence 202 ### explainable machine learning for readmissions, with a focus on sociodemographics 203 204 While at Cleveland Clinic, 205 I was one of the founding members of the 206 Center for Clinical Artificial Intelligence, 207 and its first dedicated analyst. 208 We had the delightful opportunity to work with 209 a rich dataset from one of the world's largest hospitals, 210 focused on predicting the risk of 211 readmission (discharging from the hospital and "bouncing back" too soon) 212 and extended length of stay. 213 We had access to not only health data, 214 but socioeconomic data, 215 for millions of patients, 216 and I further enriched this with census data. 217 218 With only an address and a date, 219 one can learn about a person's neighborhood 220 in incredible detail 221 (though not the person themselves - 222 the US Census is wise about privacy in the data they publish). 223 It turns out that a person's neighborhood 224 is a major predictor in their health outcomes, 225 as is their insurance provider, 226 in addition to a host of health parameters and hospital process clues. 227 228 I used interpretable machine learning techniques, 229 with a focus on revealing actionable items 230 that the patients and medical teams could hope to influence, 231 as well as to call out structural issues 232 that organizations and governments need to know. 233 One of the problems with machine learning 234 is that it can only ever restate 235 (and in some cases, as it feeds back on itself, more firmly entrench) 236 the biases that led to the historical data you fed to it. 237 If you design it to show you these biases, 238 show you *all* of its biases, explicitly, 239 to tell you exactly how much a person's race, ethnicity, gender, and neighborhood 240 played a role in its predictions (alongside medical diagnoses and lab values), 241 you have transformed an algorithm from a 242 potentially destructive tool, 243 a heinous thing that pushes hurting people farther down, 244 into a tool for positive change. 245 If these biases are shown in an easy-to-read visual, 246 that patients, clinicians, and administrators can all understand, 247 all the better. 248 See what it looks like [here](https://doi.org/10.1038/s41746-020-0249-z). 249 250 --- 251 252 253 ### Cleveland Clinic Taussig Cancer Institute 254 ### explainable machine learning for blood cancer diagnostics 255 256 Initially, when I was going to be a malignant hematologist 257 (I still think it is a beautiful field, but was drawn away to other pastures), 258 I wanted to level the playing field for advanced diagnostics. 259 260 A world-class hematopathologist together with a world-class clinician, 261 preferably a whole group of these together, 262 are required to make certain diagnoses, 263 and these only after at least one invasive biopsy. 264 What if we could mobilize the rich genetic and phenotypic data 265 available in simple blood samples and from the electronic medical record 266 to support diagnostics, and, eventually, democratize them? 267 Further, what if the answer the machine gives could be not only accurate, 268 but interpretable? 269 270 What if we could get the machine to explain itself and its reasoning, 271 both as a check against biological implausibility 272 (and more insidious problems such as systemic racism), 273 and to reveal more areas for research? 274 What if a high resolution bone marrow biopsy image 275 could be read in moments by a smart phone in 276 rural Arkansas/Kandahar/Mogadishu/Kushalnagar/etc., 277 instead of having to be shipped to 278 one of a handful of academic centers while the patient waits, 279 still sick, for an answer? 280 281 We made [some](https://doi.org/10.1182/blood-2019-126967) 282 [progress](https://doi.org/10.1182/bloodadvances.2021004755), 283 but these problems remain largely unsolved. 284 285 --- 286 287 288 ## internal links 289 290 If you want to know which tools I use, visit 291 [/uses](/uses). 292 293 If you are interested in the tech stack for this website, visit 294 [/colophon](/colophon). 295 296 If you would like to throw money at me, for whatever reason, visit 297 [/pay](/pay). 298 299